Patch: Dynamic hashtable sizing
Jamie McCarthy
jamie at mccarthy.vg
Wed Aug 9 14:20:49 UTC 2006
Neat idea. Perhaps an optional command-line argument could be
added: the number of items expected to eventually exist. This
could tune performance by starting the hashtable at the estimated
best size. Actually that might be useful for some people even
without the hashtable-resizing code.
I'd have been surprised if a number of items of 2x the hashtable
size had a significant effect on performance. How about 3x, 4x,
10x, 50x? Did you run tests on those? What numbers did you get?
--
Jamie McCarthy
http://mccarthy.vg/
jamie at mccarthy.vg
More information about the memcached
mailing list