MARK CALLAGHAN wrote:
> On Fri, Oct 10, 2008 at 5:23 PM, MARK CALLAGHAN <[EMAIL PROTECTED]> wrote:
>> On Fri, Oct 10, 2008 at 4:52 PM,  <[EMAIL PROTECTED]> wrote:
>>> I've just joined the list and have only poked about drizzle a bit, so please
>>> don't beat me up too much.
>>>
>>> On 10/10/08 14:51, Monty Taylor wrote:
>>>
>>> (snip...)
>>>> Coming soon:
>>>>
>>>>
>>> (snip...)
>>>>  Examine HOARD allocator
>>>>
>>> Is HOARD already used by Drizzle?
>>> One other worth investigating that has a lot of miles on it already
>>> (especially in terms of scaling to lots of cores/threads and handling cache
>>> alignment) is libumem.  It too is portable (ported from OpenSolaris by the
>>> OmniTI guys) and is also a dropin replacement for standard allocators.  It's
>>> newer than HOARD in that it has only been available in userspace since
>>> Solaris 9, but it's a port of the original multithreaded slab allocator
>>> which has been around since the mid '90s.  In addition to it's
>>> speed/concurrency, it has a bunch of things to simplify development.
>>>
>>> Two quick references:
>>> https://labs.omniti.com/trac/portableumem/
>>> http://blogs.sun.com/ahl/entry/solaris_10_top_11_20
>>>
>>> Anyway, if you're going to study baseline/HOARD, it could be worth looking
>>> at baseline/HOARD/libumem.
>>>
>>> - Matt
>>>
>>> --
>>> Matt Ingenthron http://blogs.sun.com/mingenthron/
>>> email: [EMAIL PROTECTED]
>>>
>>>
>>> _______________________________________________
>>> Mailing list: https://launchpad.net/~drizzle-discuss
>>> Post to     : [email protected]
>>> Unsubscribe : https://launchpad.net/~drizzle-discuss
>>> More help   : https://help.launchpad.net/ListHelp
>>>
>> I would be interested in the results if anyone runs perf tests for
>> Drizzle on a server with more than 4 cores and Hoard or libumem. Using
>> tcmalloc with MySQL makes a huge difference for the oltp readonly
>> sysbench test on that setup.
>>
>> http://code.google.com/p/google-mysql-tools/wiki/SmpPerformance
>>
>> --
>> Mark Callaghan
>> [EMAIL PROTECTED]
>>
> 
> Not that I am pushing tcmalloc. We are still trying to figure it out...
> http://code.google.com/p/google-perftools/issues/detail?id=78
>

We've got drizzle rigged with a  --enable-tcmalloc configure flag
already, btw...

Monty

_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to     : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help   : https://help.launchpad.net/ListHelp

Reply via email to