I concur with this. As long as index segment files are cached in OS file cache performance is as about good as it gets. Pulling segment files into RAM inside JVM process may actually be slower, given Lucene's existing data structures and algorithms for reading segment file data. If you have very large index (much bigger than available RAM) then it will only be slow when accessing disk for uncached segment files. In that case you might consider sharding index across more than one server and using distributed searching (possibly SOLR cloud, etc.).
How large is your index in GB? You can also try making index files smaller by removing indexed/stored fields you dont need, compressing large stored fields, etc. Also maybe turn off storing norms, term frequencies, positions, vectors and stuff if you dont need them. On Feb 8, 2012, at 3:17 AM, Ted Dunning wrote: > This is true with Lucene as it stands. It would be much faster if there > were a specialized in-memory index such as is typically used with high > performance search engines. > > On Tue, Feb 7, 2012 at 9:50 PM, Lance Norskog <goks...@gmail.com> wrote: > >> Experience has shown that it is much faster to run Solr with a small >> amount of memory and let the rest of the ram be used by the operating >> system "disk cache". That is, the OS is very good at keeping the right >> disk blocks in memory, much better than Solr. >> >> How much RAM is in the server and how much RAM does the JVM get? How >> big are the documents, and how large is the term index for your >> searches? How many documents do you get with each search? And, do you >> use filter queries- these are very powerful at limiting searches. >> >> 2012/2/7 James <ljatreey...@163.com>: >>> Is there any practice to load index into RAM to accelerate solr >> performance? >>> The over all documents is about 100 million. The search time around >> 100ms. I am seeking some method to accelerate the respond time for solr. >>> Just check that there is some practice use SSD disk. And SSD is also >> cost much, just want to know is there some method like to load the index >> file in RAM and keep the RAM index and disk index synchronized. Then I can >> search on the RAM index. >> >> >> >> -- >> Lance Norskog >> goks...@gmail.com >>
