o**********e 发帖数: 18403 | 1 【 以下文字转载自 JobHunting 讨论区 】
发信人: dongfeiwww (在路上), 信区: JobHunting
标 题: Hadoop Spark 学习小结[2014版]
关键字: Spark,Big Data,Hadoop
发信站: BBS 未名空间站 (Sun Aug 17 12:28:34 2014, 美东)
基于很多朋友希望更新这个学习资料,我就尽力按我的积累补充,are you ready, 享
受技术饕餮大餐
#Hadoop
Hadoop社区依然发展迅速,2014年推出了2.3,2.4, 2.5 的社区版本,比如增强
Resource Manager HA,
YARN Rest API, ACL on HDFS...
http://hadoop.apache.org/releases.html
根据我的观察,主要更新在Yarn,HDFS,而Mapreduce几乎停滞了,还有一些feature
属于安全,稳定可靠性一方面也是比较稳定了,但也可以说是瓶颈了。
http://hadoop.apache.org/who.html
这个是Hadoop project member and committee, 里面好多来自Hortonworks,不过也有
不少国人加入了,都是未来的希望啊。
# Spark
Spark今年大放溢彩,Spark简单说就是内存计算(或者迭代式计算,DAG计算,流式计算
)框架,
MapReduce因效率低下大家经常嘲笑, Spark号称性能超Hadoop百倍,算法实现仅有其
1/10或1/100 Reynold 作为Spark核心开发者,介绍
http://www.csdn.net/article/2013-04-26/2815057-Spark-Reynold
http://www.csdn.net/article/2014-08-07/2821098-6-sparkling-feat
起源于2010年Berkeley AMPLab,发表在hotcloud上
https://www.usenix.org/legacy/events/hotcloud10/tech/full_papers/Zaharia.pdf
是一个从学术界到工业界的成功典范,所以也吸引了顶级VC:Andreessen Horowitz的
注资
BTW: 这个实验室非常厉害,做大数据,云计算,跟工业界结合很紧密,比如Twitter也
Berkeley开了门课程
http://blogs.ischool.berkeley.edu/i290-abdt-s12/
还有个BDAS (Bad Ass)引以为傲: https://amplab.cs.berkeley.edu/software/
http://gigaom.com/2014/08/02/the-lab-that-created-spark-wants-t
在2013年,这些大牛出动把Berkeley AMPLab的人拉出去成立了Databricks,半年就做
了2次summit
参会1000人,引无数Hadoop大佬尽折腰,大家看一下Summit的sponsor,所有hadoop厂
商全来了,并且各个技术公司也在巴结,cloudrea, hortonworks, mapr, datastax,
yahoo, ooyala, 根据CTO说 Spark新增代码量活跃度今年远远超过了Hadoop本身
,马上要推出商业化产品Cloud
Spark 核心人物
Ion Stoica :http://www.cs.berkeley.edu/~istoica/ Berkeley教授,AMPLab 领军
Matei Zaharia:http://people.csail.mit.edu/matei/ 天才,MIT助理教授
Reynold Xin: http://www.eecs.berkeley.edu/~rxin/
Haoyuan Li: http://www.cs.berkeley.edu/~haoyuan/
http://www.wired.com/2013/06/yahoo-amazon-amplab-spark/all/
其实起名字也很重要,Spark就占了先机,CTO说 Where there's spark, there's fire!
http://inside-bigdata.com/2014/07/15/theres-spark-theres-fire-s
Spark核心数据结构:
Resilient Distributed Datasets: A Fault-Tolerant Abstraction for
In-Memory Cluster Computing
https://www.usenix.org/system/files/conference/nsdi12/nsdi12-final138.pdf
Spark目前是1.0.2最新版本:https://spark.apache.org/docs/1.0.2/
目前还有一些子项目,比如 Spark SQL, Spark Streaming, MLLib, Graphx
如;http://spark.apache.org/streaming/
工业界也引起广泛兴趣,国内Taobao, baidu也开始使用:
https://cwiki.apache.org/confluence/display/SPARK/Powered+By+Spark
Apache Spark支持4种分布式部署方式,分别是Amazon EC2, standalone、spark on
mesos和 spark on YARN
比如AWS:
http://www.getblueshift.com/blog/?p=56
至于如何入门,还是得好好看官方文档,上面有入门,搭建环境,Summit上的视频也是
http://spark-summit.org/2014/
也有个training视频:
http://spark-summit.org/2014/training
今年的Summit 回顾
http://www.csdn.net/article/2014-07-17/2820713
今年最叫好的demo是Dtabricks Cloud, 把twitter上面实时收集的数据做作为machine
learning素材,
用类似IPython notebook,可视化呈现惊艳,而搭建整个sampling系统就花了20分钟!
http://databricks.com/cloud
CSDN上面也有个Spark专栏,大家可以多去看看
spark.csdn.net
还有一些第三方的项目基于Spark上面
Web interactive UI on Hadoop/Spark: http://gethue.com/
Spark on cassandra:
http://planetcassandra.org/getting-started-with-apache-spark-an
https://github.com/datastax/spark-cassandra-connector
http://tuplejump.github.io/calliope/
H2O + Spark:
http://databricks.com/blog/2014/06/30/sparkling-water-h20-spark
Shark - Hive and SQL on top of Spark
MLbase - Machine Learning research project on top of Spark
BlinkDB - a massively parallel, approximate query engine built on top of
Shark and Spark
GraphX - a graph processing & analytics framework on top of Spark (GraphX
has been merged into Spark 0.9)
Apache Mesos - Cluster management system that supports running Spark
Tachyon - In memory storage system that supports running Spark
Apache MRQL - A query processing and optimization system for large-scale,
distributed data analysis, built on top of Apache Hadoop, Hama, and Spark
OpenDL - A deep learning algorithm library based on Spark framework. Just
kick off.
SparkR - R frontend for Spark
Spark Job Server - REST interface for managing and submitting Spark jobs on
the same cluster |
|