Mickael,
I just wonder you have considered BlockJoin? it performs much better than
query time approaches
http://blog.griddynamics.com/2012/08/block-join-query-performs.html ,but
faceting hasn't been implemented for it yet.
On Tue, Jan 15, 2013 at 2:01 PM, Mickael Magniez
wrote:
> Hi,
>
> Retry
Hi,
Retry on a better machine (2CPU, 8GB RAM, 1.5GB for java half used according
admin interface) still have the same issue.
It seems to grow with matches count : with a search matching 100k documents,
it takes 700ms, vs 70ms without ngroup (CPU is 100% during request)
For information, my index
group.ngroups=true is always going to be somewhat expensive, but in your
case it seems more expensive than I would expect. You should check to see
that you have enough Java JVM heap to hold more of the index and to avoid
any excessive GCs.
-- Jack Krupansky
-Original Message-
From: M
