Several options:
- use tmpfs (loads everything in RAM)
- force the FS/OS to cache your index (e.g. cat /path/to/index/* > /dev/null)
- warm up your index well (e.g. *:* + sort queries if you sort)
- make use of Solr caches
Otis
--
Sematext -- http://sematext.com/ -- Lucene - Solr - Nutch
---
>From Erik Hatcher:
Try adding &debugQuery=true to your request and look at
the timings in the response. What is the QueryComponent time?
As I understand it, if you look carefully at the output you'll
see where time is spent, including the raw time the search
takes. That'll inform your next step
