Perl cache modules performance

32 views
Skip to first unread message

Roland Lammel

unread,
Jan 5, 2010, 7:06:03 PM1/5/10
to Perl-Cache Discuss
Hi all,

As I'm in the process of deciding upon the cache backend to use for a
web application I did a (not very thorough) benchmarking of the
different perl cache modules. So I thought I'd share that.

Btw. I decided to go with CHI using the Memcached::libmemcached driver
over IP with a Memory L1_cache which gives quite stable performance
values (I created the currently missing debian packages for testing,
just in case someone is interested, drop me a note)

As it's quite lengthy here is the link to this:
http://www.brabbel.net/wp/2010/01/perl-cache-modules-benchmark/

I'm still tuning it a little to get a better overview, but these were
the early results and might be of interest, as it's stated in the docs
of CHI anyway.

Comments welcome of course. When I'm done with it, I probably post it
on blogs.perl.org for a wider audience to see.

Cheers

+rl

Tim Bunce

unread,
Jan 6, 2010, 4:32:39 AM1/6/10
to perl-cach...@googlegroups.com
On Tue, Jan 05, 2010 at 04:06:03PM -0800, Roland Lammel wrote:
> Hi all,
>
> As I'm in the process of deciding upon the cache backend to use for a
> web application I did a (not very thorough) benchmarking of the
> different perl cache modules. So I thought I'd share that.
>
> Btw. I decided to go with CHI using the Memcached::libmemcached driver
> over IP with a Memory L1_cache which gives quite stable performance
> values (I created the currently missing debian packages for testing,
> just in case someone is interested, drop me a note)
>
> As it's quite lengthy here is the link to this:
> http://www.brabbel.net/wp/2010/01/perl-cache-modules-benchmark/

Typo:
C:Mc:lIP … Cache::Memcached::libmemcached over IP
C:Mc:l … Cache::Memcached::libmemcached over IP <= Socket?

It would be good to include your perl -V output as your perl coing could
influence the results in various ways.

Tim.

Roland Lammel

unread,
Jan 6, 2010, 1:18:42 PM1/6/10
to perl-cach...@googlegroups.com
On Wed, Jan 6, 2010 at 10:32, Tim Bunce <Tim....@pobox.com> wrote:
On Tue, Jan 05, 2010 at 04:06:03PM -0800, Roland Lammel wrote:
> Hi all,
>
> As I'm in the process of deciding upon the cache backend to use for a
> web application I did a (not very thorough) benchmarking of the
> different perl cache modules. So I thought I'd share that.
>
> Btw. I decided to go with CHI using the Memcached::libmemcached driver
> over IP with a Memory L1_cache which gives quite stable performance
> values (I created the currently missing debian packages for testing,
> just in case someone is interested, drop me a note)
>
> As it's quite lengthy here is the link to this:
> http://www.brabbel.net/wp/2010/01/perl-cache-modules-benchmark/

Typo:
   C:Mc:lIP … Cache::Memcached::libmemcached over IP
   C:Mc:l …   Cache::Memcached::libmemcached over IP <= Socket?

Thanks, corrected...
 
It would be good to include your perl -V output as your perl coing could
influence the results in various ways.

And added, of course not all people have Debian testing ;-)

There are substantially differences when doing the same benchmarking with compiled perl without ithreads I use for my AnyEvent stuff. About 5-10% increase in performance, but that's going to be a different post.

Cheers

+rl

Tim.

--
You received this message because you are subscribed to the Google Groups "Perl-Cache Discuss" group.
To post to this group, send email to perl-cach...@googlegroups.com.
To unsubscribe from this group, send email to perl-cache-disc...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/perl-cache-discuss?hl=en.

Perrin Harkins

unread,
Jan 21, 2010, 4:21:28 PM1/21/10
to perl-cach...@googlegroups.com
On Wed, Jan 6, 2010 at 1:18 PM, Roland Lammel <roland...@gmail.com> wrote:
> There are substantially differences when doing the same benchmarking with
> compiled perl without ithreads I use for my AnyEvent stuff. About 5-10%
> increase in performance, but that's going to be a different post.

That's true in general. I usually compile my own perl without threads
for this reason.

Thanks for sharing your benchmark.

- Perrin

Jonathan Swartz

unread,
Jan 21, 2010, 4:38:44 PM1/21/10
to perl-cach...@googlegroups.com
Rats. I always wondered if I should turn this off.

Any idea if an existing Perl can be recompiled without threads,
without rebuilding all its CPAN modules? Or does the threadedness
potentially affect how CPAN modules are built?

Jon

Perrin Harkins

unread,
Jan 22, 2010, 1:09:10 PM1/22/10
to perl-cach...@googlegroups.com
On Thu, Jan 21, 2010 at 4:38 PM, Jonathan Swartz <swa...@pobox.com> wrote:
> Any idea if an existing Perl can be recompiled without threads, without
> rebuilding all its CPAN modules? Or does the threadedness potentially affect
> how CPAN modules are built?

I'm not sure, but I would rebuild modules any time I rebuild perl.

- Perrin

Reply all
Reply to author
Forward
0 new messages