[openstack-dev] blueprint AMQP RPC single per process response queue performance improvement
eric at cloudscaling.com
Mon Jan 21 19:32:38 UTC 2013
Quickly replying on list since I'm having trouble using gerrit on my
-1; The tests need to come out of common since these tests are used by the
zeromq tests, which will break. Otherwise, it looks fairly reasonable at a
glance, although I would like to evaluate it a bit further. As others
indicated, there might be ways to improve backwards compat.
As an aside, we're already using some similar patterns in the zeromq
driver. There might be further room for reuse here, but as I said, I'll
review the patch in more depth when I'm back at a keyboard.
On Jan 21, 2013 8:32 AM, "Ray Pekowski" <pekowski at gmail.com> wrote:
> On Thu, Jan 17, 2013 at 6:24 PM, Ray Pekowski <pekowski at gmail.com> wrote:
>> I got approval to share the results of my scalability study of RabbitMQ
>> as used by OpentStack done as part of my work at Dell. It shows the
>> improvement that comes from the single response queue change proposed by
>> the blueprint. Here is a link:
> For those not having enough time to go over the Dell presentation at the
> above link, the short summary is the amqp-rpc-fast-reply-queue blueprint
> provides a 23 times improvement to RPC call over a cluster of three
> RabbitMQ servers with mirroring enabled. The RPC call rate goes from 35
> RPC calls/sec to 812 RPC calls/sec.
> The blueprint can be found here:
> The code review here:
> For those interested, please review the code so that it can be approved.
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the OpenStack-dev