[ANN] beanstalkd 0.5

Keith Rarick kr at essembly.com
Tue Dec 11 22:25:30 UTC 2007


On Dec 11, 2007 2:17 PM, mike <mike503 at gmail.com> wrote:
> Does this -use- memcached, or is it basically a copy of memcached that
> has been modified to fit a job queue type model?
>
> Does it use memcached libraries, or is it totally independent?

Beanstalkd neither uses memcached, nor is it a copy of memcached. It
is an independent project that has taken inspiration from the design
of the memcache protocol. The implementation is completely new.

> I think it's a neat idea, however, for a work queue, I would think
> that it would be important to have guarantees (unless it guarantees it
> by processing it until it's done)
>
> memcached is volatile and can be empty or unavailable at any time, and
> the application should be able to function without it. Does this job
> queueing system function the same way? Since it is designed
> specifically with that purpose in mind it would not make sense to make
> a second job queue system (like the difference of using memcached as a
> CACHE vs. a DATA STORE) ...
>
> I hope this makes sense.

Yes, absolutely. Beanstalk intends to provide reasonable guarantees,
but this first release is in-memory only for the sake of simplicity.
We just wanted to get the code out in public!

Right now, if power goes out on a beanstalkd box, you will lose all
the jobs that were on that box. We're planning to add persistence to
beanstalk to make recovery possible in such a situation.

However, in practice, the queue is usually empty -- jobs run almost
immediately after they were submitted, so the chance of losing
important work is lower than you  might expect.

kr


More information about the memcached mailing list