Confused on scaling memcached & persistant connections (w/ the PECL client)

Simon Forman forman.simon at gmail.com
Sat Mar 11 15:51:07 UTC 2006


I use that pooling class and it seems to work just fine.  We haven't
"gone live" with the system we're using it in yet, so I can't post any
real world usage stats or anything, but a few odd problems I was
having calling [python client] memcache from a multi-threaded server
went away using MemcachePool...

Peace, and thanks!
~Simon

On 3/10/06, Jehiah Czebotar <jehiah at gmail.com> wrote:
> I use the python memcache client, and i'm not sure if i mentioned it
> on the list before or not, but I found it pretty easy to write a
> pooling memcache class. While I agree with the comments about linux
> being able to handle the connections, a smart use of resources is
> always nice, and a pooling class addresses that in a much
> easier/cleaner way than a proxy would. For me it dropped my
> connections to a tenth of their previous number.
>
> http://jehiah.com/download/MemcachePool.py.txt
> --
> Jehiah
>
> On 3/10/06, Brad Fitzpatrick <brad at danga.com> wrote:
> >
> > "What's all this talk about reducing connections?  With epoll/kqueue you
> > can get a ridiculous number of connections.  How's a proxy daemon going
> > to help at all?  Then your proxy daemon is going to have a ridiculous
> > number of connections (which is no problem, if you use epoll/kqueue in
> > the proxy daemon), but now you have latency increases."
> >
> >
> > - Brad
> >
>


More information about the memcached mailing list