Group.ngroup query slower with docValues

[vicky desai <vi...@germinait.com>]

Hi,

I am planning to use the docValues feature of Solr. I have added "docValues=
true" parameter to a few fields in my schema on which there is heavy
faceting and grouping query involved. While I noticed a considerable
improvement in faceting queries. I didnt get any improvements in grouping
query and instead places where i used "group.ngroups=true" in my query there
was around 3-4 times degradation in performance.

Any know issues or possible workarounds for this case



--
View this message in context: http://lucene.472066.n3.nabble.com/Group-ngroup-query-slower-with-docValues-tp4099581.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Group.ngroup query slower with docValues

[vicky desai <vi...@germinait.com>]

Hi Shawn,

Thanks for the reply. But the issue you pointed out talk about general
performance issue of ngroups. However what i noticed is that after using
docValues the performance of group.ngroups had degraded about 2-3 times.
This is stopping me from using docValues which otherwise in case of faceting
and grouping gives tremendous improvement



--
View this message in context: http://lucene.472066.n3.nabble.com/Group-ngroup-query-slower-with-docValues-tp4099581p4099721.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Group.ngroup query slower with docValues

[Shawn Heisey <so...@elyograg.org>]

On 11/6/2013 6:28 AM, vicky desai wrote:
> I am planning to use the docValues feature of Solr. I have added "docValues=
> true" parameter to a few fields in my schema on which there is heavy
> faceting and grouping query involved. While I noticed a considerable
> improvement in faceting queries. I didnt get any improvements in grouping
> query and instead places where i used "group.ngroups=true" in my query there
> was around 3-4 times degradation in performance.
> 
> Any know issues or possible workarounds for this case

The group.ngroups=true feature has bad performance.  It's a fact that's
well known.  There is an issue to improve the situation, but mostly it's
just a feature request with no patch and no significant ideas.

https://issues.apache.org/jira/browse/SOLR-2963

There has been no activity on the issue since January, but the first
comment lists what may be a workaround that improves performance when
grouping on a single field.

I personally have never used the grouping feature and don't know if the
workaround is solid.  This is all the information I could find.

Thanks,
Shawn