Caching: Memcached vs locmem

[Brian]

Can someone run down the differences between using Memcached vs
locmem?

The docs indicate Memcached is "the best" solution, but seems
considerably harder to setup. Just curious what the trade-offs are.

Thanks.

[the.re...@gmail.com]

I managed to get memcached up and running in about 5 minutes, so
whilst it does involve more setup than locmem I wouldn't say it's
difficult to the point of not being worth doing.

Memcached is basically more efficient in general and supports
clustering which makes it much better for scalability

[Brian]

On May 17, 7:37 pm, "a...@andrewingram.net" <the.real.a...@gmail.com>
wrote:

> I managed to get memcached up and running in about 5 minutes, so
> whilst it does involve more setup than locmem I wouldn't say it's
> difficult to the point of not being worth doing.

Thanks for that. I'll likely give it a try.

> Memcached is basically more efficient in general and supports
> clustering which makes it much better for scalability

But if you aren't clustering, say you have only a single server, is
there an advantage?

I turned on locmem today, but I don't think it is working. How can I
tell? I looked at my site with two browsers, one logged into admin,
and the other as anonymous. I made changes to the site as admin, and
the anonymous guy picked them up immediately.

Here are my settings:

CACHE_BACKEND = 'locmem:///?timeout=3600'
CACHE_MIDDLEWARE_SECONDS = 3600
CACHE_MIDDLEWARE_KEY_PREFIX = ''
CACHE_MIDDLEWARE_ANONYMOUS_ONLY = False

MIDDLEWARE_CLASSES = (
'django.middleware.cache.CacheMiddleware',
'django.middleware.common.CommonMiddleware',
'django.contrib.sessions.middleware.SessionMiddleware',
'django.contrib.auth.middleware.AuthenticationMiddleware',
'django.middleware.doc.XViewMiddleware',
)

[Brett Hoerner]

On Sat, May 17, 2008 at 8:12 PM, Brian <bgn...@gmail.com> wrote:
> But if you aren't clustering, say you have only a single server, is
> there an advantage?

Yes, locmem is memory local to a single Python process. If you're
running Django in some sort of multi-process server (as most people
do) e.g. Apache with the prefork MPM, each Python process will have to
cache for itself. If you use memcached then all of your Python
processes will speak to another process (memcached) and thus have the
same cache state at any given time.

> I turned on locmem today, but I don't think it is working. How can I
> tell? I looked at my site with two browsers, one logged into admin,
> and the other as anonymous. I made changes to the site as admin, and
> the anonymous guy picked them up immediately.

Is this using Django runserver, or under a 'production' setup such as
Apache? If you're using Apache this could be an example of the above,
your other user's request went to a different process which hadn't
cached the page yet in its own local memory.

Brett

[Brian]

On May 18, 12:55 pm, "Brett Hoerner" <bretthoer...@gmail.com> wrote:
> On Sat, May 17, 2008 at 8:12 PM, Brian <bgn...@gmail.com> wrote:
> > But if you aren't clustering, say you have only a single server, is
> > there an advantage?
>
> Yes, locmem is memory local to a single Python process. If you're
> running Django in some sort of multi-process server (as most people
> do) e.g. Apache with the prefork MPM, each Python process will have to
> cache for itself. If you use memcached then all of your Python
> processes will speak to another process (memcached) and thus have the
> same cache state at any given time.

Thanks, this makes sense. So each process has its own locmem cache.
But if you use memcached all processes would share memcached. Gotcha.

>
> > I turned on locmem today, but I don't think it is working. How can I
> > tell? I looked at my site with two browsers, one logged into admin,
> > and the other as anonymous. I made changes to the site as admin, and
> > the anonymous guy picked them up immediately.
>
> Is this using Django runserver, or under a 'production' setup such as
> Apache? If you're using Apache this could be an example of the above,
> your other user's request went to a different process which hadn't
> cached the page yet in its own local memory.

Yes this is under a production Apache server. Thanks.