Xavier, 100-700 QPS is still high. I'm guessing your 1 box won't handle that without sweating a lot (read: slow queries). Otis ---- Sematext :: http://sematext.com/ :: Solr - Lucene - Nutch Lucene ecosystem search :: http://search-lucene.com/
----- Original Message ---- > From: Xavier Schepler <xavier.schep...@sciences-po.fr> > To: solr-user@lucene.apache.org > Sent: Fri, April 23, 2010 11:53:23 AM > Subject: Re: What hardware do I need ? > > Le 23/04/2010 17:08, Otis Gospodnetic a écrit : > Xavier, > > > 0-1000 QPS is a pretty wide range. Plus, it depends on how good your > auto-complete is, which depends on types of queries it issues, among other > things. > 100K short docs is small, so that will all fit in RAM nicely, > assuming those other processes leave enough RAM for the OS to cache the > index. > > That said, you do need more than 1 box if you want > your auto-complete more fault tolerant. > > Otis > > ---- > Sematext :: > >http://sematext.com/ :: Solr - Lucene - Nutch > Lucene ecosystem > search :: > >http://search-lucene.com/ > > > > ----- Original > Message ---- > >> From: Xavier Schepler< > ymailto="mailto:xavier.schep...@sciences-po.fr" > href="mailto:xavier.schep...@sciences-po.fr">xavier.schep...@sciences-po.fr> >> > To: > href="mailto:solr-user@lucene.apache.org">solr-user@lucene.apache.org >> > Sent: Fri, April 23, 2010 11:01:24 AM >> Subject: What hardware do I > need ? >> >> Hi, >> > I'm > working with Solr 1.4. > My schema has about 50 fields. > > I'm > >> using full text search in short strings (~ > 30-100 terms) and facetted >> search. >> > > My index will have 100 000 documents. > > The number of > requests > >> per second will be low. Let's say > between 0 and 1000 because of >> auto-complete. >> > > Is a standard server (3ghz proc, 4gb ram) with the > client > >> application (apache + php5 + ZF + apc) > and Tomcat + Solr enough ??? >> > Do I > need > >> more hardware ? >> > > Thanks in advance, > > Xavier S. > > Well my auto-complete is built on the facet prefix search > component. I think that 100-700 requests per seconds is maybe a better > approximation.