From: Steven McPhillips <smcphillips@nl...> - 2008-01-18 23:16:51
If your problem is linked to faceting, this isn't so much a java tuning =
/ hardware issue as a data issue. The more unique terms you have, the =
more intensive / expensive faceting becomes. You might want to consider =
creating some normalised facet-only indexes for use in faceting. For our =
subject faceting, we only look at 6xx$a, whereas the proper subject =
index has all subfields.
We run SOLR with around 4mil records without issue in a 3.5Gb jvm. We've =
created quite a few additional indexes though, and I think our space =
requirements are now up around the 12Gb mark.
From: vufind-general-bounces@... on behalf of Antonio =
Sent: Sat 1/19/2008 2:49 AM
Subject: Re: [VuFind-General] VuFind and large collections
At Princeton, my first test was on a standard newish dell desktop that
we use for public machines. IE not very powerful. We did import 4
million records without much of a problem. But we had a ton of problems
with the facet calls bogging down in Solr. Unfortunately, I'm not still
waiting server quality hardware as a replacement. Our system
administrator said I should be getting an older server previously used
for Voyager soon. He guessed 4+ processors 16-20 gig of RAM. I can't
really give you much insight yet though.
Princeton University Library
[mailto:vufind-general-bounces@...] On Behalf Of
Sent: Friday, January 18, 2008 9:45 AM
Subject: [VuFind-General] VuFind and large collections
At the VuFind meeting during ALA, at least two institutions mentioned
using VuFind for large collections. One was Princeton and I think the
second was Western Michigan. Does anyone know the exact numbers?=20
And/or if the actual attendees are reading this list would you please
contact me? I'd like to get some idea of the hardware needed to support
discovery for the 12 million records in the Yale library.
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
VuFind-General mailing list