十位一线专家分享Spark现状与未来----峰会摘录

CSDN大数据技术:

十位一线专家分享Spark现状与未来(一)

十位一线专家分享Spark现状与未来(二)

十位一线专家分享Spark现状与未来(三)


部分摘录:

加州大学伯克利分校AMP实验室博士Matei Zaharia:Spark的现状和未来 ----(Matei Zaharia是加州大学伯克利分校AMP实验室博士研究生,Databricks公司的联合创始人兼现任CTO。Zaharia致力于于大规模数据密集型计算的系统和算法。研究项目包括:Spark、Shark、Multi-Resource Fairness、MapReduce Scheduling、SNAP Sequence Aligner)

Spark是发源于美国加州大学伯克利分校AMPLab的集群计算平台,立足于内存计算,从多迭代批量处理出发,兼收并蓄数据仓库、流处理和图计算等多种计算范式,是罕见的全能选手。

Project History:

  Spark started as research project in 2009

  Open sourced in 2010

  Growing community since

  Entered Apache lncubator in June 2013

Release Growth:

  Spark 0.6 ---- Java API、Maven、standalone mode ,17 contributors

  Spark 0.7 ---- Python API、Spark Streaming ,31 contributors

  Spark 0.8 ---- YARN、MLlib、monitoring UI ,67 contributors ---- High availability for standalone mode (0.8.1)

  Spark 0.9 ---- Scala 2.10 support、Configuration system、Spark Streaming improvement

Projects Bulit on Spark:

  Shark(SQL)、Spark Streaming(real-time)、GraphX(graph)、MLbase(machine learning)

十位一线专家分享Spark现状与未来----峰会摘录

Databricks公司CEO Ion Stoica:将数据转化为价值 ----(Ion Stoica是UC Berkeley计算机教授,AMPLab共同创始人,弹性P2P协议Chord、集群内存计算框架Spark、集群资源管理平台Mesos都出自他)

Turning Data into Value

What do We Need?

  interactive queries(交互式查询) ---- enable faster decision

  Queries on streaming data(基于数据流的查询) ---- enable decisions on real-time data ---- Eg:fraud detection(欺诈检测)、detect DDoS attacks(检测DDoS攻击)

  Sophisticated data processing(复杂的数据处理) ---- enable "better" decision

十位一线专家分享Spark现状与未来----峰会摘录

Our Goal:

  Support batch、Streaming、and interactive computation(批处理、流处理、交互计算)...... in a unified framework

  Easy to develop sophisticated algorithms(e.g..,graph,ML algos)

十位一线专家分享Spark现状与未来----峰会摘录

Big Data Challenge:Time 、Money 、Answer Quality

十位一线专家分享Spark现状与未来----峰会摘录

处理速度与精确性的权衡:反比

十位一线专家分享Spark现状与未来----峰会摘录

Tim Tully :集成Spark/Shark到雅虎数据分析平台

Sharethrough数据专家Ryan Weald:产品化Spark流媒体

Keys to Fault Tolerance:

  Receive fault tolerance ---- Use Actors with supervisor、Use self healing connection pools

  Monitoring job progress

RDDs:弹性分布式数据集

  Low latency & Scale (低延时&大规模)

  iterative and Interactive computation (迭代式和交互式计算)

Databricks创始人Patrick Wendell:理解Spark应用程序的性能 ---- (专注于大规模数据密集型计算。致力于Spark的性能基准测试,同时是spark-perf的合著者。此次峰会他就Spark 深度挖掘、UI概述和测试设备、普通性能和错误)

Summary of Components:

  Tasks:Fundamental unit of work

  Stage:Set of tasks that run in parallel

  DAG:Logical graph of RDD operations

  RDD:Parallel dataset with partitions

Demo of perf UI ---- Problems:

  Scheduling and launching tasks

  Execution of tasks

  Writing data between stages

  Collecting results

Databricks客户端解决方案主管Pat McDonough:用Spark并行程序设计 ---- (从Spark的性能、组件等方面全面介绍Spark的各种优异性能)

十位一线专家分享Spark现状与未来----峰会摘录

十位一线专家分享Spark现状与未来----峰会摘录

十位一线专家分享Spark现状与未来----峰会摘录

十位一线专家分享Spark现状与未来----峰会摘录

UC Berkeley博士Tathagata Das:用Spark流实时大数据处理 ---- (什么是Spark流,为什么选择Spark流,其性能和容错机制)

十位一线专家分享Spark现状与未来----峰会摘录

十位一线专家分享Spark现状与未来----峰会摘录十位一线专家分享Spark现状与未来----峰会摘录

DStreams+RDDs=Power

Fault-tolerance:

  Batches of input data are replicated in memory for fault-tolerance

  Data lost due to worker failure,can be recomputed from replicated input data

  All transformations are fault-tolerant,and exactly-once transformations

Higher throughput than Storm:

  Spark Streaming:670K records/sec/node

  Storm:115K records/sec/node

Fast Fault Recovery:

  Recovers from faults/stragglers within 1 sec

Spark 0.9 in Jan 2014 ---- out of alpha

  Automated master fault recovery

  Performance optimizations

  Web UI,and better monitoring capabilities

    Cluster Manager UI ---- Standalone mode:<master>:8080

    Executor Logs ---- Stored by cluster manager on each worker

    Spark Driver Logs ---- Spark initializes a log4j when created ,Include log4j.properties file on the classpath

    Application Web UI ---- http://spark-application-host:4040 ---- For executor / task / stage / memory status,etc

上一篇:C#中string.format用法详解


下一篇:业余草基于JAVA的模块化开发框架JarsLink