[
https://issues.apache.org/jira/browse/SOLR-16549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17637421#comment-17637421
]
David Smiley commented on SOLR-16549:
-------------------------------------
See SOLR-6820 which resulted in changing a default of 256 to 65536. I'd need
to review it more thoroughly to make a more informed decision. Also, an array
of longs would be more efficient than an array of {{VersionBucket}} instances
considering that a VersionBucket is just a long. But a TimedVersionBucket is
bigger :-/
Also, maybe this bucket array could be in some TTL cache such that a core that
is not doing indexing doesn't have it. A non-leader non-NRT core would have no
need for this I think.
> Solr core memory usage study
> ----------------------------
>
> Key: SOLR-16549
> URL: https://issues.apache.org/jira/browse/SOLR-16549
> Project: Solr
> Issue Type: Task
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Noble Paul
> Priority: Major
>
> This is a ticket to study and compare memory usage of cores
> The study is performed on the 9.1 branch
> a simple tests reveals the following. The conclusion is that on an average a
> Solr Core consumes ~3MB of memory for an empty index
> {{cores heap usage }}
> {{---------------------- }}
> {{0 36,623,632 }}
> {{10 68,259,704 }}
> {{20 96,774,456 }}
> {{70 240,501,176 }}
> {{120 379,437,928 }}
> {{200 602,049,784 }}
> {{300 904,252,536 }}
> {{400 1,181,341,200 }}
> {{500 1,455,186,296 }}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
