Hadoop MapReduce:
MapReduce在每次执行的时候都要从磁盘读数据，计算完毕后都要把数据放到磁盘
spark map reduce:
?
?
RDD is everything for dev:
Basic Concepts:
Graph RDD:
Spark Runtime:
schedule:
Depency Type:
Scheduler Optimizations：
Event Flow:
Submit Job:
New Job Instance:
Job In Detail:
executor.launchTask:
Standalone：
?
?
Work Flow:
Standalone detail:
Driver application to Clustor:
Worker Exception:
Executor Exception:
Master Exception:
Master HA:
hadoop&spark mapreduce对比 & 框架设计和理解
原文：http://stark-summer.iteye.com/blog/2176583
