[zeromq-dev] [Openpgm-dev] OpenPGM Roadmap

Steven McCoy steven.mccoy at miru.hk
Thu Aug 4 11:12:21 CEST 2011


On 4 August 2011 16:55, Martin Sustrik <sustrik at 250bpm.com> wrote:

> On 08/01/2011 06:28 AM, Steven McCoy wrote:
>
>     * Investigate single-thread-locked library for optimised 0MQ usage.
>>
>
> I guess this would help a lot with performance. 0MQ invokes OpenPGM -- and
> thus executes synchronisation) for every packet (every 1500 bytes) which
> results in a lot of locking in high-perf scenarios.
>
>
In 0MQ they will all be non-contended spinlocks so all you save is the
overhead of a few atomic locks per call. Any performance difference would
likely be untraceable on PGM alone, but it looks like it shouldn't be too
difficult to implement hence on the list.

The costs of memory management, checksumming and rate limiting are more
pertinent but also significantly more challenging to resolve.

-- 
Steve-o
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20110804/5f611704/attachment.htm>


More information about the zeromq-dev mailing list