[Flink基础]-- 1.4 与其他实时计算框架对比

对比其他计算框架

Spark Storm Flink
Streaming Model Micro-batch Native Native
Fault Tolerance RDD based checkpointing Record ACKs checkpoint
Latency Medium Very Low Low
Throughout High Low High
Guarantees At-least-once At Least Once,Exactly Once At Least Once,Exactly Once
State Management Dstream Not build-in Statedful Operators
EcoSystem Active、Perfect Not Active、Not Perfect Active、Function Perfect

参考

  • Jira:https://issues.apache.org/jira/projects/FLINK
  • Wiki:https://cwiki.apache.org/confluence/display/FLINK/Powered+by+Flink
  • Github:https://github.com/apache/flink
  • Community:https://data-artisans.com/blog/category/flink-community
  • spark 实现 exactly once
    http://shzhangji.com/cnblogs/2017/08/01/how-to-achieve-exactly-once-semantics-in-spark-streaming/

你可能感兴趣的:(Spark,Flink)