This still sounds like a better fit for ndb than memory to me. Too bad I can't head to head them. :)
Ronald Bradford <[email protected]> wrote: >I think from memory at eBay we used BTREE for at least one index. > >The use case was: > >PK insert/update of data, 20k record of cached data per user. >Range delete of stale data by last_updated_date (e.g. older then 20 mins) >Given the volume of data, like 400,000 rows per memory table per >server, the lock time on the range query was still like death to >performance but better then hash. > >Of course this is an extreme case, how many use memory engine for such >high volume of transactions. > >Regards > >Ronald > >_______________________________________________ >Mailing list: https://launchpad.net/~drizzle-discuss >Post to : [email protected] >Unsubscribe : https://launchpad.net/~drizzle-discuss >More help : https://help.launchpad.net/ListHelp >
_______________________________________________ Mailing list: https://launchpad.net/~drizzle-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~drizzle-discuss More help : https://help.launchpad.net/ListHelp
