有状态和无状态流（火花）

Question

I know the difference between stateful and stateless streaming processes. I read that Storm is stateless, while Trident is stateful. I also read that Hadoop (for batch processing) is stateful and that Spark can compute stateful operations.

Can someone clarify on each of these? Specifically,

1. Can spark do both stateful and stateless operations?
2. What does it mean that Hadoop is stateful since we talk only about batch processing when it comes to Hadoop.
   3. How does Apache Storm handle stateful streams? (using Trident?)

Answer 1

1-yes spark have stateful and stateless operations Stateful Stream Processing with mapWithState

for more information see https://databricks.com/blog/2016/02/01/faster-stateful-stream-processing-in-apache-spark-streaming.html

2-Hadoop is stateful because its read once