I'd actually recommend against running multiple instances if possible.
Having larger instances makes multiget's more efficient, and causes
fewer roundtrips.
If it doesn't work to simply list a server multiple times, those clients
should be fixed.
-Dormando
Mark Maunder wrote:
Excellent, thanks for the quick reply Adam (and Dormando) and the
suggestion. cc'ing the list to get this in the archive for the next guy.
I've also added this to the faq:
http://www.socialtext.net/memcached/index.cgi?faq
Mark.
On Thu, Apr 3, 2008 at 3:26 PM, Adam Lee <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
It doesn't take it into account.
You could, however, run multiple instances of the same size on one
server. Not ideal, but works.
On Thu, Apr 3, 2008 at 6:24 PM, Mark Maunder <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
> Is it sensible to have different cache sizes across multiple
servers running
> memcached. I have one machine with 16GB of memory not doing much
and I could
> use 12GB for memcached. My other servers have 2GB available for
memcached.
> Will it include memory size in it's decision on where to store
stuff? Or
> does it simply load share evenly across all machines and if one
out of three
> servers runs out of memory then a third of new data won't be cached?
>
> This faq entry seems to suggest that the hashing algorithm
doesn't take into
> account how much cache is available on each server:
>
>
http://www.socialtext.net/memcached/index.cgi?faq#how_does_memcached_work
>
> Thanks,
>
> Mark.
>
>
--
awl
--
Mark Maunder <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>>
http://markmaunder.com/
+1-206-6978723
