[zeromq-dev] Java NIO selector minimum possible latency

Michel Pelletier pelletier.michel at gmail.com
Sat Aug 25 19:24:36 CEST 2012


On Thu, Aug 23, 2012 at 1:12 PM, Julie Anderson
<julie.anderson.uk at gmail.com> wrote:
>
> Hi,
>

<Java NIO UDP benchmark results snipped>

> So my questions for the community are:
>
> 1 - Is my minimum time of 13 micros with average of 19 micros optimum for
> this round trip packet test. It looks like I am beating ZeroMQ by far so I
> may be missing something here. From this benchmark it looks like ZeroMQ has
> a 49 micros avg time (99% percentile) on a standard kernel =>
> http://www.zeromq.org/results:rt-tests-v031

It seems to me that you are comparing two very different things here.
The time to send a udp packet back and forth in a tight single
threaded selector loop is definitely going to be faster than a message
being sent back and forth using tcp as 0mq does, and then add on top
the overhead of 0mq managing the message across a thread boundary from
your application code to the core library.  The udp approach is going
to be faster, by far.  Of course it doesn't seem likely your NIO code
offers the same feature set as the 0mq library does from a messaging
framework point of view.

> 2 - Is there anything I can do to improve the selector reaction time when I
> spin a single or very few messages? 150 micros does not look good. Or should
> I assume that on a prod environment the selector will never be quite?

So is this a Java NIO question?  Or are you asking how our 49ms beats
your 150?  I think you meant "quiet" instead of "quite", but I'm still
confused as to the question.  Maybe post your benchmark code and we
can get some insight into what you're asking.

-Michel



More information about the zeromq-dev mailing list