Memcached uses way too much memory and then "dies"

[GRiNSER]

Hi gals and guys!

We run memcached on a single server website quite successfully. We pushed
the bar from about 200 oncurrent users w/o memcached to 400 with memcached.

Lately, memcached won't accept any requests at a certain level of memory
usage, which is much higher than configured at the time of the "crash". Then
the only solution is to restart the memcached instance or the server as a
whole, because mysql alone isn't able to face the load which causes the
server to swap and stall.

We allow memcached (v 2.1) to use 150 megs of our costly 1 gig ram, which
should be enough for our needs (as the stats of memcached itself show (usage
about 20% of the 150meg at max)), but it uses much more than it should
(which I can see in the linux process list).

Our problem sounds similar to problems other posters of the mailing list
faced recently.

Has anyone a clue about how to solve the crashing issue other than
restarting the server or memcached instance?
Will there be a solution for the memory "issues" in a upcoming version of
memcached?

Cheers ;-)

GRiNSER

PS: An option for the future will be to rent another server if we have
enough money...