In any case I browsed the code a bit more, it'd be easy to test out the MMap'ing because the FST simply reads directly from a byte[], which could be converted to a ByteBuffer (that's MMap'd). Nice!
On Thu, Jun 2, 2011 at 8:57 PM, Robert Muir <rcm...@gmail.com> wrote: > On Thu, Jun 2, 2011 at 11:49 PM, Jason Rutherglen > <jason.rutherg...@gmail.com> wrote: >>> because of how the datastructure is laid out. if you use mmap, instead >>> of lots of seeks, it will simply be lots of page faults instead. >> >> Ok, right, however assuming the pages are in RAM it could/should be >> ok. Eg, in HBase we'd want to store the keys alongside the column >> family values, on disk. Loading the keys into heap with the values >> still on disk/system IO cache probably is non-optimal. >> > > you don't have to believe me... if you want to use an FST, load it up in RAM. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > For additional commands, e-mail: dev-h...@lucene.apache.org > > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org
