Memcached implementation inquiry

Don MacAskill don at smugmug.com
Sat Apr 21 16:57:55 UTC 2007


I know this whole thread is going a little off-topic, but I think it's 
relevant...

There's some debate as to how useful Google's malloc is, especially for 
general purpose workloads.  Also, someone seems to have found as imple 
patch for glibc malloc that solves or reduces the problem.

Here's some relevant posts on LKML:

http://lkml.org/lkml/2007/3/13/384

http://lkml.org/lkml/2007/3/19/471

Very interesting.

Don




marc at corky.net wrote:
> The exact story can be found here: http://ozlabs.org/~anton/linux/sysbench/
> 
> 
> 
> 
> Lexington Luthor wrote:
>> marc at corky.net wrote:
>>> Hi,
>>>
>>> mysql's scaling problem, especially on Linux, was recently found to 
>>> be a sub-optimal implementation of malloc when it comes to threads.
>>>
>>> If you preload google's malloc lib for mysqld you'll see a huge 
>>> difference past ~1000 TPS allowing it to scale almost linearly as well.
>>>
>>> Search for 'mysqld google malloc'
>>>
>>
>> At what kind of (OLTP) load does mysql begin to be hampered by lock 
>> contention? Can you link to any thorough tests done to find this out? 
>> Every test that I can find using google seems to be done by mysql 
>> themselves or storage engine vendors, which are no doubt biased.
>>
>> Redards,
>> LL
>>
>>
> 
> 


More information about the memcached mailing list