Hi Cameron,

You'll need to upgrade to Solr 1.4 as the 1.3 method of faceting
is quite slow (i.e. intersecting bitsets). 1.4 uses
UnInvertedField which caches the terms per doc and
iterates/counts them. The 1.3 method is slow because for every
term (i.e. unique field value) there needs to be a bitset.

-J

On Tue, Aug 18, 2009 at 2:17 PM, CameronL<cameron.develo...@gmail.com> wrote:
>
> Our current search is faceting on a single integer field. The field is
> multi-valued.
>
> facet=true
> facet.mincount=1
> facet.limit=-1
> facet.field=fieldA
>
> The number of unique values in our index for fieldA is around 8000, and a
> typical query can return about 500 counts. A typical single document can
> have anywhere from 5 to 20 values for fieldA. The performance we are getting
> for this implementation is pretty acceptable (under 2 seconds).
>
> Now, we are trying to add in a 2nd facet, also an integer and also
> multi-valued.
>
> facet=true
> facet.mincount=1
> facet.limit=-1
> facet.field=fieldA
> facet.field=fieldB
>
> The number of unique values in our index for fieldB is around 100k, but a
> typical query still only returns about 400 counts. However, a single
> document will only have 5 or 6 values for fieldB. The performance of our
> queries dropped significantly (about 15-20 seconds per query!).
>
> I'm unable to figure out why there is such a significant drop in performance
> here. Is it the fact that there are more than 10x more possible unique
> values for fieldB? Hopefully I have provided enough info above, but do any
> of these strike you as a big contributing factor to the drop in performance?
>
> We are currently using Solr 1.3 and upgrading to 1.4 will not be an option
> until it is finalized.
>
> Thanks for the help.
> --
> View this message in context: 
> http://www.nabble.com/Faceting-Performance-Factors-tp25033622p25033622.html
> Sent from the Solr - User mailing list archive at Nabble.com.
>
>

Reply via email to