Improving .NET Application Performance and Scalability

J.D. Meier, Srinath Vasireddy, Ashish Babbar, John Allen and Alex Mackman
Microsoft Corporation

May 2004

Related Links

Home Page for Improving .NET Application Performance and Scalability

Chapter 5 — Improving Managed Code Performance

Send feedback to Scale@microsoft.com

patterns & practices Library

Summary: This chapter provides common issues, design guidelines, and coding techniques to improve the performance of your ASP.NET pages and controls. This chapter provides a common language runtime thread pool tuning formula for reducing contention, threading guidelines, specific techniques to reduce page size, scalable solutions for managing state, data caching guidelines and much more.

Contents

Objectives
Overview
How to Use This Chapter
Architecture
Performance and Scalability Issues
Design Considerations
Implementation Considerations
Threading Explained
Threading Guidelines
Resource Management
Pages
Server Controls
Data Binding
Caching Explained
Caching Guidelines
State Management
Application State
Session State
View State
HTTP Modules
String Management
Exception Management
COM Interop
Data Access
Security Considerations
IIS 6.0 Considerations
Deployment Considerations
Summary
Additional Resources

Objectives

• Improve page response times.
• Design scalable Web applications.
• Use server controls efficiently.
• Use efficient caching strategies.
• Analyze and apply appropriate state management techniques.
• Minimize view state impact.
• Improve performance without impacting security.
• Minimize COM interop scalability issues.
• Optimize threading.
• Optimize resource management.
• Avoid common data binding mistakes.
• Use security settings to reduce server load.
• Avoid common deployment mistakes.

Overview

To build ASP.NET applications that meet your performance objectives, you need to understand the places where bottlenecks typically occur, the causes of the bottlenecks, and the steps to take to prevent the bottlenecks from occurring in your application. A combination of sound architecture and design, best practice coding techniques, and optimized platform and Microsoft® .NET Framework configuration is required. This chapter addresses each of these areas.

The chapter starts by examining the architecture of an ASP.NET application and then explains the anatomy of a Web request as it progresses through the HTTP and ASP.NET pipeline. The chapter explains the processing that occurs at each stage and identifies common performance bottlenecks. The chapter then provides a series of ASP.NET design guidelines. By following the guidelines in this section, you can help ensure that your top-level design does not create performance issues that can only be corrected by costly reengineering efforts. Finally, the chapter provides a series of sections that discusses top ASP.NET performance issues. These issues include page and control issues, caching, resource management, session and view state issues, threading, exception and string management, COM interop, and more.

How to Use This Chapter

Use this chapter to help improve the performance of your ASP.NET applications. You can apply the design considerations, coding techniques, and optimized platform and .NET Framework configuration information in this chapter to new and existing applications. To get the most out of this chapter, do the following:

• Jump to topics or read from beginning to end. The main headings in this chapter help you locate the topics that interest you. Alternatively, you can read the chapter from beginning to end to gain a thorough appreciation of performance and scalability design issues.
• Use the checklist. Use "Checklist: ASP.NET Performance" in the "Checklists" section of this guide to quickly view and evaluate the guidelines presented in this chapter.
• Know the ASP.NET runtime infrastructure. Understanding the runtime infrastructure can help you write code that is optimized for performance.
• Know the major performance and scalability issues. Read "Performance and Scalability Issues" in this chapter to learn about the major issues that affect the performance and scalability of your ASP.NET application. It is important to understand these key issues so that you can effectively identify performance and scalability problems and apply the recommendations presented in this chapter.
• Design with performance in mind. Read "Design Considerations" in this chapter to learn about best practice design guidelines.
• Use the "Architecture" section of this chapter. This section helps you understand how ASP.NET works. By understanding the architecture, you can make better design and implementation choices.
• Use the "Design Considerations" section of this chapter. This section helps you understand the high-level decisions that affect implementation choices for ASP.NET code.
• Read Chapter 13, "Code Review: .NET Application Performance." See the "ASP.NET" section for specific guidance.
• Measure your application performance. Read the "ASP.NET" and ".NET Framework Technologies" sections of Chapter 15, "Measuring .NET Application Performance" to learn about the key metrics that you can use to measure application performance. You have to measure application performance so that you can identify and resolve performance issues.
• Test your application performance. Read Chapter 16, "Testing .NET Application Performance" to learn how to apply performance testing to your application. It is important for you to apply a coherent testing process and to analyze the results.
• Tune your application performance. Read the "ASP.NET" section of Chapter 17, "Tuning .NET Application Performance" to learn how to resolve performance issues that you identify through the use of tuning metrics.

Architecture

ASP.NET requires a host. On Windows Server™ 2003, the default host is the Internet Information Services (IIS) 6.0 worker process (W3wp.exe). When you use the ASP.NET Process Model, the host is the ASP.NET worker process (Aspnet_wp.exe).

When a request is received by ASP.NET, the request is handled by the HttpRuntime object. The HttpRuntime is responsible for application creation and initialization, managing the request queue and thread pool, and dispatching the incoming requests to the correct application. After the request is dispatched to the appropriate application, the request is passed through a pipeline. This pipeline is a staged, event-based execution framework consisting of multiple HttpModule objects and a single HttpHandler object. This architecture is shown in Figure 6.1.

Figure 6.1: ASP.NET runtime infrastructure

HttpModule objects participate in the pipeline by handling predefined events that ASP.NET exposes. These events include BeginRequest,AuthenticateRequest, and EndRequest. The request flows through the pipeline of HttpModule objects and is then run by a single HttpHandler. After the event handler is completed, the request then flows back through the pipeline and is sent to the client.

Throughout the entire lifetime of a request, a context is exposed. The HttpContext object encapsulates information about individual requests and their associated responses.

Performance and Scalability Issues

The main issues that can adversely affect the performance and scalability of your ASP.NET application are summarized below. Subsequent sections in this chapter provide strategies and technical information to prevent or resolve each of these issues.

• Resource affinity. Resource affinity can prevent you from adding more servers, or resource affinity can reduce the benefits of adding more CPUs and memory. Resource affinity occurs when code needs a specific thread, CPU, component instance, or server.
• Excessive allocations. Applications that allocate memory excessively on a per-request basis consume memory and create additional work for garbage collection. The additional garbage collection work increases CPU utilization. These excessive allocations may be caused by temporary allocations. For example, the excessive allocations may be caused by excessive string concatenation that uses the += operator in a tight loop.
• Failure to share expensive resources. Failing to call the Dispose or Close method to release expensive resources, such as database connections, may lead to resource shortages. Closing or disposing resources permits the resources to be reused more efficiently.
• Blocking operations. The single thread that handles an ASP.NET request is blocked from servicing additional user requests while the thread is waiting for a downstream call to return. Calls to long-running stored procedures and remote objects may block a thread for a significant amount of time.
• Misusing threads. Creating threads for each request incurs thread initialization costs that can be avoided. Also, using single-threaded apartment (STA) COM objects incorrectly may cause multiple requests to queue up. Multiple requests in the queue slow performance and create scalability issues.
• Making late-bound calls. Late-bound calls require extra instructions at runtime to identify and load the code to be run. Whether the target code is managed or unmanaged, you should avoid these extra instructions.
• Misusing COM interop. COM interop is generally very efficient, although many factors affect its performance. These factors include the size and type of the parameters that you pass across the managed/unmanaged boundary and crossing apartment boundaries. Crossing apartment boundaries may require expensive thread switches.
• Large pages. Page size is affected by the number and the types of controls on the page. Page size is also affected by the data and images that you use to render the page. The more data you send over the network, the more bandwidth you consume. When you consume high levels of bandwidth, you are more likely to create a bottleneck.
• Failure to use data caching appropriately. Failure to cache static data, caching too much data so that the items get flushed out, caching user data instead of application-wide data, and caching infrequently used items may limit your system's performance and scalability.
• Failure to use output caching appropriately. If you do not use output caching or if you use it incorrectly, you can add avoidable strain to your Web server.
• Inefficient rendering. Interspersing HTML and server code, performing unnecessary initialization code on page postback, and late-bound data binding may all cause significant rendering overhead. This may decrease the perceived and true page performance.

Design Considerations

Building high-performance ASP.NET applications is significantly easier if you design with performance in mind. Make sure you develop a performance plan from the outset of your project. Never try to add performance as a post-build step. Also, use an iterative development process that incorporates constant measuring between iterations.

By following best practice design guidelines, you significantly increase your chances of creating a high-performance Web application. Consider the following design guidelines:

• Consider security and performance.
• Partition your application logically.
• Evaluate affinity.
• Reduce round trips.
• Avoid blocking on long-running tasks.
• Use caching.
• Avoid unnecessary exceptions.

Consider Security and Performance

Your choice of authentication scheme can affect the performance and scalability of your application. You need to consider the following issues:

• Identities. Consider the identities you are using and the way that you flow identity through your application. To access downstream resources, you can use the ASP.NET process identity or another specific service identity. Or, you can enable impersonation and flow the identity of the original caller. If you connect to Microsoft SQL Server™, you can also use SQL authentication. However, SQL authentication requires you to store credentials in the database connection string. Storing credentials in the database connection string is not recommended from a security perspective. When you connect to a shared resource, such as a database, by using a single identity, you benefit from connection pooling. Connection pooling significantly increases scalability. If you flow the identity of the original caller by using impersonation, you cannot benefit from efficient connection pooling, and you have to configure access control for multiple individual user accounts. For these reasons, it is best to use a single trusted identity to connect to downstream databases.
• Managing credentials. Consider the way that you manage credentials. You have to decide if your application stores and verifies credentials in a database, or if you want to use an authentication mechanism provided by the operating system where credentials are stored for you in the Active Directory® directory service.

  You should also determine the number of concurrent users that your application can support and determine the number of users that your credential store (database or Active Directory) can handle. You should perform capacity planning for your application to determine if the system can handle the anticipated load.

• Protecting credentials. Your decision to encrypt and decrypt credentials when they are sent over the network costs additional processing cycles. If you use authentication schemes such as Windows® Forms authentication or SQL authentication, credentials flow in clear text and can be accessed by network eavesdroppers. In these cases, how important is it for you to protect them as they are passed across the network? Decide if you can choose authentication schemes that are provided by the operating system, such as NTLM or the Kerberos protocol, where credentials are not sent over the network to avoid encryption overhead.
• Cryptography. If your application only needs to ensure that information is not tampered with during transit, you can use keyed hashing. Encryption is not required in this case, and it is relatively expensive compared to hashing. If you need to hide the data that you send over the network, you require encryption and probably keyed hashing to ensure data validity. When both parties can share the keys, using symmetric encryption provides improved performance in comparison to asymmetric encryption. Although larger key sizes provide greater encryption strength, performance is slower relative to smaller key sizes. You must consider this type of performance and balance the larger key sizes against security tradeoffs at design time.

More Information

For more information, see "Performance Comparison: Security Design Choices" on MSDN at http://msdn.microsoft.com/en-us/library/ms978415.aspx.

Partition Your Application Logically

Use layering to logically partition your application logic into presentation, business, and data access layers. This helps you create maintainable code, but it also permits you to monitor and optimize the performance of each layer separately. A clear logical separation also offers more choices for scaling your application. Try to reduce the amount of code in your code-behind files to improve maintenance and scalability.

Do not confuse logical partitioning with physical deployment. A logical separation enables you to decide whether to locate presentation and business logic on the same server and clone the logic across servers in a Web farm, or to decide to install the logic on servers that are physically separate. The key point to remember is that remote calls incur a latency cost, and that latency increases as the distance between the layers increases.

For example, in-process calls are the quickest calls, followed by cross-process calls on the same computer, followed by remote network calls. If possible, try to keep the logical partitions close to each other. For optimum performance you should place your business and data access logic in the Bin directory of your application on the Web server.

For more information about these and other deployment issues, see "Deployment Considerations" later in this chapter.

Evaluate Affinity

Affinity can improve performance. However, affinity may affect your ability to scale. Common coding practices that introduce resource affinity include the following:

• Using in-process session state. To avoid server affinity, maintain ASP.NET session state out of process in a SQL Server database or use the out-of-process state service running on a remote machine. Alternatively, design a stateless application, or store state on the client and pass it with each request.
• Using computer-specific encryption keys. Using computer-specific encryption keys to encrypt data in a database prevents your application from working in a Web farm because common encrypted data needs to be accessed by multiple Web servers. A better approach is to use computer-specific keys to encrypt a shared symmetric key. You use the shared symmetric key to store encrypted data in the database.

More Information

For more information about how to encrypt and decrypt data in a shared database, without introducing affinity, see Chapter 14, "Building Secure Data Access," in Improving Web Application Security: Threats and Countermeasures on MSDN at http://msdn.microsoft.com/en-us/library/ms994921.aspx.

Reduce Round Trips

Use the following techniques and features in ASP.NET to minimize the number of round trips between a Web server and a browser, and between a Web server and a downstream system:

• HttpResponse.IsClientConnected. Consider using the HttpResponse.IsClientConnected property to verify if the client is still connected before processing a request and performing expensive server-side operations. However, this call may need to go out of process on IIS 5.0 and can be very expensive. If you use it, measure whether it actually benefits your scenario.
• Caching. If your application is fetching, transforming, and rendering data that is static or nearly static, you can avoid redundant hits by using caching.
• Output buffering. Reduce roundtrips when possible by buffering your output. This approach batches work on the server and avoids chatty communication with the client. The downside is that the client does not see any rendering of the page until it is complete. You can use theResponse.Flush method. This method sends output up to that point to the client. Note that clients that connect over slow networks where buffering is turned off, affect the response time of your server. The response time of your server is affected because your server needs to wait for acknowledgements from the client. The acknowledgements from the client occur after the client receives all the content from the server.
• Server.Transfer. Where possible, use the Server.Transfer method instead of the Response.Redirect method. Response.Redirect sends a response header to the client that causes the client to send a new request to the redirected server by using the new URL. Server.Transfer avoids this level of indirection by simply making a server-side call.

  You cannot always just replace Response.Redirect calls with Server.Transfer calls because Server.Transfer uses a new handler during the handler phase of request processing. If you need authentication and authorization checks during redirection, use Response.Redirect instead ofServer.Transfer because the two mechanisms are not equivalent. When you use Response.Redirect, ensure you use the overloaded method that accepts a Boolean second parameter, and pass a value of false to ensure an internal exception is not raised.

  Also note that you can only use Server.Transfer to transfer control to pages in the same application. To transfer to pages in other applications, you must use Response.Redirect.

More Information

For more information, see Knowledge Base article 312629, "PRB: ThreadAbortException Occurs If You Use Response.End, Response.Redirect, or Server.Transfer," at http://support.microsoft.com/default.aspx?scid=kb;en-us;312629.

Avoid Blocking on Long-Running Tasks

If you run long-running or blocking operations, consider using the following asynchronous mechanisms to free the Web server to process other incoming requests:

• Use asynchronous calls to invoke Web services or remote objects when there is an opportunity to perform additional parallel processing while the Web service call proceeds. Where possible, avoid synchronous (blocking) calls to Web services because outgoing Web service calls are made by using threads from the ASP.NET thread pool. Blocking calls reduce the number of available threads for processing other incoming requests.

  For more information, see "Avoid Asynchronous Calls Unless You Have Additional Parallel Work" later in this chapter.

• Consider using the OneWay attribute on Web methods or remote object methods if you do not need a response. This "fire and forget" model allows the Web server to make the call and continue processing immediately. This choice may be an appropriate design choice for some scenarios.
• Queue work, and then poll for completion from the client. This permits the Web server to invoke code and then let the Web client poll the server to confirm that the work is complete.

More Information

For more information about how to implement these mechanisms, see "Threading Guidelines" later in this chapter.

Use Caching

A well-designed caching strategy is probably the single most important performance-related design consideration. ASP.NET caching features include output caching, partial page caching, and the cache API. Design your application to take advantage of these features.

Caching can be used to reduce the cost of data access and rendering output. Knowing how your pages use or render data enables you to design efficient caching strategies. Caching is particularly useful when your Web application constantly relies on data from remote resources such as databases, Web services, remote application servers, and other remote resources. Applications that are database intensive may benefit from caching by reducing the load on the database and by increasing the throughput of the application. As a general rule, if caching is cheaper than the equivalent processing, you should use caching. Consider the following when you design for caching:

• Identify data or output that is expensive to create or retrieve. Caching data or output that is expensive to create or retrieve can reduce the costs of obtaining the data. Caching the data reduces the load on your database server.
• Evaluate the volatility. For caching to be effective, the data or output should be static or infrequently modified. Lists of countries, states, or zip codes are some simple examples of the type of data that you might want to cache. Data or output that changes frequently is usually less suited to caching but can be manageable, depending upon the need. Caching user data is typically only recommended when you use specialized caches, such as the ASP.NET session state store.
• Evaluate the frequency of use. Caching data or output that is frequently used can provide significant performance and scalability benefits. You can obtain performance and scalability benefits when you cache static or frequently modified data and output alike. For example, frequently used, expensive data that is modified on a periodic basis may still provide large performance and scalability improvements when managed correctly. If the data is used more often than it is updated, the data is a candidate for caching.
• Separate volatile data from nonvolatile data. Design user controls to encapsulate static content such as navigational aids or help systems, and keep them separate from more volatile data. This permits them to be cached. Caching this data decreases the load on your server.
• Choose the right caching mechanism. There are many different ways to cache data. Depending on your scenario, some are better than others. User-specific data is typically stored in the Session object. Static pages and some types of dynamic pages such as non-personalized pages that are served to large user sets can be cached by using the ASP.NET output cache and response caching. Static content in pages can be cached by using a combination of the output cache and user controls. The ASP.NET caching features provide a built-in mechanism to update the cache. Application state, session state, and other caching means do not provide a built-in mechanism to update the cache.

Avoid Unnecessary Exceptions

Exceptions add significant overhead to your application. Do not use exceptions to control logic flow, and design your code to avoid exceptions where possible. For example, validate user input, and check for known conditions that can cause exceptions. Also, design your code to fail early to avoid unnecessary processing.

If your application does not handle an exception, it propagates up the stack and is ultimately handled by the ASP.NET exception handler. When you design your exception handling strategy, consider the following:

• Design code to avoid exceptions. Validate user input and check for known conditions that can cause exceptions. Design code to avoid exceptions.
• Avoid using exceptions to control logic flow. Avoid using exception management to control regular application logic flow.
• Avoid relying on global handlers for all exceptions. Exceptions cause the runtime to manipulate and walk the stack. The further the runtime traverses the stack searching for an exception handler, the more expensive the exception is to process.
• Catch and handle exceptions close to where they occur. When possible, catch and handle exceptions close to where they occur. This avoids excessive and expensive stack traversal and manipulation.
• Do not catch exceptions you cannot handle. If your code cannot handle an exception, use a try/finally block to ensure that you close resources, regardless of whether an exception occurs. When you use a try/finally block, your resources are cleaned up in the finally block if an exception occurs, and the exception is permitted to propagate up to an appropriate handler.
• Fail early to avoid expensive work. Design your code to avoid expensive or long-running work if a dependent task fails.
• Log exception details for administrators. Implement an exception logging mechanism that captures detailed information about exceptions so that administrators and developers can identify and remedy any issues.
• Avoid showing too much exception detail to users. Avoid displaying detailed exception information to users, to help maintain security and to reduce the amount of data that is sent to the client.

Implementation Considerations

When you move from application design to application development, consider the technical details of your ASP.NET application. Key ASP.NET performance measures include response times, speed of throughput, and resource management.

You can improve response times by reducing page sizes, reducing your reliance on server controls, and using buffering to reduce chatty communication with the client. You can avoid unnecessary work by caching resources.

Throughput can be improved by making effective use of threads. Tune the thread pool to reduce connections, and to avoid blocking threads because blocking threads reduce the number of available worker threads.

Poor resource management can place excessive loads on server CPU and memory. You can improve resource utilization by effectively using pooled resources, by explicitly closing or disposing resources you open, and by using efficient string management.

When you follow best practice implementation guidelines, you increase the performance of your application by using well-engineered code and a well-configured application platform. The following sections describe performance considerations for ASP.NET features and scenarios.

Threading Explained

ASP.NET processes requests by using threads from the .NET thread pool. The thread pool maintains a pool of threads that have already incurred the thread initialization costs. Therefore, these threads are easy to reuse. The .NET thread pool is also self-tuning. It monitors CPU and other resource utilization, and it adds new threads or trims the thread pool size as needed. You should generally avoid creating threads manually to perform work. Instead, use threads from the thread pool. At the same time, it is important to ensure that your application does not perform lengthy blocking operations that could quickly lead to thread pool starvation and rejected HTTP requests.

Formula for Reducing Contention

The formula for reducing contention can give you a good empirical start for tuning the ASP.NET thread pool. Consider using the Microsoft product group-recommended settings that are shown in Table 6.1 if the following conditions are true:

• You have available CPU.
• Your application performs I/O bound operations such as calling a Web method or accessing the file system.
• The ASP.NET Applications/Requests In Application Queue performance counter indicates that you have queued requests.

Table 6.1: Recommended Threading Settings for Reducing Contention

To address this issue, you need to configure the following items in the Machine.config file. Apply the recommended changes that are described in the following section, across the settings and not in isolation. For a detailed description of each of these settings, see "Thread Pool Attributes" in Chapter 17, "Tuning .NET Application Performance."

• Set maxconnection to 12 * # of CPUs. This setting controls the maximum number of outgoing HTTP connections that you can initiate from a client. In this case, ASP.NET is the client. Set maxconnection to 12 * # of CPUs.
• Set maxIoThreads to 100. This setting controls the maximum number of I/O threads in the .NET thread pool. This number is automatically multiplied by the number of available CPUs. Set maxloThreads to 100.
• Set maxWorkerThreads to 100. This setting controls the maximum number of worker threads in the thread pool. This number is then automatically multiplied by the number of available CPUs. Set maxWorkerThreads to 100.
• Set minFreeThreads to 88 * # of CPUs. This setting is used by the worker process to queue all the incoming requests if the number of available threads in the thread pool falls below the value for this setting. This setting effectively limits the number of requests that can run concurrently tomaxWorkerThreads – minFreeThreads. Set minFreeThreads to 88 * # of CPUs. This limits the number of concurrent requests to 12 (assumingmaxWorkerThreads is 100).
• Set minLocalRequestFreeThreads to 76 * # of CPUs. This setting is used by the worker process to queue requests from localhost (where a Web application sends requests to a local Web service) if the number of available threads in the thread pool falls below this number. This setting is similar to minFreeThreads but it only applies to localhost requests from the local computer. Set minLocalRequestFreeThreads to 76 * # of CPUs.

Note   The recommendations that are provided in this section are not rules. They are a starting point. Test to determine the appropriate settings for your scenario. If you move your application to a new computer, ensure that you recalculate and reconfigure the settings based on the number of CPUs in the new computer.

If your ASPX Web page makes multiple calls to Web services on a per-request basis, apply the recommendations.

The recommendation to limit the ASP.NET runtime to 12 threads for handling incoming requests is most applicable for quick-running operations. The limit also reduces the number of context switches. If your application makes long-running calls, first consider the design alternatives presented in the "Avoid Blocking on Long-Running Tasks" section. If the alternative designs cannot be applied in your scenario, start with 100 maxWorkerThreads, and keep the defaults for minFreeThreads. This ensures that requests are not serialized in this particular scenario. Next, if you see high CPU utilization and context-switching when you test your application, test by reducing maxWorkerThreads or by increasing minFreeThreads.

The following occurs if the formula has worked:

• CPU utilization increases.
• Throughput increases according to the ASP.NET Applications\Requests/Sec performance counter.
• Requests in the application queue decrease according to the ASP.NET Applications/Requests In Application Queue performance counter.

If using the recommended settings does not improve your application performance, you may have a CPU bound scenario. By adding more threads you increase thread context switching. For more information, see "ASP.NET Tuning" in Chapter 17, "Tuning .NET Application Performance."

More Information

For more information, see Knowledge Base article 821268, "PRB: Contention, Poor Performance, and Deadlocks When You Make Web Service Requests from ASP.NET Applications," at http://support.microsoft.com/default.aspx?scid=kb;en-us;821268.

Threading Guidelines

This section discusses guidelines that you can use to help improve threading efficiency in ASP.NET. The guidelines include the following:

• Tune the thread pool by using the formula to reduce contention.
• Consider minIoThreads and minWorkerThreads for burst load.
• Do not create threads on a per-request basis.
• Avoid blocking threads.
• Avoid asynchronous calls unless you have additional parallel work.

Tune the Thread Pool by Using the Formula to Reduce Contention

If you have available CPU and if requests are queued, configure the ASP.NET thread pool. For more information about how to do this, see "Formula for Reducing Contention" in the preceding "Threading Explained" section. The recommendations in "Threading Explained" are a starting point.

When your application uses the common language runtime (CLR) thread pool, it is important to tune the thread pool correctly. Otherwise, you may experience contention issues, performance problems, or possible deadlocks. Your application may be using the CLR thread pool if the following conditions are true:

• Your application makes Web service calls.
• Your application uses the WebRequest or HttpWebRequest classes to make outgoing Web requests.
• Your application explicitly queues work to the thread pool by calling the QueueUserWorkItem method.

More Information

For more information, see Knowledge Base article 821268, "PRB: Contention, Poor Performance, and Deadlocks When You Make Web Service Requests from ASP.NET Applications," at http://support.microsoft.com/default.aspx?scid=kb;en-us;821268.

Consider minIoThreads and minWorkerThreads for Burst Load

If your application experiences burst loads where there are prolonged periods of inactivity between the burst loads, the thread pool may not have enough time to reach the optimal level of threads. A burst load occurs when a large number of users connect to your application suddenly and at the same time. The minIoThreads and minWorkerThreads settings enable you to configure a minimum number of worker threads and I/O threads for load conditions.

At the time of this writing, you need a supported fix to configure these settings. For more information, see the following Knowledge Base articles:

• 810259, "FIX: SetMinThreads and GetMinThreads API Added to Common Language Runtime ThreadPool Class," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;810259
• 827419, "PRB: Sudden Requirement for a Larger Number of Threads from the ThreadPool Class May Result in Slow Computer Response Time," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;827419

Do Not Create Threads on a Per-Request Basis

Creating threads is an expensive operation that requires initialization of both managed and unmanaged resources. You should avoid manually creating threads on each client request for server-based applications such as ASP.NET applications and Web services.

Consider using asynchronous calls if you have work that is not CPU bound that can run in parallel with the call. For example, this might include disk I/O bound or network I/O bound operations such as reading or writing files, or making calls to another Web method.

You can use the infrastructure provided by the .NET Framework to perform asynchronous operations by calling the Beginsynchronous andEndsynchronous methods (where synchronous represents the synchronous method name). If this asynchronous calling pattern is not an option, then consider using threads from the CLR thread pool. The following code fragment shows how you queue a method to run on a separate thread from the thread pool.

WaitCallback methodTarget = new WaitCallback(myClass.UpdateCache);

bool isQueued = ThreadPool.QueueUserWorkItem(methodTarget);

Avoid Blocking Threads

Any operation that you perform from an ASP.NET page that causes the current request thread to block means that one less worker thread from the thread pool is available to service other ASP.NET requests. Avoid blocking threads.

Avoid Asynchronous Calls Unless You Have Additional Parallel Work

Make asynchronous calls from your Web application only when your application has additional parallel work to perform while it waits for the completion of the asynchronous calls, and the work performed by the asynchronous call is not CPU bound. Internally, the asynchronous calls use a worker thread from the thread pool; in effect, you are using additional threads.

At the same time that you make asynchronous I/O calls, such as calling a Web method or performing file operations, the thread that makes the call is released so that it can perform additional work, such as making other asynchronous calls or performing other parallel tasks. You can then wait for completion of all of those tasks. Making several asynchronous calls that are not CPU bound and then letting them run simultaneously can improve throughput.

More Information

For more information about ASP.NET threading and asynchronous communication, see "ASP.NET Pipeline: Use Threads and Build Asynchronous Handlers in Your Server-Side Web Code" at http://msdn.microsoft.com/en-us/magazine/cc164128.aspx.

Resource Management

Poor resource management from pages and controls is one of the primary causes of poor Web application performance. Poor resource management can place excessive loads on CPUs and can consume vast amounts of memory. When CPU or memory thresholds are exceeded, applications might be recycled or blocked until the load on the server is lower. For more information, see "Resource Management" in Chapter 3, "Design Guidelines for Application Performance." Use the following guidelines to help you manage your resources efficiently:

• Pool resources.
• Explicitly call Dispose or Close on resources you open.
• Do not cache or block on pooled resources.
• Know your application allocation pattern.
• Obtain resources late and release them early.
• Avoid per-request impersonation.

Pool Resources

ADO.NET provides built-in database connection pooling that is fully automatic and requires no specific coding. Make sure that you use the same connection string for every request to access the database.

Make sure you release pooled resources so that they can be returned to the pool as soon as possible. Do not cache pooled resources or make lengthy blocking calls while you own the pooled resource, because this means that other clients cannot use the resource in the meantime. Also, avoid holding objects across multiple requests.

Explicitly Call Dispose or Close on Resources You Open

If you use objects that implement the IDisposable interface, make sure you call the Dispose method of the object or the Close method if one is provided. Failing to call Close or Dispose prolongs the life of the object in memory long after the client stops using it. This defers the cleanup and can contribute to memory pressure. Database connection and files are examples of shared resources that should be explicitly closed. The finally clause of the try/finally block is a good place to ensure that the Close or Dispose method of the object is called. This technique is shown in the following Visual Basic® .NET code fragment.

Try

  conn.Open()

…Finally

  If Not(conn Is Nothing) Then

    conn.Close()

  End If

End Try

In Visual C#®, you can wrap resources that should be disposed, by using a using block. When the using block completes, Dispose is called on the object listed in the brackets on the using statement. The following code fragment shows how you can wrap resources that should be disposed by using a using block.

SqlConnection conn = new SqlConnection(connString);

using (conn)

{

  conn.Open();

  . . .

} // Dispose is automatically called on the connection object conn here.

More Information

For more information, see "Finalize and Dispose Guidelines" in Chapter 5, "Improving Managed Code Performance." Also, see "Explicitly Close Connections" in Chapter 12, "Improving ADO.NET Performance."

Do Not Cache or Block on Pooled Resources

If your application uses resources that are pooled, release the resource back to the pool. Caching the pooled resources or making blocking calls from a pooled resource reduces the availability of the pooled resource for other users. Pooled resources include database connections, network connections, and Enterprise Services pooled objects.

Know Your Application Allocation Pattern

Poor memory allocation patterns may cause the garbage collector to spend most of its time collecting objects from Generation 2. Collecting objects from Generation 2 leads to poor application performance and high loads on the CPU.

Coding techniques that cause large numbers of temporary allocations during a short interval put pressure on the garbage collector. For example, when you perform a large number of string concatenation operations by using the += operator in a tight loop, or when you use String.Split for every request, you may put pressure on the garbage collector. All of these operations create hidden objects (temporary allocations). Use tools such as the CLR Profiler and System Monitor to better understand allocation patterns in your application.

More Information

For more information, see "How To: Use CLR Profiler" in the "How To" section of this guide. Also, see "CLR and Managed Code" in Chapter 15, "Measuring .NET Application Performance."

For more information about the mechanics of garbage collection and generations, see "Garbage Collection Explained" in Chapter 5, "Improving Managed Code Performance."

Obtain Resources Late and Release Them Early

Open critical, limited, and shared resources just before you need them, and release them as soon as you can. Critical, limited, and shared resources include resources such as database connections, network connections, and transactions.

Avoid Per-Request Impersonation

Identify and, if necessary, authorize the caller at the Web server. Obtain access to system resources or application-wide resources by using the identity of the Web application process or by using a fixed service account. System resources are resources such as event logs. Application-wide resources are resources such as databases. Avoiding per-request impersonation minimizes security overhead and maximizes resource pooling.

Note   Impersonation on its own does not cause performance issues. However, impersonation often prevents efficient resource pooling. This is a common cause of performance and scalability problems.

Pages

The efficiency of your ASP.NET page and code-behind page logic plays a large part in determining the overall performance of your Web application. The following guidelines relate to the development of individual .aspx and .ascx Web page files.

• Trim your page size.
• Enable buffering.
• Use Page.IsPostBack to minimize redundant processing.
• Partition page content to improve caching efficiency and reduce rendering.
• Ensure pages are batch compiled.
• Ensure debug is set to false.
• Optimize expensive loops.
• Consider using Server.Transfer instead of Response.Redirect.
• Use client-side validation.

Trim Your Page Size

Processing large page sizes increases the load on the CPU, increases the consumption of network bandwidth, and increases the response times for clients. Avoid designing and developing large pages that accomplish multiple tasks, particularly where only a few tasks are normally executed for each request. Where possible logically partition your pages.

To trim your page size, you can do one or all of the following:

• Use script includes for any static scripts in your page to enable the client to cache these scripts for subsequent requests. The following script element shows how to do this.

  <script language=jscript src="scripts\myscript.js">
  

• Remove characters such as tabs and spaces that create white space before you send a response to the client. Removing white spaces can dramatically reduce the size of your pages. The following sample table contains white spaces.

  // with white space
  
  <table>
  
    <tr>
  
      <td>hello</td>
  
      <td>world</td>
  
     </tr>
  
  </table>
  

  The following sample table does not contain white spaces.

  // without white space
  
  <table>
  
  <tr><td>hello</td><td>world</td></tr>
  
  </table>
  

Save these two tables in separate text files by using Notepad, and then view the size of each file. The second table saves several bytes simply by removing the white space. If you had a table with 1,000 rows, you could reduce the response time by just removing the white spaces. In intranet scenarios, removing white space may not represent a huge saving. However, in an Internet scenario that involves slow clients, removing white space can increase response times dramatically. You can also consider HTTP compression; however, HTTP compression affects CPU utilization.

You cannot always expect to design your pages in this way. Therefore, the most effective method for removing the white space is to use an Internet Server API (ISAPI) filter or an HttpModule object. An ISAPI filter is faster than an HttpModule; however, the ISAPI filter is more complex to develop and increases CPU utilization. You might also consider IIS compression. IIS compression can be added by using a metabase entry.

Additionally, you can trim page size in the following ways:

• Disable view state when you do not need it. For more information, see "View State" later in this chapter.
• Limit the use of graphics, and consider using compressed graphics.
• Consider using cascading style sheets to avoid sending the same formatting directives to the client repeatedly.
• Avoid long control names; especially ones that are repeated in a DataGrid or Repeater control. Control names are used to generate unique HTML ID names. A 10-character control name can easily turn into 30 to 40 characters when it is used inside nested controls that are repeated.

Note   When using the ASP.NET process model, the ASP.NET worker process sends responses back to the client, it first sends them through IIS in 31-kilobyte (KB) chunks. This applies to .NET Framework 1.1, but it could change in future versions. The more 31-KB chunks that ASP.NET has to send through IIS, the slower your page runs. You can determine how many chunks ASP.NET requires for your page by browsing the page, viewing the source, and then saving the file to disk. To determine the number of chunks, divide the page size by 31.

More Information

For more information about IIS compression, see Knowledge Base article, 322603, "HOW TO: Enable ASPX Compression in IIS," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;322603.

Enable Buffering

Because buffering is enabled by default, ASP.NET batches work on the server and avoid chatty communication with the client. The disadvantage to this approach is that for a slow page, the client does not see any rendering of the page until it is complete. You can use Response.Flush to mitigate this situation because Response.Flush sends output up to that point to the client. Clients that connect over slow networks affect the response time of your server because your server has to wait for acknowledgements from the client to proceed. Because you sent headers with the first send, there is no chance to do it later.

If buffering is turned off, you can enable buffering by using the following methods:

• Enable buffering programmatically in a page.

  // Response.Buffer is available for backwards compatibility; do not use.
  
  Response.BufferOutput = true;
  

• Enable buffering at the page level by using the @Page directive.

  <%@ Page Buffer = "true" %>
  

• Enable buffering at the application or computer level by using the <pages> element in the Web.config or Machine.config file.

  <pages buffer="true" …>
  

When you run your ASP.NET application by using the ASP.NET process model, it is even more important to have buffering enabled. The ASP.NET worker process first sends responses to IIS in the form of response buffers. After the ISAPI filter is running, IIS receives the response buffers. These response buffers are 31 KB in size., After IIS receives the response buffers, it then sends that actual response back to the client. With buffering disabled, instead of using the entire 31-KB buffer, ASP.NET can only send a few characters to the buffer. This causes extra CPU processing in both ASP.NET as well as in IIS. This may also cause memory consumption in the IIS process to increase dramatically.

Use Page.IsPostBack to Minimize Redundant Processing

Use the Page.IsPostBack property to ensure that you only perform page initialization logic when a page is first loaded and not in response to client postbacks. The following code fragment shows how to use the Page.IsPostBack property.

if (Page.IsPostBack == false) {

  // Initialization logic

} else {

  // Client post-back logic

}

Partition Page Content to Improve Caching Efficiency and Reduce Rendering

Partition the content in your page to increase caching potential. Partitioning your page content enables you to make different decisions about how you retrieve, display, and cache the content. You can use user controls to segregate static content, such as navigational items, menus, advertisements, copyrights, page headers, and page footers. You should also separate dynamic content and user-specific content for maximum flexibility when you want to cache content.

More Information

For more information, see "Partial Page or Fragment Caching" later in this chapter.

Ensure Pages Are Batch Compiled

As the number of assemblies that are loaded in a process grows, the virtual address space can become fragmented. When the virtual address space is fragmented, out-of-memory conditions are more likely to occur. To prevent a large number of assemblies from loading in a process, ASP.NET tries to compile all pages that are in the same directory into a single assembly. This occurs when the first request for a page in that directory occurs. Use the following techniques to reduce the number of assemblies that are not batch compiled:

• Do not mix multiple languages in the same directory. When multiple languages such as C# or Visual Basic .NET are used in pages in the same directory, ASP.NET compiles a separate assembly for each language.
• Ensure content updates do not cause additional assemblies to be loaded. For more information, see "Deployment Considerations" later in this chapter.
• Ensure that the debug attribute is set to false at the page level and in the Web.config file, as described in the following section.

Ensure Debug Is Set to False

When debug is set to true, the following occurs:

• Pages are not batch compiled.
• Pages do not time out. When a problem occurs, such as a problem with a Web service call, the Web server may start to queue requests and stop responding.
• Additional files are generated in the Temporary ASP.NET Files folder.
• The System.Diagnostics.DebuggableAttribute attribute is added to generated code. This causes the CLR to track extra information about generated code, and it also disables certain optimizations.

Before you run performance tests and before you move your application into production, be sure that debug is set to false in the Web.config file and at the page level. By default, debug is set to false at the page level. If you do need to set this attribute during development time, it is recommended that you set it at the Web.config file level, as shown in the following fragment.

<compilation debug="false" ... />

The following shows how to set debug to false at the page level.

<%@ Page debug="false" ... %>

Note   A common pitfall is to set this attribute at the page level during development and then forget to set it back when the application is moved to production.

Optimize Expensive Loops

Expensive loops in any application can cause performance problems. To reduce the overhead that is associated with code inside loops, you should follow these recommendations:

• Avoid repetitive field or property access.
• Optimize code inside the loop.
• Copy frequently called code into the loop.
• Replace recursion with looping.
• Use For instead of ForEach in performance-critical code paths.

More Information

For more information about the recommendations in this section, see "Iterating and Looping" in Chapter 5, "Improving Managed Code Performance."

Consider Using Server.Transfer Instead of Response.Redirect

Response.Redirect sends a metatag to the client that makes the client send a new request to the server by using the new URL. Server.Transfer avoids this indirection by making a server-side call. When you use Server.Transfer, the URL in the browser does not change, and load test tools may incorrectly report the page size because different pages are rendered for the same URL.

The Server.Transfer, Response.Redirect, and Response.End methods all raise ThreadAbortException exceptions because they internally callResponse.End. The call to Response.End causes this exception. Consider using the overloaded method to pass false as the second parameter so that you can suppress the internal call to Response.End.

More Information

For more information, see Knowledge Base article 312629, "PRB: ThreadAbortException Occurs If You Use Response.End, Response.Redirect, or Server.Transfer," at http://support.microsoft.com/default.aspx?scid=kb;en-us;312629.

Use Client-Side Validation

Prevalidating data can help reduce the round trips that are required to process a user's request. In ASP.NET, you can use validation controls to implement client-side validation of user input.

Note   Ensure that you also use server-side validation for security reasons.

More Information

For more information on validation controls, see the following:

• "Web Forms Validation" in Visual Basic and Visual C# Concepts on MSDN at http://msdn.microsoft.com/en-us/library/465kys5b(VS.71).aspx
• Knowledge Base article 316662, "HOW TO: Use ASP.NET Validation Controls from Visual Basic .NET" athttp://support.microsoft.com/default.aspx?scid=kb;en-us;316662

Server Controls

You can use server controls to encapsulate and to reuse common functionality. Server controls provide a clean programming abstraction and are the recommended way to build ASP.NET applications. When server controls are used properly, they can improve output caching and code maintenance. The main areas you should review for performance optimizations are view state and control composition. Use the following guidelines when you develop server controls:

• Identify the use of view state in your server controls.
• Use server controls where appropriate.
• Avoid creating deep hierarchies of controls.

Identify the Use of View State in Your Server Controls

View state is serialized and deserialized on the server. To save CPU cycles, reduce the amount of view state that your application uses. Disable view state if you do not need it. Disable view state if you are doing at least one of the following:

• Displaying a read-only page where there is no user input
• Displaying a page that does not post back to the server
• Rebuilding server controls on each post back without checking the postback data

More Information

For more information about view state, see "View State" later in this chapter.

Use Server Controls Where Appropriate

The HTTP protocol is stateless; however, server controls provide a rich programming model that manages state between page requests by using view state. Server controls require a fixed amount of processing to establish the control and all of its child controls. This makes server controls relatively expensive compared to HTML controls or possibly static text. Scenarios where server controls are expensive include the following:

• Large payload over low bandwidth. The more controls that you have on a page, the higher the network payload is. Therefore, multiple controls decreases the time to last byte (TTLB) and the time to first byte (TTFB) for the response that is sent to the client. When the bandwidth between client and server is limited, as is the case when a client uses a low-speed dial-up connection, pages that carry a large view state payload can significantly affect performance.
• View state overhead. View state is serialized and deserialized on the server. The CPU effort is proportional to the view state size. In addition to server controls that use view state, it is easy to programmatically add any object that can be serialized to the view state property. However, adding objects to the view state adds to the overhead. Other techniques such as storing, computed data or storing several copies of common data adds unnecessary overhead.
• Composite controls or large number of controls. Pages that have composite controls such as DataGrid may increase the footprint of the view state. Pages that have a large number of server controls also may increase the footprint of the view state. Where possible, consider the alternatives that are presented later in this section.

When you do not need rich interaction, replace server controls with an inline representation of the user interface that you want to present. You might be able to replace a server control under the following conditions:

• You do not need to retain state across postbacks.
• The data that appears in the control is static. For example, a label is static data.
• You do not need programmatic access to the control on the server-side.
• The control is displaying read-only data.
• The control is not needed during postback processing.

Alternatives to server controls include simple rendering, HTML elements, inline Response.Write calls, and raw inline angle brackets (<% %>). It is essential to balance your tradeoffs. Avoid over optimization if the overhead is acceptable and if your application is within the limits of its performance objectives.

Avoid Creating Deep Hierarchies of Controls

Deeply nested hierarchies of controls compound the cost of creating a server control and its child controls. Deeply nested hierarchies create extra processing that could be avoided by using a different design that uses inline controls, or by using a flatter hierarchy of server controls. This is especially important when you use list controls such as Repeater, DataList, and DataGrid because they create additional child controls in the container.

For example, consider the following Repeater control.

<asp:repeater id=r runat=server>

  <itemtemplate>

  <asp:label runat=server><%# Container.DataItem %><br></asp:label>

  </itemtemplate>

</asp:repeater>

Assuming there are 50 items in the data source, if you enable tracing for the page that contains the Repeater control, you would see that the page actually contains more than 200 controls.

Table 6.2: Partial Repeater Control Hierarchy

The ASP.NET list controls are designed to handle many different scenarios and may not be optimized for your scenario. In situations where performance is critical, you can choose from the following options:

• If you want to display data that is not very complex, you might render it yourself by calling Response.Write. For example, the following code fragment would produce the same output, as noted earlier in the section.

  for(int i=0;i<datasource.Count;i++)
  
  {
  
    Response.Write(datasource[i] + "<br>");
  
  }

• If you want to display more complex data, you can create a custom control to do the rendering. For example, the following custom control produces the same output as noted earlier in the section.

  public class MyControl : Control
  
  {
  
    private IList _dataSource;
  
    public IList DataSource
  
    {
  
      get {return _dataSource;}
  
      set {_dataSource=value;}
  
    }
  
    protected override void Render(HtmlTextWriter writer)
  
    {
  
      for(int i=0;i<_dataSource.Count;i++)
  
      {
  
        writer.WriteLine(_dataSource[i] + "<br>");
  
      }
  
    }
  
  }
  

More Information

For general background information about server controls, see Knowledge Base article 306459, "INFO: ASP.NET Server Controls Overview," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;306459.

Data Binding

Data binding is another common area that often leads to performance problems if it is used inefficiently. If you use data binding, consider the following recommendations:

• Avoid using Page.DataBind.
• Minimize calls to DataBinder.Eval.

Avoid Using Page.DataBind

Calling Page.DataBind invokes the page-level method. The page-level method in turn calls the DataBind method of every control on the page that supports data binding. Instead of calling the page-level DataBind, call DataBind on specific controls. Both approaches are shown in the following examples.

The following line calls the page level DataBind. The page level DataBind in turn recursively calls DataBind on each control.

DataBind();

The following line calls DataBind on the specific control.

yourServerControl.DataBind();

Minimize Calls to DataBinder.Eval

The DataBinder.Eval method uses reflection to evaluate the arguments that are passed in and to return the results. If you have a table that has 100 rows and 10 columns, you call DataBinder.Eval 1,000 times if you use DataBinder.Eval on each column. Your choice to use DataBinder.Eval is multiplied 1,000 times in this scenario. Limiting the use of DataBinder.Eval during data binding operations significantly improves page performance. Consider the following ItemTemplate element within a Repeater control using DataBinder.Eval.

<ItemTemplate>

  <tr>

    <td><%# DataBinder.Eval(Container.DataItem,"field1") %></td>

    <td><%# DataBinder.Eval(Container.DataItem,"field2") %></td>

  </tr>

</ItemTemplate>

There are alternatives to using DataBinder.Eval in this scenario. The alternatives include the following:

• Use explicit casting. Using explicit casting offers better performance by avoiding the cost of reflection. Cast the Container.DataItem as aDataRowView.

  <ItemTemplate>
  
    <tr>
  
      <td><%# ((DataRowView)Container.DataItem)["field1"] %></td>
  
      <td><%# ((DataRowView)Container.DataItem)["field2"] %></td>
  
    </tr>
  
  </ItemTemplate>
  

  You can gain even better performance with explicit casting if you use a DataReader to bind your control and use the specialized methods to retrieve your data. Cast the Container.DataItem as a DbDataRecord.

  <ItemTemplate>
  
    <tr>
  
       <td><%# ((DbDataRecord)Container.DataItem).GetString(0) %></td>
  
       <td><%# ((DbDataRecord)Container.DataItem).GetInt(1) %></td>
  
    </tr>
  
  </ItemTemplate>
  

  The explicit casting depends on the type of data source you are binding to; the preceding code illustrates an example.

• Use the ItemDataBound event. If the record that is being data bound contains many fields, it may be more efficient to use the ItemDataBoundevent. By using this event, you only perform the type conversion once. The following sample uses a DataSet object.

  protected void Repeater_ItemDataBound(Object sender, RepeaterItemEventArgs e)
  
  {
  
    DataRowView drv = (DataRowView)e.Item.DataItem;
  
    Response.Write(string.Format("<td>{0}</td>",drv["field1"]));
  
    Response.Write(string.Format("<td>{0}</td>",drv["field2"]));
  
    Response.Write(string.Format("<td>{0}</td>",drv["field3"]));
  
    Response.Write(string.Format("<td>{0}</td>",drv["field4"]));
  
  }
  

More Information

For more information about data binding, see the following Knowledge Base articles:

• 307860, "INFO: ASP.NET Data Binding Overview" at http://support.microsoft.com/default.aspx?scid=kb;en-us;307860
• 313481, "INFO: Roadmap for Web Forms Data Binding" at http://support.microsoft.com/default.aspx?scid=kb;en-us;313481
• 314809, "PRB: ASP.NET Fires Change Events Even If You Do Not Change the Control Value" at http://support.microsoft.com/default.aspx?scid=kb;en-us;314809

Caching Explained

Caching avoids redundant work. If you use caching properly, you can avoid unnecessary database lookups and other expensive operations. You can also reduce latency.

The ASP.NET cache is a simple, scalable, in-memory caching service provided to ASP.NET applications. It provides a time-based expiration facility, and it also tracks dependencies on external files, directories, or other cache keys. It also provides a mechanism to invoke a callback function when an item expires in the cache. The cache automatically removes items based on a least recently used (LRU) algorithm, a configured memory limit, and theCacheItemPriority enumerated value of the items in the cache. Cached data is also lost when your application or worker process recycles.

ASP.NET provides the following three caching techniques:

• Cache API
• Output caching
• Partial page or fragment caching

These caching techniques are briefly summarized in the following sections.

Cache API

You should use the cache API to programmatically cache application-wide data that is shared and accessed by multiple users. The cache API is also a good place for data that you need to manipulate in some way before you present the data to the user. This includes data such as strings, arrays, collections, and data sets.

Some common scenarios where you might want to use the cache API include the following:

• Headlines. In most cases, headlines are not updated in real time; they are often delayed for 10 to 20 minutes. Because headlines are shared by multiple users and are updated infrequently, this makes them good candidates for the cache API.
• Product catalogs. Product catalogs are good candidates for the cache API because the data typically needs to be updated at specific intervals, shared across the application, and manipulated before sending the content to the client.

You should avoid using the cache API in the following circumstances:

• The data you are caching is user-specific. Consider using session state instead.
• The data is updated in real time.
• Your application is already in production, and you do not want to update the code base. In this case, consider using output caching.

The cache API permits you to insert items in the cache that have a dependency upon external conditions. Cached items are automatically removed from the cache when the external conditions change. You use this feature by using a third parameter on the Cache.Insert method that accepts an instance of a CacheDependency class. The CacheDependency class has eight different constructors that support various dependency scenarios. These constructors include file-based, time-based, and priority-based dependencies, together with dependencies that are based on existing dependencies.

You can also run code before serving data from the cache. For example, you might want to serve cached data for certain customers, but for others you might want to serve data that is updated in real time. You can perform this type of logic by using the HttpCachePolicy.AddValidationCallbackmethod.

Output Caching

The output cache enables you to cache the contents of entire pages for a specific duration of time. It enables you to cache multiple variations of the page based on query strings, headers, and userAgent strings. The output cache also enables you to determine where to cache the content, for example on a proxy, server, or a client. Like the cache API, output caching enables you to save time retrieving data. Output caching also saves time rendering content. You should enable output caching on dynamically generated pages that do not contain user-specific data in scenarios where you do not need to update the view on every request.

Some common scenarios that are ideal for output caching include the following:

• Pages that are frequently visited. Output caching can be used to increase the overall performance of an application after the application is released to production by identifying the heavily visited pages and by enabling output caching on those specific pages if possible.
• Reports. Reports that only contain a low number of variations are good candidates for using output caching because you save time by not retrieving and processing the data each time the page is accessed.

Avoid using output caching in the following circumstances:

• You need programmatic access to the data on your page. Consider using the cache API instead.
• The number of page variants becomes too large.
• The page contains a mixture of static, dynamic, and user-specific data. Consider using fragment caching instead.
• The page contains content that must be refreshed with every view.

Partial Page or Fragment Caching

Partial page or fragment caching is a subset of output caching. It includes an additional attribute that allows you to cache a variation based on the properties of the user control (.ascx file.)

Fragment caching is implemented by using user controls in conjunction with the @OutputCache directive. Use fragment caching when caching the entire content of a page is not practical. If you have a mixture of static, dynamic, and user-specific content in your page, partition your page into separate logical regions by creating user controls. These user controls can then be cached, independent of the main page, to reduce processing time and to increase performance.

Some common scenarios that make good candidates for fragment caching include the following:

• Navigation menus. Navigation menus that are not user-specific are great candidates for fragment caching because menus are usually rendered with each request and are often static.
• Headers and footers. Because headers and footers are essentially static content that does not need to be regenerated with every request, they make good candidates for fragment caching.

You should avoid using fragment caching under the following conditions:

• The number of page variants becomes too large.
• The cached user controls contain content that must be refreshed with every view.

If your application uses the same user control on multiple pages, make the pages share the same instance by setting the Shared attribute of the user control @ OutputCache directive to true. This can save a significant amount of memory.

Caching Guidelines

Consider the following guidelines when you are designing a caching strategy:

• Separate dynamic data from static data in your pages.
• Configure the memory limit.
• Cache the right data.
• Refresh your cache appropriately.
• Cache the appropriate form of data.
• Use output caching to cache relatively static pages.
• Choose the right cache location.
• Use VaryBy attributes for selective caching.
• Use kernel caching on Windows Server 2003.

Separate Dynamic Data from Static Data in Your Pages

Partial page caching enables you to cache parts of a page by using user controls. Use user controls to partition your page. For example, consider the following simple page which contains static, dynamic, and user-specific information.

[main.aspx]

<html>

<body>

<table>

<tr><td colspan=3>Application Header – Welcome John Smith</td></tr>

<tr><td>Menu</td><td>Dynamic Content</td><td>Advertisments</td></tr>

<tr><td colspan=3>Application Footer</td></tr>

</table>

</html>

You can partition and cache this page by using the following code:

[main.aspx]

<%@ Register TagPrefix="app" TagName="header" src="header.ascx" %>

<%@ Register TagPrefix="app" TagName="menu" src="menu.ascx" %>

<%@ Register TagPrefix="app" TagName="advertisements"

src="advertisements.ascx" %>

<%@ Register TagPrefix="app" TagName="footer" src="footer.ascx" %>

<html>

<body>

<table>

<tr><td colspan=3><app:header runat=server /></td></tr>

<tr><td><app:menu runat=server /></td><td>Dynamic

Content</td><td><app:advertisements runat=server /></td></tr>

<tr><td colspan=3><app:footer runat=server /></td></tr>

</table>

</html>

[header.ascx]

<%@Control %>

Application Header – Welcome <% GetName() %>

[menu.ascx]

<%@Control %>

<%@ OutputCache Duration="30" VaryByParam="none" %>

Menu

[advertisements.ascx]

<%@Control %>

<%@ OutputCache Duration="30" VaryByParam="none" %>

Advertisements

[footer.ascx]

<%@Control %>

<%@ OutputCache Duration="60" VaryByParam="none" %>

Footer

By partitioning the content, as shown in the sample, you can cache selected portions of the page to reduce processing and rendering time.

Configure the Memory Limit

Configuring and tuning the memory limit is critical for the cache to perform optimally. The ASP.NET cache starts trimming the cache based on a LRU algorithm and the CacheItemPriority enumerated value assigned to the item after memory consumption is within 20 percent of the configured memory limit. If the memory limit is set too high, it is possible for the process to be recycled unexpectedly. Your application might also experience out-of-memory exceptions. If the memory limit is set too low, it could increase the amount of time spent performing garbage collections, which decreases overall performance.

Empirical testing shows that the likelihood of receiving out-of-memory exceptions increases when private bytes exceed 800 megabytes (MB). A good rule to follow when determining when to increase or decrease this number is that 800 MB is only relevant for .NET Framework 1.0. If you have .NET Framework 1.1 and if you use the /3 GB switch, you can go up to 1,800 MB.

When using the ASP.NET process model, you configure the memory limit in the Machine.config file as follows.

<processModel memoryLimit="50">

This value controls the percentage of physical memory that the worker process is allowed to consume. The process is recycled if this value is exceeded. In the previous sample, if there are 2 gigabytes (GB) of RAM on your server, the process recycles after the total available physical RAM falls below 50 percent of the RAM; in this case 1 GB. In other words, the process recycles if the memory used by the worker process goes beyond 1 GB. You monitor the worker process memory by using the process performance counter object and the private bytes counter.

More Information

For more information about how to tune the memory limit and about the /3GB switch, see "Configure the Memory Limit" and "/3GB Switch" in Chapter 17, "Tuning .NET Application Performance."

Cache the Right Data

It is important to cache the right data. If you cache the wrong data, you may adversely affect performance.

Cache application-wide data and data that is used by multiple users. Cache static data and dynamic data that is expensive to create or retrieve. Data that is expensive to retrieve and that is modified on a periodic basis can still provide performance and scalability improvements when managed properly. Caching data even for a few seconds can make a big difference to high volume sites. Datasets or custom classes that use optimized serialization for data binding are also good candidates for caching. If the data is used more often than it is updated, it is also a candidate for caching.

Do not cache expensive resources that are shared, such as database connections, because this creates contention. Avoid storing DataReader objects in the cache because these objects keep the underlying connections open. It is better to pool these resources. Do not cache per-user data that spans requests — use session state for that. If you need to store and to pass request-specific data for the life of the request instead of repeatedly accessing the database for the same request, consider storing the data in the HttpContext.Current.Cache object.

Refresh Your Cache Appropriately

Just because your data updates every ten minutes does not mean that your cache needs to be updated every ten minutes. Determine how frequently you have to update the data to meet your service level agreements. Avoid repopulating caches for data that changes frequently. If your data changes frequently, that data may not be a good candidate for caching.

Cache the Appropriate Form of the Data

If you want to cache rendered output, you should consider using output caching or fragment caching. If the rendered output is used elsewhere in the application, use the cache API to store the rendered output. If you need to manipulate the data, then cache the data by using the cache API. For example, if you need the data to be bound to a combo box, convert the retrieved data to an ArrayList object before you cache it.

Use Output Caching to Cache Relatively Static Pages

If your page is relatively static across multiple user requests, consider using page output caching to cache the entire page for a specified duration. You specify the duration based on the nature of the data on the page. A dynamic page does not always have to be rebuilt for every request just because it is a dynamic page. For example, you might be able to cache Web-based reports that are expensive to generate for a defined period. Caching dynamic pages for even a minute or two can increase performance drastically on high volume pages.

If you need to remove an item from the cache instead of waiting until the item expires, you can use the HttpResponse.RemoveOutputCacheItemmethod. This method accepts an absolute path to the page that you want to remove as shown in the following code fragment.

HttpResponse.RemoveOutputCacheItem("/Test/Test.aspx");

The caveat here is that this is specific to a server, because the cache is not shared across a Web farm. Also, it cannot be used from a user control.

Note   The next version of ASP.NET (code-named "Whidbey") is likely to support a database cache dependency. If it is implemented, this database cache dependency will allow you to remove items from the cache when data changes in the database.

Choose the Right Cache Location

The @OutputCache directive allows you to determine the cache location of the page by using the Location attribute. The Location attribute provides the following values:

• Any. This is the default value. The output cache can be located on the browser client where the request originated, on a proxy server, or any other server that is participating in the request or on the server where the request is processed.
• Client. The output cache is located on the browser client where the request originated.
• DownStream. The output cache can be stored in any HTTP 1.1 cache-capable device except for the origin server. This includes proxy servers and the client that made the request.
• None. The output cache is disabled for the requested page.
• Server. The output cache is located on the Web server where the request was processed.
• ServerAndClient. The output cache can be stored only at the origin server or at the requesting client. Proxy servers cannot cache the response.

Unless you know for certain that your clients or your proxy server will cache responses, it is best to keep the Location attribute set to Any, Server, orServerAndClient. Otherwise, if there is not a downstream cache available, the attribute effectively negates the benefits of output caching.

Note   The Location attribute does not apply to user controls.

Use VaryBy Attributes for Selective Caching

The VaryBy attributes allow you to cache different versions of the same page. ASP.NET provides four VaryBy attributes:

• VaryByParam. Different versions of the page are stored based on the query string values.
• VaryByHeader. Different versions of the page are stored based on the specified header values.
• VaryByCustom. Different versions of the page are stored based on browser type and major version. Additionally, you can extend output caching by defining custom strings.
• VaryByControl. Different versions of the page are stored based on the property value of a user control. This only applies to user controls.

The VaryBy attribute determines the data that is cached. The following sample shows how to use the VaryBy attribute.

<%@ OutputCache Duration="30" VaryByParam="a" %>

The setting shown in the previous sample would make the following pages have the same cached version:

• http://localhost/cache.aspx?a=1
• http://localhost/cache.aspx?a=1&b=1
• http://localhost/cache.aspx?a=1&b=2

If you add b to the VaryByParam attribute, you would have three separate versions of the page rather than one version. It is important for you to be aware of the number of variations of the cached page that could be cached. If you have two variables (a and b), and a has 5 different combinations, andb has 10 different combinations, you can calculate the total number of cached pages that could exist by using the following formula:

(MAX a × MAX b) + (MAX a + MAX b) = 65 total variations

When you make the decision to use a VaryBy attribute, make sure that there are a finite number of variations because each variation increases the memory consumption on the Web server.

Use Kernel Caching on Windows Server 2003

Windows Server 2003 and IIS 6.0 provide kernel caching. ASP.NET pages can automatically benefit from the IIS 6.0 kernel cache. Kernel caching produces significant performance gains because requests for cached responses are served without switching to user mode.

More Information

For more information, see "Kernel Mode Caching" in "IIS 6.0 Considerations" later in this chapter.

More Information

For more information on caching in general, see the following Knowledge Base articles:

• 811431, "HOWTO: Cache in ASP.NET by Using Visual Basic .NET," at http://support.microsoft.com/default.aspx?scid=kb;en-us;811431
• 323290, "HOWTO: To: Cache in ASP.NET by Using Visual C# .NET," at http://msdn.microsoft.com/en-us/kb/kb00323290.aspx
• 308375, "HOW TO: Control Page Output Caching in ASP.NET by Using Visual C# .NET," at http://support.microsoft.com/default.aspx?scid=kb;en-us;308375
• 315896, "HOW TO: Improve Performance by Caching Pages in ASP.NET," at http://support.microsoft.com/default.aspx?scid=kb;en-us;315896
• 312358, "HOW TO: Implement Key-Based Dependencies for Data Caching in ASP.NET by Using Visual Basic .NET," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;312358
• 308147, "HOW TO: Implement Key-Based Dependencies for Data Caching in ASP.NET by Using Visual C# .NET," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;308147

For more information about programmatic caching, see "Using Programmatic Caching" in "Understanding Caching Technologies" of the Caching Architecture Guide for .NET Framework Applications at http://msdn.microsoft.com/en-us/library/ms978500.aspx.

State Management

Web applications present specific challenges for state management. This is especially true for Web applications that are deployed in Web farms. The choices that you make regarding where and how state is stored have a significant impact on the performance and scalability of your application. There are several different types of state:

• Application state. Application state is used for storing application-wide state for all clients. Using application state affects scalability because it causes server affinity. In a Web scenario, if you modify application state, there is no mechanism to replicate the changes across servers. Therefore, if a subsequent request from the same user goes to another server, the change is not available. You store data in application state by using a key/value pair, as shown in the following sample.

  Application["YourGlobalState"] = somevalue;
  

• Session state. Session state is used for storing per-user state on the server. The state information is tracked by using a session cookie or a mangled URL. ASP.NET session state scales across Web servers in a farm.
• View state. View state is used for storing per-page state information. The state flows with every HTTP POST request and response.
• Alternatives. Other techniques for state management include client cookies, query strings, and hidden form fields.

Guidelines that are specific to application state, session state, and view state are included in later sections. The following are guidelines that address the broad issues that concern state management in general:

• Store simple state on the client where possible.
• Consider serialization costs.

Store Simple State on the Client Where Possible

Use cookies, query strings, and hidden controls for storing lightweight, user-specific state that is not sensitive such as personalization data. Do not use them to store security-sensitive information because the information can be easily read or manipulated.

• Client cookies. Client cookies are created on the server, and they are sent and stored on the client browser. They are domain specific and are not completely secure. All subsequent requests from a browser include the cookies, which the server code can inspect and modify. The maximum amount of data that you can put in cookie is 4 KB.
• Query strings. Query strings are the data that is appended to a URL. The data is clear text and there is a limit on the overall string length. The data can easily be manipulated by the user. Therefore, do not retrieve and display sensitive data based on query parameters without using authentication or validation. For anonymous Web sites, this is less of an issue.
• Hidden controls. Hidden controls on the page store state information that is sent back and forth in requests and responses.

More Information

For more information about the security implications of using these various state management techniques, see Chapter 10, "Building Secure ASP.NET Pages and Controls" in Improving Web Application Security: Threats and Countermeasures on MSDN at http://msdn.microsoft.com/en-us/library/aa302426.aspx.

Consider Serialization Costs

If you need to serialize state, consider the serialization costs. For example, you might want to serialize state to store in a remote state store. Only store what is absolutely necessary, and prefer simple types rather than complex objects to reduce the impact of serialization.

Application State

Application state is used to store application-wide static information. ASP.NET includes application state primarily for compatibility with classic Active Server Pages (ASP) technology so that it is easier to migrate existing applications to ASP.NET.

If you use application state, use the following guidelines to ensure your application runs optimally:

• Use static properties instead of the Application object to store application state.
• Use application state to share static, read-only data.
• Do not store STA COM objects in application state.

Use Static Properties Instead of the Application Object to Store Application State

You should store data in static members of the application class instead of in the Application object. This increases performance because you can access a static variable faster than you can access an item in the Application dictionary. The following is a simplified example.

<%

private static string[] _states[];

private static object _lock = new object();

public static string[] States

{

  get {return _states;}

}

public static void PopulateStates()

{

  //ensure this is thread safe

  if(_states == null)

  {

    lock(_lock)

    {

        //populate the states…    }

  }

}

public void Application_OnStart(object sender, EventArgs e)

{

  PopulateStates();

}

%>

Use Application State to Share Static, Read-Only Data

Application state is application-wide and specific to a server. Even though you can store read-write data, it advisable to only store read-only data to avoid server affinity. Consider using the Cache object. The Cache object is a better alternative for read-only data.

Do Not Store STA COM Objects in Application State

Storing STA COM objects in application state bottlenecks your application because the application uses a single thread of execution when it accesses the component. Avoid storing STA COM objects in application state.

More Information

For more information about application state, see Knowledge Base article, 312607, "INFO: Application Instances, Application Events, and Application State in ASP.NET," at http://support.microsoft.com/default.aspx?scid=kb;en-us;312607.

Session State

If you need session state in ASP.NET, there are three session state modes that you can choose from. Each mode offers varying degrees of performance and scalability as described in the following list:

• InProc. The in-process store provides the fastest access to session state. There are no serialization or marshaling costs involved because state is maintained within the managed memory of the ASP.NET process. The ASP.NET process is the Aspnet_wp.exe file on Windows 2000 Server, and the W3wp.exe file on Windows Server 2003. When the process recycles, the state data is lost, although you can disable process recycling in IIS 6 if process recycling affects your application. The in-process store limits application scalability because you cannot use it in conjunction with multiple worker processes; for example, it prevents Web farm or Web garden deployment. Also, high numbers of large or concurrent sessions can cause your application to run out of memory.
• StateServer. The session state service, a Microsoft Win32® service, can be installed on your local Web server or on a remote server that is accessible by all Web servers in a Web farm. This approach scales well, but performance is reduced in comparison to the in-process provider because of the additional serialization and marshaling that is required to transfer the state to and from the state store.
• SQL Server. Microsoft SQL Server provides a highly scalable and easily available solution. SQL Server is a solution that is well-suited to large amounts of session state. The serialization and marshalling costs are the same as the costs for the session state service, although overall performance is slightly lower. SQL Server provides clustering for failover, although this is not supported in the default configuration for session state. To enable clustering for failover, you have to apply configuration changes, and the session data must be stored in a non temporary table.

  For more information, see Knowledge Base article 323262,"INFO: ASP.NET Session State with SqlServer Mode in a Failover Cluster," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;323262.

Choosing a State Store

The in-process state store provides excellent performance and scales well. However, most high volume Web applications run in a Web farm. To be able to scale out, you need to choose between the session state service and the SQL Server state store. With either of these choices, you have to understand the associated impact of network latency and serialization, and you have to measure them to ensure that your application meets its performance objectives. Use the following information to help choose a state store:

• Single Web server. Use the in-process state store when you have a single Web server, when you want optimum session state performance, and when you have a reasonable and limited number of concurrent sessions. Use the session state service running on the local Web server when your sessions are expensive to rebuild and when you require durability in the event of an ASP.NET restart. Use the SQL Server state store when reliability is your primary concern.
• Web farm. Avoid the in-process option, and avoid running the session state service on the local Web server. These cause server affinity. You can use Internet Protocol (IP) affinity to ensure that the same server handles subsequent requests from the same client, but Internet service providers (ISP) that use a reverse proxy cause problems for this approach. Use a remote session state service or use SQL Server for Web farm scenarios.
• StateServer versus SQLServer. Use a remote state service, if you do not have a SQL Server database. Use SQL Server for enterprise applications or high volume Web applications. If your remote state service and your Web server are separated by a firewall, then you need to open a port. The default port is port 42424. You can change the port in the following registry key:

  HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\aspnet_state\Parameters.

To ensure optimized session state performance, follow these guidelines:

• Prefer basic types to reduce serialization costs.
• Disable session state if you do not use it.
• Avoid storing STA COM objects in session state.
• Use the ReadOnly attribute when you can.

Prefer Basic Types to Reduce Serialization Costs

You incur serialization overhead if you use the StateServer or the SQLServer out-of-process state stores. The simpler the object graph, the faster it should serialize. To minimize serialization costs, use basic types such as Int, Byte, Decimal, String, DateTime, TimeSpan, Guid, IntPtr, and UintPrt. ASP.NET uses an optimized internal serialization method to serialize basic types. Complex types are serialized using a relatively slow BinaryFormatterobject. For complex types, you can use the Serializable attribute, or you can implement the ISerializable interface. Using this interface provides you with more precise control and may speed up serialization.

Minimize what you serialize. Disable serialization when you do not use it, and mark specific fields from a serializable class that you want to exclude with the NonSerialized attribute. Alternatively, control the serialization process by using the ISerializable interface.

Note   You should only implement the ISerializable interface as a last resort. New formatters provided by future versions of the .NET Framework and improvements to the framework provided serialization will not be utilized once you take this approach. Prefer theNonSerialized attribute.

More Information

For more information, see "How To: Improve Serialization Performance" in the "How To" section of this guide.

Disable Session State If You Do Not Use It

If you do not use session state, disable session state to eliminate redundant session processing performed by ASP.NET. You might not use session state because you store simple state on the client and then pass it to the server for each request. You can disable session state for all applications on the server, for specific applications, or for individual pages, as described in the following list:

• To disable session state for all applications on your server, use the following element in the Machine.config file.

  <sessionState mode='Off'/>
  

  You can also remove the session state module from <httpModules> to completely remove session processing overhead.

• To disable session state for a specific application, use the following element in the Web.config file of the application.

  <sessionState mode='Off'/>
  

• To disable session state for a specific Web page, use the following page setting.

  <%@ Page EnableSessionState="false" . . .%>
  

Avoid Storing STA COM Objects in Session State

Storing STA COM objects in session state causes thread affinity. Thread affinity severely affects performance and scalability. If you do use STA COM objects in session state, be sure to set the AspCompat attribute of the @ Page directive.

More Information

For more information, see "COM Interop" later in this chapter.

Use the ReadOnly Attribute When You Can

Page requests that use session state internally use a ReaderWriterLock object to manage session data. This allows multiple reads to occur at the same time when no lock is held. When the writer acquires the lock to update session state, all read requests are blocked. Normally two calls are made to the database for each request. The first call connects to the database, marks the session as locked, and executes the page. The second call writes any changes and unlocks the session. For pages that only read session data, consider setting EnableSessionState to ReadOnly as shown in the following sample.

<%@ Page EnableSessionState="ReadOnly" . . .%>

Setting EnableSessionState to ReadOnly is particularly useful when you use frames. In this event, the default setting serializes the execution of the page because a ReaderWriterLock is used. By setting EnableSessionState to ReadOnly, you avoid blocking, and you send fewer calls to the database. One option is to disable sessions in the configuration file as shown earlier, and to set the ReadOnly attribute on a page-by-page basis.

More Information

For more information about session state, see "Underpinnings of the Session State Implementation in ASP.NET" on MSDN athttp://msdn.microsoft.com/en-us/library/aa479041.aspx.

For additional information on session state, see the following Knowledge Base articles:

• 307598, "INFO: ASP.NET State Management Overview," at http://support.microsoft.com/default.aspx?scid=kb;en-us;307598
• 311209, "HOW TO: Configure ASP.NET for Persistent SQL Server Session State Management," at http://support.microsoft.com/default.aspx?scid=kb;en-us;311209
• 317604, "HOW TO: Configure SQL Server to Store ASP.NET Session State," at http://support.microsoft.com/default.aspx?scid=kb;en-us;317604
• 306996, "HOW TO: Disable ASP Session State in ASP.NET," at http://support.microsoft.com/default.aspx?scid=kb;en-us;306996
• 326606, "BUG: Impersonation May Not Work When You Use ASP.NET SQL Server Session State with Integrated Security," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;326606
• 324772, "PRB: Session Data Is Lost When You Use ASP.NET InProc Session State Mode," at http://support.microsoft.com/default.aspx?scid=kb;en-us;324772

View State

View state is used primarily by server controls to retain state only on pages that post data back to themselves. The information is passed to the client and read back in a specific hidden variable called _VIEWSTATE. ASP.NET makes it easy to store any types that are serializable in view state. However, this capability can easily be misused and performance reduced. View state is an unnecessary overhead for pages that do not need it. As the view state grows larger. it affects performance in the following ways:

• Increased CPU cycles are required to serialize and to deserialize the view state.
• Pages take longer to download because they are larger.
• Very large view state can impact the efficiency of garbage collection.

Transmitting a huge amount of view state can significantly affect application performance. The change in performance becomes more marked when your Web clients use slow, dial-up connections. Consider testing for different bandwidth conditions when you work with view state. Optimize the way your application uses view state by following these recommendations:

• Disable view state if you do not need it.
• Minimize the number of objects you store in view state.
• Determine the size of your view state.

Disable View State If You Do Not Need It

View state is turned on in ASP.NET by default. Disable view state if you do not need it. For example, you might not need view state because your page is output-only or because you explicitly reload data for each request. You do not need view state when the following conditions are true:

• Your page does not post back. If the page does not post information back to itself, if the page is only used for output, and if the page does not rely on response processing, you do not need view state.
• You do not handle server control events. If your server controls do not handle events, and if your server controls have no dynamic or data bound property values, or they are set in code on every request, you do not need view state.
• You repopulate controls with every page refresh. If you ignore old data, and if you repopulate the server control each time the page is refreshed, you do not need view state.

There are several ways to disable view state at various levels:

• To disable view state for all applications on a Web server, configure the <pages> element in the Machine.config file as follows.

  <pages enableViewState="false" />
  

  This approach allows you to selectively enable view state just for those pages that need it by using the EnableViewState attribute of the @ Pagedirective.

• To disable view state for a single page, use the @ Page directive as follows.

  <%@ Page EnableViewState="false" %>
  

• To disable view state for a single control on a page, set the EnableViewState property of the control to false, as shown in the following code fragment.

  //programatically
  
  yourControl.EnableViewState = false;
  
  //something
  
  <asp:datagrid EnableViewState="false" runat= "server" />
  

Minimize the Number of Objects You Store In View State

As you increase the number of objects you put into view state, the size of your view state dictionary grows, and the processing time required to serialize and to deserialize the objects increases. Use the following guidelines when you put objects into view state:

• View state is optimized for serializing basic types such as strings, integers, and Booleans, and objects such as arrays, ArrayLists, and Hashtablesif they contain these basic types. When you want to store a type which is not listed previously, ASP.NET internally tries to use the associated type converter. If it cannot find one, it uses the relatively expensive binary serializer.
• The size of the object is directly proportional to the size of the view state. Avoid storing large objects.

Determine the Size of Your View State

By enabling tracing for the page, you can monitor the view state size for each control. The view state size for each control appears in the leftmost column in the control tree section of the trace output. Use this information as a guide to determine if there are any controls that you can reduce the amount of view state for or if there are controls that you can disable view state for.

More Information

For related information, see "Taking a Bite out of ASP.NET ViewState" on MSDN at http://msdn.microsoft.com/en-us/library/ms972427.aspx.

HTTP Modules

HTTP modules are filters that allow you to add preprocessing and postprocessing to HTTP request and response messages as they flow through the ASP.NET pipeline. They are commonly used for authorization, authentication, logging, and machine-level error handling. HTTP modules run on every request, so whatever processing they perform has global impact either on the application or the computer, depending on where you register them.

If you develop HTTP modules, consider the following:

• Avoid long-running and blocking calls in pipeline code.
• Consider asynchronous events.

Avoid Long-Running and Blocking Calls in Pipeline Code

Avoid placing long-running code in an HTTP module for the following reasons:

• ASP.NET pages are processed in a synchronous fashion.
• HTTP modules typically use synchronous events.

Long-running or blocking code reduces the concurrent requests that can run in ASP.NET.

Consider Asynchronous Events

For every synchronous event, there is also an asynchronous version of that event. Although asynchronous events still logically block the request for the duration of the asynchronous work, they do not block the ASP.NET thread.

More Information

For more information on HTTP modules, see the MSDN Magazine article, "ASP.NET Pipeline: Use Threads and Build Asynchronous Handlers in Your Server-Side Web Code," at http://msdn.microsoft.com/en-us/magazine/cc164128.aspx.

In addition, see Knowledge Base article 307985, "INFO: ASP.NET HTTP Modules and HTTP Handlers Overview," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;307985.

String Management

When you build output, you often need to concatenate strings. This is an expensive operation because it requires temporary memory allocation and subsequent collection. As a result, you should minimize the amount of string concatenation that you perform. There are three common ways to concatenate strings in your pages to render data:

• Using the += operator. Use the += operator when the number of appends is known.
• StringBuilder. Use the StringBuilder object when the number of appends is unknown. Treat StringBuffer as a reusable buffer.
• Response.Write <% %>. Use the Response.Write method. It is one of the fastest ways to return output back to the browser.

The most effective way to determine the option to choose is to measure the performance of each option. If your application relies heavily on temporary buffers, consider implementing a reusable buffer pool of character arrays or byte arrays.

Use the following guidelines when you are managing your strings:

• Use Response.Write for formatting output.
• Use StringBuilder for temporary buffers.
• Use HtmlTextWriter when building custom controls.

Use Response.Write for Formatting Output

Where possible, avoid using loops to concatenate strings for formatting page layout. Consider using Response.Write instead. This approach writes output to the ASP.NET response buffers. When you are looping through datasets or XML documents, using Response.Write is a highly efficient approach. It is more efficient than concatenating the content by using the += operator before writing the content back to the client. Response.Writeinternally appends strings to a reusable buffer so that it does not suffer the performance overhead of allocating memory, in addition to cleaning that memory up.

Use StringBuilder for Temporary Buffers

In many cases it is not feasible to use Response.Write. For example, you might need to create strings to write to a log file or to build XML documents. In these situations, use a StringBuilder object as a temporary buffer to hold your data. Measure the performance of your scenario by trying various initial capacity settings for the StringBuilder object.

Use HtmlTextWriter When Building Custom Controls

When you are building custom controls, the Render, RenderChildren, and RenderControl methods provide access to the HtmlTextWriter object. TheHtmlTextWriter writes to the same reusable buffer as Response.Write. In the same way as Response.Write, HtmlTextWriter does not suffer the performance overhead of allocating memory in addition to cleaning up the memory.

More Information

For more information about strings, see "String Operations" in Chapter 5, "Improving Managed Code Performance."

To determine if your application is creating excessive temporary memory allocation due to inefficient string concatenations, use performance counters, as discussed in the "Memory" topic in "CLR and Managed Code" in Chapter 15, "Measuring .NET Application Performance." To determine the source of the problem, use the CLR Profiler. For more information, see "How To: Use CLR Profiler" in the "How To" section of this guide.

Exception Management

Exceptions are expensive. By knowing the causes of exceptions, and by writing code that avoids exceptions and that handles exceptions efficiently, you can significantly improve the performance and scalability of your application. When you design and implement exception handling, consider the following guidelines to ensure optimum performance:

• Implement a Global.asax error handler.
• Monitor application exceptions.
• Use try/finally on disposable resources.
• Write code that avoids exceptions.
• Set timeouts aggressively.

Implement a Global.asax Error Handler

The first step in managing exceptions is to implement a global error handler in the Global.asax file or in the code-behind file. Implementing a global error handler traps all unhandled exceptions in your application. Inside the handler, you should, at a minimum, log the following information to a data store such as a database, the Windows event log, or a log file:

• The page that the error occurred on
• Call stack information
• The exception name and message

In your Global.asax file or your code-behind page, use the Application_Error event to handle your error logic, as shown in the following code sample:

public void Application_Error(object s, EventArgs ev)

{

   StringBuilder message = new StringBuilder();

   if (Server != null) {

     Exception e;

     for (e = Server.GetLastError(); e != null; e = e.InnerException)

     {

        message.AppendFormat("{0}: {1}{2}",

                             e.GetType().FullName,

                             e.Message,

                             e.StackTrace);

     }

     //Log the exception and inner exception information.

  }

}

More Information

For more information, see "Rich Custom Error Handling with ASP.NET" on MSDN at http://msdn.microsoft.com/en-us/library/aa479319.aspx.

Also, see Knowledge Base article 306355, "HOW TO: Create Custom Error Reporting Pages in ASP.NET by Using Visual C# .NET," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;306355.

Monitor Application Exceptions

To reduce the number of exceptions occurring in your application, you need to effectively monitor your application for exceptions. You can do the following:

• If you have implemented exception handling code, review your exception logs periodically.
• Monitor the # of Exceps Thrown / sec counter under the .NET CLR Exceptions Performance Monitor object. This value should be less then 5 percent of your average requests per second.

Use Try/Finally on Disposable Resources

To guarantee resources are cleaned up when an exception occurs, use a try/finally block. Close the resources in the finally clause. Using a try/finallyblock ensures that resources are disposed even if an exception occurs. The following code fragment demonstrates this.

try

{

   conn.Open();

   ...

}

finally

{

   if(null!=conn)

     conn.close;

}

Write Code That Avoids Exceptions

The following is a list of common techniques you can use to avoid exceptions:

• Check for null values. If it is possible for an object to be null, check to make sure it is not null, rather then throwing an exception. This commonly occurs when you retrieve items from view state, session state, application state, or cache objects as well as query string and form field variables. For example, do not use the following code to access session state information.

  try {
  
    loginid = Session["loginid"].ToString();
  
  }
  
  catch(Exception ex) {
  
    Response.Redirect("login.aspx", false);
  
  }
  

  Instead, use the following code to access session state information.

  if(Session["loginid"]!=null)
  
    loginid = Session["loginid"].ToString();
  
  else
  
    Response.Redirect("login.aspx", false);
  

• Do not use exceptions to control logic. Exceptions are just that — exceptions. A database connection that fails to open is an exception. A user who mistypes his password is simply a condition that needs to be handled. For example, consider the following function prototype used to log in a user.

  public void Login(string UserName, string Password) {}
  

  The following code is used to call the login.

  try
  
  {
  
    Login(userName,password);
  
  }
  
  catch (InvalidUserNameException ex)
  
  {…}
  
  catch (InvalidPasswordException ex)
  
  {…}
  

  It is better to create an enumeration of possible values and then change the Login method to return that enumeration, as follows.

  public enum LoginResult
  
  {
  
    Success,InvalidUserName, InvalidPassword, AccountLockedOut
  
  }
  
  public LoginResult Login(string UserName, string Password) {}
  

  The following code is used to call Login.

  LoginResult result = Login(userName,password)
  
  switch(result)
  
  {
  
    case Success:
  
     . . .
  
    case InvalidUserName:
  
     . . .
  
    case InvalidPassword:
  
  }
  

• Suppress the internal call to Response.End. The Server.Transfer, Response.Redirect, Response.End methods all raise exceptions. Each of these methods internally call Response.End. The call to Response.End, in turn, causes a ThreadAbortException exception. If you useResponse.Redirect, consider using the overloaded method and passing false as the second parameter to suppress the internal call toResponse.End.

  For more information, see Knowledge Base article 312629, "PRB: ThreadAbortException Occurs If You Use Response.End, Response.Redirect, or Server.Transfer," at http://support.microsoft.com/default.aspx?scid=kb;en-us;312629.

• Do not catch exceptions you cannot handle. If your code cannot handle an exception, use a try/finally block to ensure that you close resources, regardless of whether an exception occurs. Do not catch the exception if you cannot try recovery. Permit the exception to propagate to an appropriate handler that can deal with the exception condition.

Set Timeouts Aggressively

Page timeouts that are set too high can cause problems if parts of your application are operating slowly. For example, page timeouts that are set too high may cause the following problems:

• Browsers stop responding.
• Incoming requests start to queue.
• IIS rejects requests after the request queue limit is reached.
• ASP.NET stops responding.

The default page timeout is 90 seconds. You can change this value to accommodate your application scenario.

Consider the following scenario where an ASP.NET front-end application makes calls to a remote Web service. The remote Web service then calls a mainframe database. If, for any reason, the Web service calls to the mainframe start blocking, your front-end ASP.NET pages continue to wait until the back end calls time out, or the page timeout limit is exceeded. As a result, the current request times out, ASP.NET starts to queue incoming requests, and those incoming requests may time out, too. It is more efficient for your application to time out these requests in less than 90 seconds. Additionally, timing out the requests in less than 90 seconds improves the user experience.

In most Internet and intranet scenarios, 30 seconds is a very reasonable timeout limit. For high traffic pages such as a home page, you might want to consider lowering the timeout limit. If your application takes a long time to generate certain pages, such as report pages, increase the timeout limit for those pages.

More Information

For more information about exception handling, see the following MSDN articles:

• "An Exception to the Rule, Part 1," at http://msdn.microsoft.com/en-us/library/ms972393.aspx
• "An Exception to the Rule, Part 2," at http://msdn.microsoft.com/en-us/library/ms972399.aspx

For more information about the various timeout parameters and how to configure them, see "Configure Timeouts Aggressively" in Chapter 17, "Tuning .NET Application Performance."

COM Interop

Calling COM objects from ASP.NET may present performance challenges because you have to deal with threading issues, marshaling data types, and transitions across the boundary between managed and unmanaged code. Because ASP.NET runs requests on multithreaded apartment (MTA) threads, working with STA COM components may be especially challenging.

Use the following guidelines to improve COM interop performance:

• Use ASPCOMPAT to call STA COM objects.
• Avoid storing COM objects in session state or application state.
• Avoid storing STA objects in session state.
• Do not create STA objects in a page constructor.
• Supplement classic ASP Server.CreateObject with early binding.

Use ASPCOMPAT to Call STA COM Objects

When you call an STA object, such as a Visual Basic 6.0 component, from an ASP.NET page, use the page-level ASPCOMPAT attribute. Use theASPCOMPAT attribute as shown in the following sample, to denote that the events in your page should run using a thread from the STA thread pool rather than a default MTA thread.

<%@ Page ASPCOMPAT="true" language="c#" %>

STA object calls require an STA thread. If you do not use the ASPCOMPAT attribute, all STA object calls are serialized on the host STA thread and a serious bottleneck occurs.

Avoid Storing COM Objects in Session State or Application State

Avoid storing COM objects in state containers such as session state or application state. COM objects are not serializable, and although calling the object may work with a single-server deployment, affinity and serialization issues will prevent your application from working when it is moved to a Web farm.

Avoid Storing STA Objects in Session State

Even though it is technically possible to store STA objects in session state, do not do so because it causes thread affinity issues. If you do so, requests to the STA object have to be run on the same thread that created the object, and this quickly becomes a bottleneck as the number of users increases.

More Information

For more information, see Knowledge Base article 817005, "FIX: Severe Performance Issues When You Bind Session State to Threads in ASPCompat Mode," at http://support.microsoft.com/default.aspx?scid=kb;en-us;817005.

Do Not Create STA Objects in a Page Constructor

Do not create STA objects in a page constructor because this causes a thread switch to the host STA and causes all calls to be serialized. Although theASPCOMPAT attribute ensures that an STA thread from the STA thread pool is used for page events such as onload, button_click, and other page events, other parts of your page such as the constructor are run by using an MTA thread.

Supplement Classic ASP Server.CreateObject with Early Binding

Late binding requires extra instructions to locate the target code, whether this is a COM class or executing a method by name. Methods such asServer.CreateObject, Activator.CreateInstance and MethodInfo.Invoke allow late bound execution of code. When you migrate ASP code, use thenew keyword to allow early bound calls.

The following example uses early binding. The new operator is used to create a classic ActiveX® Data Objects (ADO) connection.

<%@ Import namespace="ADODB" %>

<%@ Assembly name="ADODB" %>

…   Connection con = new Connection();

The following example uses late binding. The <object> tag along with the class attribute is used to create an ADO connection object.ADODB.Connection represents the namespace and class name. The second ADODB represents the assembly name.

<object id="con" runat="server" class="ADODB.Connection, ADODB" />

The following example also uses late binding. GetType is used to obtain the type and this is passed to the overloaded Server.CreateObject method that is provided by ASP.NET.

con = Server.CreateObject(Type.GetType("ADODB.Connection, ADODB"));

For these code samples to work, add a reference to the Microsoft ActiveX Data Objects X.X Library in Visual Studio®.NET. Replace X.X with the version number that you want to use. This approach causes an interop assembly to be used if one exists or creates one automatically for you. If you are not using Visual Studio .NET, use the TlbImp.exe file to generate the interop assembly. It is recommended that you look for and use a primary interop assembly. Copy the generated interop assembly to the Bin directory of your application.

More Information

For more information about COM interop performance and issues, see Chapter 7, "Improving Interop Performance" in this guide. In addition, see the following Knowledge Base articles:

• 308095, "PRB: Creating STA Components in the Constructor in ASP.NET ASPCOMPAT Mode Negatively Affects Performance," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;308095
• 817005, "FIX: Severe Performance Issues When You Bind Session State to Threads in ASPCompat Mode" athttp://support.microsoft.com/default.aspx?scid=kb;en-us;817005
• 818612, "FIX: "COM Object Can Not Be Used" Error Message When You Use an STA COM Object That You Created by Using Server.CreateObject and Stored in Session Scope in a Different Web Page," at http://support.microsoft.com/default.aspx?scid=kb;en-us;818612
• 827164, "FIX: References to STA Objects That Are Stored in Session State May Become Corrupted If They Are Called from a Session_End Event,"http://support.microsoft.com/default.aspx?scid=kb;en-us;827164

Data Access

Almost all ASP.NET applications use some form of data access. Data access is typically a focal point for improving performance because the majority of application requests require data that comes from a database.

Use the following guidelines to improve your data access:

• Use paging for large result sets.
• Use a DataReader for fast and efficient data binding.
• Prevent users from requesting too much data.
• Consider caching data.

Use Paging for Large Result Sets

Paging large query result sets can significantly improve the performance of an application. If you have large result sets, implement a paging solution that achieves the following:

• The paging solution reduces back-end work on the database.
• The paging solution reduces the size of data that is sent to the client.
• The paging solution limits client work.

Several paging solutions are available; each solution solves the problems that are inherent to specific scenarios. The following paragraphs briefly summarize the solutions. For implementation-specific details, see the "How To: Page Records in .NET Applications" in the "How To" section of this guide.

A relatively quick and easy solution is to use the automatic paging provided by the DataGrid object. However, this solution works only for tables that have unique incrementing columns; it is not suitable for large tables. With the custom paging approach, you set AllowPaging andAllowCustomPaging properties to true, and then set the PageSize and VirtualItemCount properties. Then the StartIndex (the last browsed row) andNextIndex (StartIndex + PageSize) properties are calculated. The StartIndex and NextIndex values are used as ranges for the identity column to retrieve and display the requested page. This solution does not cache data; it pulls only the relevant records across the network.

There are several solutions available for tables that do not have unique incrementing column numbers. For tables that have a clustered index and and that do not require special server-side coding, use the subquery solution to track the number of rows to skip from the start. From the resulting records, use the TOP keyword in conjunction with the <pagesize> element to retrieve the next page of rows. Only the relevant page records are retrieved over the network. Other solutions use either the Table data type or a global temporary table with an additional IDENTITY column to store the queried results. This column is used to limit the range of rows fetched and displayed. This requires server-side coding.

More Information

For more information and implementation details for paging solutions, see "How To: Page Records in .NET Applications" in the "How To" section of this guide. Also, see Knowledge Base article 318131, "HOW TO: Page Through a Query Result for Better Performance," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;318131.

Use a DataReader for Fast and Efficient Data Binding

Use a DataReader object if you do not need to cache data, if you are displaying read - only data, and if you need to load data into a control as quickly as possible. The DataReader is the optimum choice for retrieving read-only data in a forward-only manner. Loading the data into a DataSet object and then binding the DataSet to the control moves the data twice. This method also incurs the relatively significant expense of constructing a DataSet.

In addition, when you use the DataReader, you can use the specialized type-specific methods to retrieve the data for better performance.

Prevent Users from Requesting Too Much Data

Allowing users to request and retrieve more data than they can consume puts an unnecessary strain on your application resources. This unnecessary strain causes increased CPU utilization, increased memory consumption, and decreased response times. This is especially true for clients that have a slow connection speed. From a usability standpoint, most users do not want to see thousands of rows presented as a single unit.

Limit the amount of data that users can retrieve by using one of the following techniques:

• Implement a paging mechanism. For more information, see "How To: Page Records in .NET Applications" in the "How To" section of this guide.
• Design a master/detail form. Instead of giving users all of the information for each piece of data, only display enough information to allow the users to recognize the piece of data they are interested in. Permit the user to select that piece of data and obtain more details.
• Enable users to filter the data.

Consider Caching Data

If you have application-wide data that is fairly static and expensive to retrieve, consider caching the data in the ASP.NET cache.

More Information

For more information about data access, see Chapter 12, "Improving ADO.NET Performance"

Security Considerations

Security and performance are often at the center of design tradeoffs, because additional security mechanisms often negatively impacts performance. However, you can reduce server load by filtering unwanted, invalid, or malicious traffic, and by constraining the requests that are allowed to reach your Web server. The earlier that you block unwanted traffic, the greater the processing overhead that you avoid. Consider the following recommendations:

• Constrain unwanted Web server traffic.
• Turn off authentication for anonymous access.
• Validate user input on the client.
• Avoid per-request impersonation.
• Avoid caching sensitive data.
• Segregate secure and non-secure content.
• Only use SSL for pages that require it.
• Use absolute URLs for navigation.
• Consider using SSL hardware to offload SSL processing.
• Tune SSL timeout to avoid SSL session expiration.

Constrain Unwanted Web Server Traffic

Constrain the traffic to your Web Server to avoid unnecessary processing. For example, block invalid requests at your firewall to limit the load on your Web server. In addition, do the following:

• Map unsupported extensions to the 404.dll file in IIS.
• Use the UrlScan filter to control the verbs and the URL requests that you allow. Verbs that you might want to control include Get, Post, andSOAP.
• Review your IIS logs. If the logs are full of traffic that you do not allow, investigate blocking that traffic at the firewall or filtering the traffic by using a reverse proxy.

Turn Off Authentication for Anonymous Access

Partition pages that require authenticated access from pages that support anonymous access. To avoid authentication overhead, set the authentication mode to None in the Web.config file in the directory that contains the anonymous pages. The following line shows how to set the authentication mode in the Web.config file.

<authentication mode="None" />

Validate User Input on the Client

Consider using client-side validation to avoid sending unwanted traffic to the server. However, do not trust client-side validation alone because it can easily be bypassed. For security reasons, you should implement the equivalent server-side checks for every client check.

Avoid Per-Request Impersonation

Per-request impersonation where you use the original caller's identity to access the database places severe scalability constraints on your application. Per-request impersonation prevents the effective use of database connection pooling. The trusted subsystem model is the preferred and scalable alternative. With this approach, you use a fixed service account to access the database and to pass the identity of the original caller at the application level if the identity of the original caller is required. For example, you might pass the identity of the original caller through stored procedure parameters.

More Information

For more information about the trusted subsystem model, see Chapter 3, "Authentication and Authorization," in Building Secure ASP.NET Applicationson MSDN at http://msdn.microsoft.com/en-us/library/aa302383.aspx.

Avoid Caching Sensitive Data

Instead of caching sensitive data, retrieve the data when you need it. When you measure application performance, if you discover that retrieving the data on a per-request basis is very costly, measure the cost to encrypt, cache, retrieve, and decrypt the data. If the cost to retrieve the data is higher than the cost to encrypt and decrypt the data, consider caching encrypted data.

Segregate Secure and Non-Secure Content

When you design the folder structure of your Web site, clearly differentiate between the publicly accessible areas and restricted areas that require authenticated access and Secure Sockets Layer (SSL). Use separate subfolders beneath the virtual root folder of your application to hold restricted pages such as forms logon pages, checkout pages, and any other pages that users transmit sensitive information to that needs to be secured by using HTTPS. By doing so, you can use HTTPS for specific pages without incurring the SSL performance overhead across your entire site.

Only Use SSL for Pages That Require It

Using SSL is expensive. Only use SSL for pages that require it. This includes pages that contain or capture sensitive data, such as pages that accept credit card numbers and passwords. Use SSL only if the following conditions are true:

• You want to encrypt the page data.
• You want to guarantee that the server to which you send the data is the server that you expect.

For pages where you must use SSL, follow these guidelines:

• Make the page size as small as possible.
• Avoid using graphics that have large file sizes. If you use graphics, use graphics that have smaller file sizes and resolution. Or, use graphics from a site that is not secure. However, when you use graphics from a site that is not secure, Web browsers display a dialog box that asks the user if the user wants to display the content from the site that is not secure.

Use Absolute URLs for Navigation

Navigating between HTTP and HTTPs using redirects uses the protocol of the current page instead of the protocol of the target page. When your redirects use relative links (..\publicpage.aspx) to sites that are not secure from a site that uses HTTPS, these public pages are served by using the HTTPS protocol. This use of the HTTPS protocol incurs unnecessary overhead. To avoid this problem, use the absolute link instead of the relative link for your redirects. For example, use an absolute link such as http://yourserver/publicpage.aspx. The same applies when you navigate from pages that use HTTP to pages that use HTTPS. The following code fragment shows how to create a redirect from a page that uses HTTP to a page that uses HTTPS.

string serverName =

HttpUtility.UrlEncode(Request.ServerVariables["SERVER_NAME"]);

string vdirName = Request.ApplicationPath;

Response.Redirect("https://" + serverName + vdirName +

"/Restricted/Login.aspx", false);

Consider Using SSL Hardware to Offload SSL Processing

Consider a hardware solution for SSL processing. Terminating SSL sessions at a load balancer by using a hardware accelerator generally offers better performance, particularly for sites that experience heavy use.

Tune SSL Timeout to Avoid SSL Session Expiration

If you are not using SSL hardware, tune the ServerCacheTimer property to avoid having to renegotiate the SSL handshakes with browser clients. The largest use of resources when you use SSL occurs during the initial handshake, where asymmetric public/private-key encryption is used. After a secure session key is generated and exchanged, faster, symmetric encryption is used to encrypt application data.

Monitor your SSL connections and increase the value of the ServerCacheTime registry entry if you find that a longer time is better for your scenario.

More Information

For more information about how to change the ServerCacheTime value, see Knowledge Base article 247658, "HOW TO: Configure Secure Sockets Layer Server and Client Cache Elements," at http://support.microsoft.com/default.aspx?scid=kb;en-us;247658.

More Information

For more information about security-related performance considerations, see the following:

• "Cryptographic Services" in the .Net Framework Developer's Guide at http://msdn.microsoft.com/en-us/library/93bskf9z(VS.80).aspx
• "Performance Comparison: Security Design Choices" on MSDN at http://msdn.microsoft.com/en-us/library/ms978415.aspx

IIS 6.0 Considerations

On Microsoft Windows Server 2003, the IIS 6.0 architecture is different from IIS 5.0 on Windows 2000 Server. IIS 6.0 enables multiple processes to be used to host separate Web applications. This is shown in Figure 6.2.

Figure 6.2: IIS 6.0 architecture

IIS 6.0 includes a new HTTP listener (HTTP.Sys) that is implemented in the kernel. Requests are routed to one of the multiple worker process instances (W3wp.exe) that host ASP.NET applications and Web services.

Process Model

The primary difference between the ASP.NET architecture under Windows 2000 and Windows Server 2003 is that under Windows 2003, you can use separate IIS worker process instances to host Web applications

By default, the IIS worker process instances run using the NT Authority\NetworkService account. This account is a least-privileged local account that acts as the computer account over the network. A Web application that runs in the context of the Network Service account presents the computer's credentials to remote servers for authentication.

IIS 6.0 also supports a backwards-compatibility mode that supports the IIS 5.0 ASP.NET worker process model.

Kernel Mode Caching

If you deploy your application on Windows Server 2003, ASP.NET pages automatically benefit from the IIS 6.0 kernel cache. The kernel cache is managed by the HTTP.sys kernel-mode device driver. This driver handles all HTTP requests. Kernel mode caching may produce significant performance gains because requests for cached responses are served without switching to user mode.

The following default setting in the Machine.config file ensures that dynamically generated ASP.NET pages can use kernel mode caching, subject to the requirements listed below.

<httpRunTime enableKernelOutputCache="true" . . ./>

Dynamically generated ASP.NET pages are automatically cached subject to the following restrictions:

• Pages must be retrieved by using HTTP GET requests. Responses to HTTP POST requests are not cached in the kernel.
• Query strings are ignored when responses are cached. If you want a request for http://contoso.com/myapp.aspx?id=1234 to be cached in the kernel, all requests for http://contoso.com/myapp.aspx are served from the cache, regardless of the query string.
• Pages must have an expiration policy. In other words, the pages must have an Expires header.
• Pages must not have VaryByParams.
• Pages must not have VaryByHeaders.
• The page must not have security restrictions. In other words, the request must be anonymous and not require authentication. The HTTP.sys driver only caches anonymous responses.
• There must be no filters configured for the W3wp.exe file instance that are unaware of the kernel cache.

More Information

For more information about IIS 6.0 and kernel caching, see the IIS 6.0 Resource Kit at http://www.microsoft.com/downloads/details.aspx?FamilyID=80a1b6e6-829e-49b7-8c02-333d9c148e69&DisplayLang=en.

Web Gardens

By default, ASP.NET uses all CPUs available. In Web garden mode, ASP.NET creates one process per CPU. Each process creates an affinity to a single CPU. Web gardens offer an addition layer of reliability and robustness. If a process crashes, there are other processes that still service incoming requests.

Web gardens may perform better under the following scenarios:

• Your application uses STA objects heavily.
• Your application accesses a pool of resources that are bound by the number of processes. For example, a single process is restricted to using a particular number of resources.

To determine the effectiveness of Web gardens for your application, run performance tests, and then compare your results with and without Web gardens. Typically, in the two scenarios that are described in this section, you are likely to notice a greater benefit with servers that contain four or eight CPUs.

Note   Do not use the in-process session state store or any technique that causes process affinity if Web gardens are enabled.

IIS 6.0 vs. the ASP.NET Process Model

By default, the ASP.NET Process Model is not enabled in IIS 6.0. If you enable Web gardens, you may adversely affect the performance of the garbage collector. The performance of the garbage collector may be affected because the server version of the garbage collector is still used while bound to a single CPU. The disadvantage is that this creates one worker process per CPU. Because there is a worker process for each CPU, additional system resources are consumed.

Enabling Web Gardens by Using IIS 6.0

You can enable Web gardens in IIS 6.0 by using the Internet Information Services Manager. To do so, follow these steps:

1. Right-click the application pool that you want enable Web gardening on, and then click Properties.
2. Click the Performance tab.
3. In the Web garden section, specify the number of worker processes that you want to use.

Enabling Web Gardens by Using the ASP.NET Process Model

In the <processModel> section of the Machine.config file, set the webGarden attribute to true, and then configure the cpuMask attribute as follows.

<processModel webGarden="true" cpuMask="0xffffffff" />

Configuring the cpuMask Attribute

The cpuMask attribute specifies the CPUs on a multiprocessor server that are eligible to run ASP.NET processes. By default, all CPUs are enabled and ASP.NET creates one process for each CPU. If the webGarden attribute is set to false, the cpuMask attribute is ignored, and only one worker process runs. The value of the cpuMask attribute specifies a bit pattern that indicates the CPUs that are eligible to run ASP.NET threads. Table 6.3 shows some examples.

Table 6.3: Processor Mask Bit Patterns

More Information

For more information about how to use ASP.NET Web gardens, see Knowledge Base article 815156, "HOW TO: Restrict ASP.NET to Specific Processors in a Multiprocessor System," at http://support.microsoft.com/default.aspx?scid=kb;en-us;815156.

Garbage Collector Configuration Flag

By default, if you have a multiple processor server, the server GC is loaded. If you have a single processor server, the workstation GC is loaded. At the time of writing, .NET Framework version 1.1 Service Pack 1 (SP1) provides a switch that enables you to configure whether ASP.NET loads the server or the workstation GC. You can use this switch to configure ASP.NET to load the workstation GC even on a multiple-processor server.

When to Use the Workstation GC

If you host several isolated worker processes on Windows Server 2003 on a multiprocessor computer, use the workstation GC and nonconcurrent mode.

The server GC is optimized for throughput, memory consumption, and multiprocessor scalability. However, using the workstation GC when you are running Windows Server 2003 on a multiprocessor server can dramatically reduce the memory used per-worker process, and greatly increases the number of isolated worker processes that you can host. Disabling concurrent garbage collection further increases the number of isolated worker processes that you can run.

Configuring the Workstation GC

To configure ASP.NET to use the workstation GC, add the following configuration to the Aspnet.config file. The Aspnet.config file is in the same directory as the Aspnet_isapi.dll file.

<configuration>

  <runtime>

    <gcServer enabled="false"/>

    <gcConcurrent enabled="false"/>

  </runtime>

</configuration>

More Information

For more information about garbage collection in general and about server and workstation GCs, see "Garbage Collection Explained" in Chapter 5, "Improving Managed Code Performance"

Deployment Considerations

Physical deployment plays a key role in determining the performance and scalability characteristics of your application. Unless you have a compelling reason to introduce a remote middle tier, you should deploy your Web application's presentation, business, and data access layers on the Web server. The only remote hop should be the hop to the database. This section discusses the following key deployment considerations:

• Avoid unnecessary process hops.
• Understand the performance implications of a remote middle tier.
• Short circuit the HTTP pipeline.
• Configure the memory limit.
• Disable tracing and debugging.
• Ensure content updates do not cause additional assemblies to be loaded.
• Avoid XCOPY under heavy load.
• Consider precompiling pages.
• Consider Web garden configuration.
• Consider using HTTP compression.
• Consider using perimeter caching.

Avoid Unnecessary Process Hops

Although process hops are not as expensive as machine hops, you should avoid process hops where possible. Process hops cause added overhead because they require interprocess communication (IPC) and marshaling. For example, if your solution uses Enterprise Services, use library applications where possible, unless you need to put your Enterprise Services application on a remote middle tier.

Understand the Performance Implications of a Remote Middle Tier

If possible, avoid the overhead of interprocess and intercomputer communication. Unless your business requirements dictate the use of a remote middle tier, keep your presentation, business, and data access logic on the Web server. Deploy your business and data access assemblies to the Bin directory of your application. However, you might require a remote middle tier for any of the following reasons:

• You want to share your business logic between your Internet-facing Web applications and other internal enterprise applications.
• Your scale-out and fault tolerance requirements dictate the use of a middle tier cluster or of load-balanced servers.
• Your corporate security policy mandates that you cannot put business logic on your Web servers.

If you do have to deploy by using a remote middle tier, ensure you recognize this early so that you can measure and test by using the same environment.

Short Circuit the HTTP Pipeline

The HTTP pipeline sequence is determined by settings in the Machine.config file. Put the modules that you do not use inside comments. For example, if you do not use Forms authentication, you should either put the entry for Forms authentication in the Machine.config file in a comment or, explicitly remove the entry in your Web.config file for a particular application. The following sample shows how to comment out an entry.

<httpModules>
<!-- <add name="FormsAuthentication"

      type="System.Web.Security.FormsAuthenticationModule"/> -->

</httpModules>

The following sample from a Web.config file shows how to remove the entry for a specific application.

<httpModules>

<remove name="FormsAuthentication" />

</httpModules>

If you have other applications on your Web server that are using the HTTP module that you do not use, remove the HTTP module from the Web.config file of the application. Do this instead of using comments to disable the HTTP module in the Machine.config file.

Configure the Memory Limit

Before you deploy your application, configure the memory limit. Configuring the memory limit ensures optimal ASP.NET cache performance and server stability.

More Information

For more information, see "Configure the Memory Limit" in the "Caching Guidelines" section of this chapter.

Disable Tracing and Debugging

Before you deploy your application, disable tracing and debugging. Tracing and debugging may cause performance issues. Tracing and debugging are not recommended while your application is running in production.

Disable tracing and debugging in the Machine.config and Web.config files, as shown in the following sample.

<configuration>

  <system.web>

    <trace enabled="false" pageOutput="false" />

      <compilation debug="false" />

  </system.web>

</configuration>

Note   You may also want to verify that tracing and debugging are not enabled on individual pages. Individual page settings override the settings in the Web.config file.

More Information

For more information, see Knowledge Base article 815157, "HOW TO: Disable Debugging for ASP.NET Applications" athttp://support.microsoft.com/default.aspx?scid=kb;en-us;815157.

Ensure Content Updates Do Not Cause Additional Assemblies to Be Loaded

Problems may arise when updates to .aspx or .ascx pages occur without an application restart. Consider the following scenario. Assume you have five pages in a directory as follows.

\mydir

  Page1.aspx

  Page2.aspx

  Page3.aspx

  Page4.aspx

  Page5.aspx

When a page in the Mydir directory is first requested, all pages in that directory are compiled into a single assembly, as shown in the following sample.

Assembly1.dll {page1.aspx, page2.aspx, page3.aspx, page4.aspx}

If Page1.aspx is updated, a new single assembly is created for Page1.aspx. Now there are two assemblies, as shown in the following sample.

Assembly1.dll {page1.aspx, page2.aspx, page3.aspx, page4.aspx, page5.aspx}

Assembly2.dll {page1.aspx}

If Page2.aspx is updated, a new single assembly is created for Page2.aspx. Now there are three assemblies.

Assembly1.dll {page1.aspx, page2.aspx, page3.aspx, page4.aspx, page5.aspx}

Assembly2.dll {page1.aspx}

Assembly3.dll {page2.aspx}

To ensure that you do not experience this problem and generate multiple assemblies, follow these steps when you want to update content:

1. Remove the Web server from rotation.
2. Restart IIS.
3. Delete all files in the Temporary ASP.NET Files folder.
4. Request a single page in each directory to ensure that each directory is batch compiled.
5. Put the Web server back into rotation.

This approach to updating content also solves another problem. If a server is put into rotation before batch compilation is complete, some pages may be compiled as a single assembly. If another request is made during batch compilation for a page in the same directory that is being batch compiled, that page is compiled as a single assembly. Taking the Web server out of rotation and then putting it back in rotation helps you avoid this problem.

Avoid XCOPY Under Heavy Load

XCOPY deployment is designed to make deployment easy because you do not have to stop your application or IIS. However, for production environments you should remove a server from rotation, stop IIS, perform the XCOPY update, restart IIS, and then put the server back into rotation.

It is particularly important to follow this sequence under heavy load conditions. For example, if you copy 50 files to a virtual directory, and each file copy takes 100 milliseconds, the entire file copy takes 5 seconds. During that time, the application domain of your application may be unloaded and loaded more than once. Also, certain files may be locked by the XCOPY process (Xcopy.exe). If the XCOPY process locks certain files, the worker process and the compilers cannot access the files.

If you do want to use XCOPY deployment for updates, the .NET Framework version 1.1 includes the waitChangeNotification andmaxWaitChangeNotification settings. You can use these settings to help resolve the XCOPY issues described in this section.

Note   These settings are also available in a hotfix for .NET Framework version 1.0. For more information, see Knowledge Base article 810281, "Error Message: Cannot Access File AssemblyName Because It Is Being Used by Another Process," athttp://support.microsoft.com/default.aspx?scid=kb;en-us;810281.

The value of the waitChangeNotification setting should be based on the amount of time that it takes to use XCOPY to copy your largest file. ThemaxWaitChangeNotification setting should be based on the total amount of time that XCOPY uses to copy all the files plus a small amount of extra time.

More Information

For more information, see the following Knowledge Base articles:

• 810281, "Error Message: Cannot Access File AssemblyName Because It Is Being Used by Another Process" athttp://support.microsoft.com/default.aspx?scid=kb;en-us;810281
• 326355, "Deploy an ASP.NET Web Application Using Xcopy Deployment" at http://support.microsoft.com/default.aspx?scid=kb;en-us;326355

Consider Precompiling Pages

So that your users do not have to experience the batch compile of your ASP.NET files, you can initiate batch compiles by issuing one request to a page per directory and then waiting until the processor idles again before putting the Web server back into rotation. This increases the performance that your users experience, and it decreases the burden of batch compiling directories while handling requests at the same time.

Consider Web Garden Configuration

Consider using Web gardens if your application uses STA objects heavily or if your application accesses a pool of resources that are bound by the number of processes.

To determine the effectiveness of Web gardens for your application, run performance tests, and then compare the results with and without Web gardens.

Consider Using HTTP Compression

HTTP compression is supported by most modern browsers and by IIS. HTTP compression is likely to improve performance when the client accesses the Web server over a low bandwidth connection.

Consider Using Perimeter Caching

A perimeter network protects your intranet from intrusion by controlling access from the Internet or from other large networks. It consists of a combination of systems such as proxy servers, packet filtering, gateways, and other systems that enforce a boundary between two or more networks.

If your perimeter network includes a proxy server, consider enabling caching on your proxy server to improve performance.

Summary

This chapter discusses the common pitfalls and bottlenecks that can occur during ASP.NET application development. It shows you the steps you need to take to avoid and overcome these issues. By following the guidance and advice in this chapter, you can build extremely high performance ASP.NET applications.

Building high performance applications of any type requires you to consider performance from the outset. You have to create a sound architecture and design that takes into account any restrictions that might be imposed by your physical deployment environment. During development, you need to ensure that you adopt best practice coding techniques. You have to continually measure performance to ensure that your application operates within the boundaries determined by your performance objectives. Measuring should continue throughout the life cycle. Finally, at deployment time you have to consider the configuration of the environment that your application will run in.

Additional Resources

For more information, see the following resources:

• For a printable checklist, see "Checklist: ASP.NET Performance" in the "Checklists" section of this guide.
• Chapter 4 — Architecture and Design Review of a .NET Application for Performance and Scalability
• Chapter 13 — Code Review: .NET Application Performance
• Chapter 15 — Measuring .NET Application Performance
• Chapter 16 — Testing .NET Application Performance
• Chapter 17 — Tuning .NET Application Performance

For more information about IIS 6.0, see the following resource:

• "Web and Application Server Infrastructure - Performance and Scalability" at http://www.microsoft.com/downloads/details.aspx?FamilyID=B48F46A9-3971-4DA6-8A90-B591DD495ECC&displaylang=en&displaylang;=en