Bad performance with distributed search when sort contains relevancy queries
----------------------------------------------------------------------------
Key: SOLR-3104
URL: https://issues.apache.org/jira/browse/SOLR-3104
Project: Solr
Issue Type: Improvement
Components: search
Affects Versions: 3.6
Reporter: XJ Wang
Priority: Critical
So I found this issue when trying out distributed search with solr 3.5 and
noticed big performance degradation for some queries comparing to the single
box search.
After some query analysis and comparison, it turns out that shard queries with
"fsv=true" are much slower than the same queries w/o "fsv=true". Some examples
are like 1200ms vs 200ms (start=0, rows=30, hits<100).
>From the discussions with Yonik Seeley on solr mailing list, it may due to
>fact that I'm using lot of relevancy queries in sorting. But Solr is not
>retrieving those sort values efficiently .
This is critical for us and prevents us from moving to distributed search. I
believe users like our scenarios will also suffer from this issue. Any
patch/idea is welcomed.
Quote from Yonik Seeley on solr-user mailing list:
"OK, so basically it's slow because functions with embedded relevancy
queries are "forward only" - if you request the value for a docid
previous to the last, we need to reboot the query (re-weight, ask for
the scorer, etc). This means that for your 30 documents, that will
require rebooting the query about 15 times (assuming that roughly half
of the time the next docid will be less than the previous one).
Unfortunately there's not much you can do externally... we need to
implement optimizations at the Solr level for this."
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
