On Jun 1, 3:41 pm, Shields <[email protected]> wrote: > Does anyone know of anyone using memcached in any of the following > scenarios: > - Low expiry (on the order of 5 seconds) > - High update rate for memcached objects > > Can anyone point me to performance data on mixed read/write > scenarios. I am explicitly exploring the performance of memcached > outside the normal 99% read, 1% write scenario. Any thoughts?
I use is it as an optimistic message deduplicator in twitterspy ( http://dustin.github.com/twitterspy/ ) It's not necessarily ``high'' volume since twitter redefined ``unlimited'' to ``20,000 requests per hour,'' but the response to those requests get broken into individual messages and keys are tested against every combination of everyone interested in a particular search query for every message in every response. This polling model also means I get a lot of duplicates, so it's useful to try to sending the same message twice where I can. The practice here is to just try to add a key comprised of the message ID and the destination JID. If the add is successful, I send it. If the add fails, I know for sure I've sent it before so I suppress it.
