+1
If we use Tokyo Cabinet, we can still use b-tree index (specifically
it is b+tree) and still get the advantage of hash tables via it's hash
database API (which is the lower layer of it's b+tree database API).
Tokyo Cabinet is significantly faster than BDB and it will not bomb
out on large data. There are few very large shops that I know are
currently using it but I'm not sure if it's appropriate to name them
now.
I can say that mixi.jp is using it though.
Hmm guess the most convincing thing to do is actually try replacing
MEMORY with TC and take a sysbench of it... I'll see if I can get around
to doing that (unless someone else wants to do it of course).
Oh, and the author of Tokyo Cabinet happens to sit next to me so
I can poke him anytime if we ever come across any issues.
Cheers,
Toru
On Apr 28, 2009, at 5:26 AM, Baron Schwartz wrote:
Suggestion: consider replacing Memory with Tokyo Cabinet. Then you
get both properties of indexes in one index (hash + ordering for
ranges and scans).
I have a customer who tried out TC but I haven't done so myself. He
told me it was much faster than BerkeleyDB.
_______________________________________________
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
