Hi,

We are using Kafka-Spout and ES-Bolt in our storm topology.
Kafka-Spout is very fast while ES-Bolt is not so fast because ES cannot
index at the same speed.

In such a case, how many workers per supervisor is recommended?

1) Multiple workers per supervisor with each worker having few threads?
2) Single worker per supervisor with several threads in the worker?

I am assuming the latter because context switching among threads is less
costly than that between the processes.
Or does it really not matter in this case?


--SG

Reply via email to