cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sebastian Estevez (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CASSANDRA-11516) Make max number of streams configurable
Date Wed, 06 Apr 2016 21:59:25 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-11516?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Sebastian Estevez updated CASSANDRA-11516:
------------------------------------------
    Description: 
Today we default to num cores. In large boxes (many cores), this is suboptimal as it can generate
huge amounts of garbage that GC can't keep up with.

Usually we tackle issues like this with the streaming throughput levers but in this case the
problem is CPU consumption by StreamReceiverTasks specifically in the IntervalTree build --
https://github.com/apache/cassandra/blob/cassandra-2.1.12/src/java/org/apache/cassandra/utils/IntervalTree.java#L257

We need a max number of parallel streams lever to hanlde this.

  was:
Today we default to num cores. In large boxes (many cores), this is suboptimal as it can generate
huge amounts of garbage that GC can't keep up with.

Usually we tackle issues like this with the streaming throughput levers but in this case the
problem is CPU consumption by StreamReceiverTasks specifically in the IntervalTree build --
https://github.com/apache/cassandra/blob/cassandra-2.1.12/src/java/org/apache/cassandra/utils/IntervalTree.java#L257


> Make max number of streams configurable
> ---------------------------------------
>
>                 Key: CASSANDRA-11516
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11516
>             Project: Cassandra
>          Issue Type: New Feature
>            Reporter: Sebastian Estevez
>
> Today we default to num cores. In large boxes (many cores), this is suboptimal as it
can generate huge amounts of garbage that GC can't keep up with.
> Usually we tackle issues like this with the streaming throughput levers but in this case
the problem is CPU consumption by StreamReceiverTasks specifically in the IntervalTree build
-- https://github.com/apache/cassandra/blob/cassandra-2.1.12/src/java/org/apache/cassandra/utils/IntervalTree.java#L257
> We need a max number of parallel streams lever to hanlde this.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message