Why not let the fs handle this ??

[Constantin B]

I understand the point about cache replication but ,if we work with
1 or 2 huge memcache servers , if they fail , evrything is dead

i mean the db it self cannot provide the power needed without the cache so
it will
quickly die , untill the cache is restored , and the cache need the mysql to
refill .

How do you manage this in big setups ?
do you have a dozen of memcaches around ???

the fs has another greate advantage , it cache only used files in fs buffer
, so this mean that the ram usage is very small compared to memcache filled
with all data .

Ps : with reiser fs number of files is not a limitation
Ps2 : i think it can be cool to make memcache , cache on disk the unused
keys this and this will make it persistent on reboot too.


2006/6/6, Brian Moon <brianm at dealnews.com>:
>
> Constantin B wrote:
> > The purpose of this email is to ask what are the future limitations of
> > my setup ? I mean for the moment it do the
> > job , but where is the gotcha ?
>
> We are slowly migrating from a disk based cache to memcache.  We were
> already using NFS for our cache storage, so multiple copies was not a
> problem.  We saw issues with multiple apache processes trying to write
> the same file at the same time.  We would sometimes end up with a half
> written chunk in the file making the cache obsolete.
>
> Basically, offloading that work into memory and spread across lots of
> machines made our caches much more reliable.  Faster?  I dunno.  It
> depends on the exact parameters.  The FS can be very fast.
>
> Oh, another issue was what to do with a cache directory with 500,000
> files in it.  Even if its well structured, it can be a nightmare to
> manage.
>
> --
>
> Brian Moon
> -------------
> http://dealnews.com/
> Its good to be cheap =)
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.danga.com/pipermail/memcached/attachments/20060606/8cc7e98c/attachment.html