how goes the new perl client?

[Brad Fitzpatrick]

Artur's apparently hacking on it during his 20-leg flight around the
world.  He's already added a binary get command to the memcached server,
so then writing a really fast C get call is easy.  That's in subversion
already.  The XS code for the Perl is in his private svk repo.... he'll be
branching it on the public svn soon.  I imagine when he gets back to
reliable net access.

- Brad


On Fri, 31 Mar 2006, Cahill, Earl wrote:

> I was just doing some memcached benchmarks on a couple newish boxes, and
> I was noticing that gets were kind of "slow," and I just started
> thinking about the perl client, and wondered how it was coming.  So, how
> is it coming?  I checked cpan and it wasn't there, but I do look forward
> to it being there.
>
>
>
> In case anyone is wondering, I think the bottom of the barrel sunfire
> x2100 would make for a fine memcached server.
>
>
>
> Using the attached perl file, I got the following results
>
>
>
> Benchmark: running incr for at least 2 CPU seconds...
>
>       incr:  3 wallclock secs ( 1.70 usr +  0.39 sys =  2.09 CPU) @
> 12100.00/s (n=25289)
>
> Benchmark: running decr for at least 2 CPU seconds...
>
>       decr:  2 wallclock secs ( 1.72 usr +  0.40 sys =  2.12 CPU) @
> 12098.11/s (n=25648)
>
> Benchmark: running get for at least 2 CPU seconds...
>
>        get:  2 wallclock secs ( 1.96 usr +  0.16 sys =  2.12 CPU) @
> 4033.96/s (n=8552)
>
> Benchmark: running set for at least 2 CPU seconds...
>
>        set:  3 wallclock secs ( 1.71 usr +  0.40 sys =  2.11 CPU) @
> 9726.07/s (n=20522)
>
> Benchmark: running set_delete for at least 2 CPU seconds...
>
> set_delete:  3 wallclock secs ( 1.72 usr +  0.40 sys =  2.12 CPU) @
> 5396.70/s (n=11441)
>
>
>
> Are there other memcached benchmark tools out there?  I think it would
> be great to start to compile a set of them.
>
>
>
> Thanks,
>
> Earl
>
>