Let's see what the queries are. If you're searching for single terms that don't match many docs that's one thing. If you're looking at many terms that match many documents, I'd expect larger numbers.
Unless you're hitting the document cache and not searching at all.... Best Erick On Thu, Feb 10, 2011 at 5:08 PM, Stijn Vanhoorelbeke <stijn.vanhoorelb...@gmail.com> wrote: > Hi, > > I've done some stress testing onto my solr system ( running in the ec2 cloud > ). > From what I've noticed during the tests, the QTime drops to just 1 or 2 ms ( > on a index of ~2 million documents ). > > My first thought pointed me to the different Solr caches; so I've disabled > all of them. Yet QTime stays low. > Then the Lucence internal Field Cache came into sight. This cache is hidden > deep into Lucence and is not configurable trough Solr. > > To cope with this I thought I would lower the memory allocated to Solr - > that way a smaller cache is forced. > But yet QTime stays low. > > Can Solr be so fast to retrieve queries in just 1/2 ms - even if I only > allocate 100 Mb to Solr? >