memcached replication

[Venkatesh KS]

I am working on something similar.

I am trying the following.

1. Convert memcached to inprocess cache. There is going to be a socket
through which it can still talk
but it can also do bypass calls down to memcached routine.

2. Memcached to notify if a cache entry expires or is ejected. This
notifiction for now is just a function pointer (this will help
me cleanup some state) but in could event notified over n/w to interested
parties.

Infuture I may go towards inprocess cache as L1 and out of process L2 cache.
The L2 can reside out of box as well. The L2 could be used by many L1 caches
as well. And L1 can just cache information about L2 entries. And if there is
a way for L2 (memcached) to notify events (add/eject/delete) to the
interested parties.


-venkatesh



On 8/29/07, Matt Ingenthron <Matt.Ingenthron at sun.com> wrote:
>
> Marcus Bointon wrote:
> > On 29 Aug 2007, at 00:37, Dustin Sallings wrote:
> >
> >> My goal is not replication, but to allow for a sort of L1 cache in an
> >> application with memcached as an L2 and cache invalidation service.
> >
> > That's a really nice idea. I've seen something vaguely similar with
> > jgroups, but it lacks the best bits of both memcache and in-process
> > caches (I'm also using APC with PHP). I can see that being a very
> > efficient system.
>
> That's also what ehCache does (in process cache, with remote L2 cache)
> for Java applications.
>
> I've looked at it a bit and talked with Greg Luck about it (the night he
> released his "benchmark" between ehCache and memcached).  The
> "benchmark" shows an impressive chart but leaves out the details you
> really need to understand what's going on-- looks like his blog filled
> in the details.
>
> Personally, I see room for both approaches.  From discussions with
> others, there are times you just want an app to minimize local memory
> usage.  Plus, in talking with Greg, he specifically plans in most cases
> to have a cache that overflows OS page buffer, which tells you it's
> typically deployed in a different way than memcached.  That doesn't
> negate the fact that sometimes a well managed, in-process cache would be
> an advantage.
>
> - Matt
>
> --
> Matt Ingenthron - Web Infrastructure Solutions Architect
> Sun Microsystems, Inc. - Global Systems Practice
> http://blogs.sun.com/mingenthron/
> email: matt.ingenthron at sun.com             Phone: 310-242-6439
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.danga.com/pipermail/memcached/attachments/20070829/d47d4027/attachment.html