[
https://issues.apache.org/jira/browse/SOLR-6035?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13985896#comment-13985896
]
Marvin Justice commented on SOLR-6035:
--------------------------------------
We see average indexing times drop from 9.2 ms to 6.7 ms. The overall timing
distribution has a smaller tail for the patched version with stddev dropping
from 15.4 ms to 9.0 ms.
Good point about replication, I didn't actually test that. This work was done
on our "alpha" cluster which is 64 shards on 16 nodes spread across 4 machines
but with no replication.
> CloudSolrServer directUpdate routing should use getCoreUrl
> -----------------------------------------------------------
>
> Key: SOLR-6035
> URL: https://issues.apache.org/jira/browse/SOLR-6035
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Reporter: Marvin Justice
> Assignee: Joel Bernstein
> Priority: Minor
> Labels: performance
> Attachments: SOLR-6035.patch
>
>
> In a multisharded node environment we were seeing forward-to-leader hops when
> using CloudSolrServer directUpdate (with the hop being on the same node) .
> Consequently, there was no improvement in indexing performance over
> non-directUpdate.
> Changing buildUrlMap to use getCoreUrl eliminated the extra hop and now we
> see a dramatic improvement in performance.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
