SQL SERVER内存按存放数据的类型,大概可以分为三类:
1、buffer pool,存放数据页面的缓冲区,sql server数据都是存放在一个个8K的页面里,当用户需要使用这个页面上的数据时,都是把整个页面加载到内存的buffer pool区缓存起来。
2、各类consumer:
connect:SQL SERVER为每一个客户端连接分配一块内存,用来存储连接的信息,以及发过来的指令和缓存指令结果待待客户端取走
无数据:表、存储过程、索引等的元数据
锁:SQL SERVER中锁是稀有资源,会占用大量内存
Query plan:缓存SQL的执行计划
Optimizer:生成执行 计划过程中需要使用内存
3、线程内存:sql server会为每个线程分配0.5M的内存,用来存放线程的数据结构和相关信息
sql1:
-- 查询SqlServer总体的内存使用情况 select type , sum(virtual_memory_reserved_kb) VM_Reserved , sum(virtual_memory_committed_kb) VM_Commited , sum(awe_allocated_kb) AWE_Allocated , sum(shared_memory_reserved_kb) Shared_Reserved , sum(shared_memory_committed_kb) Shared_Commited --, sum(single_pages_kb) --SQL2005、2008 --, sum(multi_pages_kb) --SQL2005、2008 from sys.dm_os_memory_clerks group by type order by type -- 查询当前数据库缓存的所有数据页面,哪些数据表,缓存的数据页面数量 -- 从这些信息可以看出,系统经常要访问的都是哪些表,有多大? select p.object_id, object_name=object_name(p.object_id), p.index_id, buffer_pages=count(*) from sys.allocation_units a, sys.dm_os_buffer_descriptors b, sys.partitions p where a.allocation_unit_id=b.allocation_unit_id and a.container_id=p.hobt_id and b.database_id=db_id() group by p.object_id,p.index_id order by buffer_pages desc -- 查询缓存的各类执行计划,及分别占了多少内存 -- 可以对比动态查询与参数化SQL(预定义语句)的缓存量 select cacheobjtype , objtype , sum(cast(size_in_bytes as bigint))/1024 as size_in_kb , count(bucketid) as cache_count from sys.dm_exec_cached_plans group by cacheobjtype, objtype order by cacheobjtype, objtype -- 查询缓存中具体的执行计划,及对应的SQL -- 将此结果按照数据表或SQL进行统计,可以作为基线,调整索引时考虑 -- 查询结果会很大,注意将结果集输出到表或文件中 SELECT usecounts , refcounts , size_in_bytes , cacheobjtype , objtype , TEXT FROM sys.dm_exec_cached_plans cp CROSS APPLY sys.dm_exec_sql_text(plan_handle) ORDER BY objtype DESC ; GO
-- 查看SQL语句占用多大内存: SELECT s2.dbid, s1.sql_handle, (SELECT TOP 1 SUBSTRING(s2.text,statement_start_offset / 2+1 , ( (CASE WHEN statement_end_offset = -1 THEN (LEN(CONVERT(nvarchar(max),s2.text)) * 2) ELSE statement_end_offset END) - statement_start_offset) / 2+1)) AS sql_statement, execution_count, plan_generation_num, last_execution_time, total_worker_time, last_worker_time, min_worker_time, max_worker_time, total_physical_reads, last_physical_reads, min_physical_reads, max_physical_reads, total_logical_writes, last_logical_writes, min_logical_writes, max_logical_writes FROM sys.dm_exec_query_stats AS s1 CROSS APPLY sys.dm_exec_sql_text(sql_handle) AS s2 WHERE s2.objectid is null ORDER BY s1.sql_handle, s1.statement_start_offset, s1.statement_end_offset;
sql2:
select OBJECT_NAME(object_id) 表名,COUNT(*) 页数,COUNT(*)*8/1024.0 Mb from sys.dm_os_buffer_descriptors a,sys.allocation_units b,sys.partitions c where a.allocation_unit_id=b.allocation_unit_id and b.container_id=c.hobt_id and database_id=DB_ID() group by OBJECT_NAME(object_id) order by 2 desc
参考:
http://www.cnblogs.com/zhaoguan_wang/p/4602866.html
http://blog.csdn.net/shutao917/article/details/51444424
http://blog.csdn.net/burgess_liu/article/details/52813727
http://blog.csdn.net/burgess_liu/article/details/17739725
http://blog.csdn.net/burgess_liu/article/details/17733149