I have to ask a different question: Why would you disable the caches? You're trying to test worst-case times perhaps?
Because the caches are an integral part of Solr performance. Disabling them artificially reduces your performance numbers. So disabling them is useful for answering the question "how bad can it get", but it's also skewing your results FWIW, Erick On Wed, Nov 13, 2013 at 7:42 AM, fbrisbart <fbrisb...@bestofmedia.com>wrote: > Solr uses the MMap Directory by default. > > What you see is surely a filesystem cache. > Once a file is accessed, it's memory mapped. > Restarting solr won't reset it. > > > On unix, you may reset this cache with > echo 3 > /proc/sys/vm/drop_caches > > > Franck Brisbart > > > Le mercredi 13 novembre 2013 à 11:58 +0000, Jacky.J.Wang > (mis.cnsh04.Newegg) 41361 a écrit : > > > > > > Dear lucene > > > > > > > > In order to test the solr search performance ,I closed all the cache > > solr > > > > > > > > insert into the 10 million data,and find the first search very > > slowly(700ms),and the secondary search very quick(20ms),I am > > sure no solr cache。 > > > > This problem bothering me for a month, > > > > > > > > Tracing the source code found > > > > > > > > 说明: 说明: cid:image001.png@01CED80C.EF49C740 > > > > Fisrt invoke readVIntBlock method always very slowly ,and secondary > > invoke readVIntBlock method is very quick, I don't know what reason is > > this > > > > > > > > Eagerly awaiting your reply, thanks very much!!! > > > > > > > > > > > > > > >