[
https://issues.apache.org/jira/browse/CASSANDRA-9237?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14692505#comment-14692505
]
Jonathan Ellis commented on CASSANDRA-9237:
-------------------------------------------
I see two options for 3.0, both of which are better than the status quo:
# Reduce the small-message threshold
# Go back to the old heuristic of putting gossip and internal responses on a
separate socket
The problem with #2 in the past was that read responses, which are quite large,
got jumbled in too. (REQUEST_RESPONSE is too large an umbrella.) We could
split those out to their own verb, but it's not clear to me that putting write
acks on the "low traffic" socket is a win.
Any redefinition of liveness or heartbeat generation belongs in a new ticket
and is something of an open-ended research project with no clear answers imo.
> Gossip messages subject to head of line blocking by other intra-cluster
> traffic
> -------------------------------------------------------------------------------
>
> Key: CASSANDRA-9237
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9237
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Ariel Weisberg
> Assignee: Ariel Weisberg
> Fix For: 3.0.0 rc1
>
>
> Reported as an issue over less than perfect networks like VPNs between data
> centers.
> Gossip goes over the small message socket where small is 64k which isn't
> particularly small. This is done for performance to keep most traffic on one
> hot socket.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
