concerning server list management, connection pooling

[Jehiah Czebotar]

As for connection pooling; i wrote a wrapper around the python client
library which does connection pooling (i'm sure i've mentioned it on
the list before) http://jehiah.cz/download/MemcachePool.py.txt

but as brad mentioned... it's also not even needed in 99% of the
cases; memcached isn't slowed by a large number of connections.
--
Jehiah

On 1/18/07, Brad Fitzpatrick <brad at danga.com> wrote:
> The code I wrote for consistent hashing can be used even without its
> integration into the client library (if you use the consistent hashing to
> build your servers => [...] list passed to the client constructor...).
>
> But eventually it'll be integrated.  You might take a stab at that if you
> have time?
>
> As for connection pooling:  what exactly do you mean?  The server does
> epoll/kqueue... just keep a ton of connections open from every
> mod_perl/php process on all machines.
>
> - Brad
>
>
> On Fri, 12 Jan 2007, Ofer Nave wrote:
>
> > I've known about memcached for a while, but only recently starting
> > researching it for potential implementation within my company.  I love the
> > architecture, but there are two specific issues that concern me that I
> > haven't heard have been solved yet:
> >
> >
> >
> > 1)       modifying the server list (to add a new memcached server, for
> > instance) destroys the key -> server mapping. the equivalent of a
> > flush_all()
> >
> > 2)       connection pooling
> >
> >
> >
> > I realize both of these are not new issues to anyone on the list.  I've
> > tried skimming the archives to glean some insight from past discussions.  I
> > found a thread
> > (http://lists.danga.com/pipermail/memcached/2006-October/002944.html) that
> > referenced a research paper about consistent hashing.  I also found a post
> > by Brad
> > (http://lists.danga.com/pipermail/memcached/2004-October/000804.html) that
> > talks about the 'server list propagation' problem potentially being solved
> > by running a set of trackers and using virtual buckets.
> >
> >
> >
> > Unfortunately, I can't find anything in the memcached docs themselves (at
> > danga/memcached/, and in the docs of the PHP and CPAN modules) addressing
> > these issues, or covering the use of trackers and virtual buckets, so I
> > suspect that while clever individuals may have come up with various
> > solutions, none of them have made it into the code.
> >
> >
> >
> > So, my questions are as follows:
> >
> >
> >
> > 1)       What's the current consensus (or nearest to) on the list as to the
> > best strategies for these two issues?  (For the sake of context, assume an
> > apache environment with mod_perl or mod_php).
> >
> > 2)       Are there any working implementations of these solutions, either as
> > part of the client libraries, or as additional tools (one of the posts
> > mentioned creating a memcached admin tool)?
> >
> > 3)       If not, can I help?  I'm strongest with Perl, but can hack PHP if
> > need be, and I'm likely to need both clients at my work.  I'm also learning
> > Python, so I'm up for a good limited-scope project.. I can write good code,
> > but I have zero experience with memcached to date.  So while I'm happy to
> > help, I don't think I should be deciding what to implement, or how.
> >
> >
> >
> > -ofer
> >
> >
>