Measure. On an index with your real data with real queries <G>... Being a smart-aleck aside, using something like jMeter is useful. The basic idea is that you can use such a tool to fire queries at a Solr index, configuring it with some number of threads that all run in parallel, and keep upping the number of threads until the server falls over.
But it's critical that you use your real data. All of it (i.e. don't run with a partial set of data and expect the results to hold when you add the rest of the data). It's equally critical that you use real queries that reflect what the users actually send at your index. Of course, with a new app, getting "real" user queries isn't possible, and you're forced to guess. Which is much better than nothing, but you need to monitor what happens when real users do start using your system... Do be aware that what I have seen when doing this is that your QPS will plateau, but the response time for each query will increase at some threshold... FWIW Erick On Mon, Jul 18, 2011 at 10:35 AM, Siddhesh Shirode <sshir...@searchtechnologies.com> wrote: > Hi Everyone, > > I would like to know the best practices or best tools for Calculating QPS > in Solr. Thanks. > > Thanks, > SIDDHESH SHIRODE > Technical Consultant > > M +1 240 274 5183 > > SEARCH TECHNOLOGIES > THE EXPERT IN THE SEARCH SPACE > www.searchtechnologies.com<http://www.searchtechnologies.com> > >
