Sushant Kumar created STORM-431:
-----------------------------------
Summary: Topology back pressure using netty transport
Key: STORM-431
URL: https://issues.apache.org/jira/browse/STORM-431
Project: Apache Storm (Incubating)
Issue Type: Improvement
Reporter: Sushant Kumar
There are several community threads that talk about automatically managing back
pressure based on send and receive buffers between the spouts and the bolts.
This works perfectly in a single worker scenario. Supposedly, it doesn't work
across multiple workers because of zmq deficiency. Now that netty has replaced
zmq as a network transport, is it possible to officially support buffer lengths
as a flow control mechanism in non-reliable topologies with high throughput
needs?
--
This message was sent by Atlassian JIRA
(v6.2#6252)