mod_cache caching cluster

mod_cache caching cluster

am 07.12.2009 20:08:28 von mtice

--0016e64b98e8bfa9f5047a282ea7
Content-Type: text/plain; charset=ISO-8859-1

Hello, I was wondering if anyone has run across a means consolidating or
clustering their cache? Currently we have 20 nodes that only serve up
static content. Each node is configured with a 6G ramdisk
(mod_disk_cache). This works *ok* except for a couple issues. 1) We
experience intermittent performance issues (it seems to happen when
htcacheclean kicks off), and 2) the cache varies from machine to machine.

I was digging around with mod_memcache - I really like the idea but 1) it
doesn't look like it's actively developed, and 2) I can't seem to get the
caching to do what I want. I was also briefly looking at JCS - but that
maybe a little overkill?

I could have the content stored on a shared NFS mount but I wanted to stay
away from disk-based caching if I could.

If anyone has any suggestions or ideas I'd appreciate it greatly.

Matt

--0016e64b98e8bfa9f5047a282ea7
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

Hello, I was wondering if anyone has run across a means consolidating or cl=
ustering their cache?=A0 Currently we have 20 nodes that only serve up stat=
ic content.=A0 Each node is configured with a 6G ramdisk (mod_disk_cache).=
=A0 This works *ok* except for a couple issues.=A0 1) We experience intermi=
ttent performance issues (it seems to happen when htcacheclean kicks off), =
and 2) the cache varies from machine to machine.


I was digging around with mod_memcache - I really like the idea but 1) =
it doesn't look like it's actively developed, and 2) I can't se=
em to get the caching to do what I want.=A0 I was also briefly looking at J=
CS - but that maybe a little overkill?


I could have the content stored on a shared NFS mount but I wanted to s=
tay away from disk-based caching if I could.

If anyone has any sugge=
stions or ideas I'd appreciate it greatly.

Matt


--0016e64b98e8bfa9f5047a282ea7--

Re: mod_cache caching cluster

am 07.12.2009 20:23:33 von Dan Poirier

Matthew Tice writes:

> I was digging around with mod_memcache - I really like the idea but 1) it
> doesn't look like it's actively developed, and 2) I can't seem to get the
> caching to do what I want.

Yes, mod_mem_cache development is pretty much stopped, and it's been
dropped from the next release of Apache. Also consider that mem_cache
would have an independent cache per process, which would be the reverse
of the consolidation you're looking for.


------------------------------------------------------------ ---------
The official User-To-User support forum of the Apache HTTP Server Project.
See for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
" from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org

Re: mod_cache caching cluster

am 08.12.2009 00:04:17 von Francis Galiegue

On Mon, Dec 7, 2009 at 20:08, Matthew Tice wrote:
> Hello, I was wondering if anyone has run across a means consolidating or
> clustering their cache?  Currently we have 20 nodes that only serve =
up
> static content.  Each node is configured with a 6G ramdisk
> (mod_disk_cache).  This works *ok* except for a couple issues. =
1) We
> experience intermittent performance issues (it seems to happen when
> htcacheclean kicks off), and 2) the cache varies from machine to machine.
>
> I was digging around with mod_memcache - I really like the idea but 1) it
> doesn't look like it's actively developed, and 2) I can't seem to get the
> caching to do what I want.  I was also briefly looking at JCS - but =
that
> maybe a little overkill?
>
> I could have the content stored on a shared NFS mount but I wanted to sta=
y
> away from disk-based caching if I could.
>
> If anyone has any suggestions or ideas I'd appreciate it greatly.
>

Since it is only really static content, and provided that when a given
element changes, its URI changes, you should definitely look at
mod_expires:

ExpiresActive On


ExpiresDefault "access plus 1 month" # or more
Header append Cache-Control "private" # this tells that the proxy
won't cache, but the final client will


You don't even need disk-based caching. The OS' pagecache will largely
fill the "need for speed".

--=20

Francis Galiegue
ONE2TEAM
Ingénieur système
Mob : +33 (0) 683 877 875
Tel : +33 (0) 178 945 552
fge@one2team.com
40 avenue Raymond Poincaré
75116 Paris

------------------------------------------------------------ ---------
The official User-To-User support forum of the Apache HTTP Server Project.
See for more info.
To unsubscribe, e-mail: users-unsubscribe@httpd.apache.org
" from the digest: users-digest-unsubscribe@httpd.apache.org
For additional commands, e-mail: users-help@httpd.apache.org

Re: Re: mod_cache caching cluster

am 08.12.2009 14:50:16 von mtice

--00504502f6239c3118047a37da6f
Content-Type: text/plain; charset=ISO-8859-1

On Mon, Dec 7, 2009 at 12:23 PM, Dan Poirier wrote:

> Matthew Tice writes:
>
> > I was digging around with mod_memcache - I really like the idea but 1) it
> > doesn't look like it's actively developed, and 2) I can't seem to get the
> > caching to do what I want.
>
> Yes, mod_mem_cache development is pretty much stopped, and it's been
> dropped from the next release of Apache. Also consider that mem_cache
> would have an independent cache per process, which would be the reverse
> of the consolidation you're looking for.
>
>
Well, I was talking about mod_memcache (
http://code.google.com/p/modmemcache/) which seems to sort of be an
Apache-frontend for memcached servers. To get around using mod_mem_cache I
use mod_disk_cache on a ramdisk. It does great except for some reason I
can't get it to scale over 4G very well.

--00504502f6239c3118047a37da6f
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable



On Mon, Dec 7, 2009 at 12:23 PM, Dan Poi=
rier <poirier@pob=
ox.com
>
wrote:
border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; paddi=
ng-left: 1ex;">
Matthew Tice <mjti=
ce@gmail.com
> writes:



> I was digging around with mod_memcache - I really like the idea but 1)=
it

> doesn't look like it's actively developed, and 2) I can't =
seem to get the

> caching to do what I want.



Yes, mod_mem_cache development is pretty much stopped, and it's b=
een

dropped from the next release of Apache. =A0Also consider that mem_cache >
would have an independent cache per process, which would be the reverse

of the consolidation you're looking for.



We=
ll, I was talking about mod_memcache ( odmemcache/">http://code.google.com/p/modmemcache/) which seems to sort=
of be an Apache-frontend for memcached servers.=A0 To get around using mod=
_mem_cache I use mod_disk_cache on a ramdisk.=A0 It does great except for s=
ome reason I can't get it to scale over 4G very well.


--00504502f6239c3118047a37da6f--

Re: mod_cache caching cluster

am 08.12.2009 15:02:29 von mtice

--0016e64b98e84fc146047a38060e
Content-Type: text/plain; charset=ISO-8859-1

On Mon, Dec 7, 2009 at 4:04 PM, Francis GALIEGUE wrote:

> On Mon, Dec 7, 2009 at 20:08, Matthew Tice wrote:
> > Hello, I was wondering if anyone has run across a means consolidating or
> > clustering their cache? Currently we have 20 nodes that only serve up
> > static content. Each node is configured with a 6G ramdisk
> > (mod_disk_cache). This works *ok* except for a couple issues. 1) We
> > experience intermittent performance issues (it seems to happen when
> > htcacheclean kicks off), and 2) the cache varies from machine to machine.
> >
> > I was digging around with mod_memcache - I really like the idea but 1) it
> > doesn't look like it's actively developed, and 2) I can't seem to get the
> > caching to do what I want. I was also briefly looking at JCS - but that
> > maybe a little overkill?
> >
> > I could have the content stored on a shared NFS mount but I wanted to
> stay
> > away from disk-based caching if I could.
> >
> > If anyone has any suggestions or ideas I'd appreciate it greatly.
> >
>
> Since it is only really static content, and provided that when a given
> element changes, its URI changes, you should definitely look at
> mod_expires:
>
> ExpiresActive On
>
>
> ExpiresDefault "access plus 1 month" # or more
> Header append Cache-Control "private" # this tells that the proxy
> won't cache, but the final client will
>

>
> You don't even need disk-based caching. The OS' pagecache will largely
> fill the "need for speed".
>
> --
>

Thanks Francis, I'm a little confused about a couple things. 1) Is the
ExpiresDefault in mod_expires similar to CacheDefaultExpire in mod_cache?
2) This wouldn't address the need for a backend global cache? Also I
wanted to limit/reduce any kind of paging and it seems that with 9+G of
cache I would be swapping all over the place.

--0016e64b98e84fc146047a38060e
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable



On Mon, Dec 7, 2009 at 4:04 PM, Francis =
GALIEGUE <fge@one2=
team.com
>
wrote:
=3D"border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; p=
adding-left: 1ex;">
On Mon, Dec 7, 2009 at 20:08, Matthew Tic=
e <> wrote: r>
> Hello, I was wondering if anyone has run across a means consolidating =
or

> clustering their cache?=A0 Currently we have 20 nodes that only serve =
up

> static content.=A0 Each node is configured with a 6G ramdisk

> (mod_disk_cache).=A0 This works *ok* except for a couple issues.=A0 1)=
We

> experience intermittent performance issues (it seems to happen when >
> htcacheclean kicks off), and 2) the cache varies from machine to machi=
ne.

>

> I was digging around with mod_memcache - I really like the idea but 1)=
it

> doesn't look like it's actively developed, and 2) I can't =
seem to get the

> caching to do what I want.=A0 I was also briefly looking at JCS - but =
that

> maybe a little overkill?

>

> I could have the content stored on a shared NFS mount but I wanted to =
stay

> away from disk-based caching if I could.

>

> If anyone has any suggestions or ideas I'd appreciate it greatly.<=
br>
>



Since it is only really static content, and provided that when =
a given

element changes, its URI changes, you should definitely look at

mod_expires:



ExpiresActive On



<Location /some/static/URI/base>

=A0 =A0ExpiresDefault "access plus 1 month" # or more

=A0 =A0Header append Cache-Control "private" # this tells that t=
he proxy

won't cache, but the final client will

</Location>



You don't even need disk-based caching. The OS' pagecache will larg=
ely

fill the "need for speed".



--

Thanks Francis, I'm a little confuse=
d about a couple things.=A0 1) Is the ExpiresDefault in mod_expires similar=
to CacheDefaultExpire in mod_cache?=A0 2) This wouldn't address the ne=
ed for a backend global cache?   Also I wanted to limit/reduce any kind=
of paging and it seems that with 9+G of cache I would be swapping all over=
the place.


--0016e64b98e84fc146047a38060e--