Hi Martin,
> Kelly,
>
> > So, I went ahead and spent a bit of time doing an experiment and
> > making sure I was not mixing up old items with 0mq.. As mentioned, one
> of
> > my tests was pushing about 200mb's a second which is unacceptable. I
> > switched off the no_delay options on the sockets and that cut the
> bandwidth
> > down to about 150mb/s, not a bad savings, but still not acceptable.
>
> Can you give a hint of what exactly are you measuring?
The test fixture has two parts. One runs the service on my test box
which is OsX and this simply tracks the bandwidth on the entire machine. I
measure this over the time period of 10 minutes using standard deviation. I
turn off everything which uses bandwidth while the test is running so the
error is probably a couple k second from samba, bonjour etc communications
running in the background.
The second test fixture runs on my main dev box which is Win7, 8
core 16 gigs. This fixture here has the "perfect" run log which is used to
validate the test run. At the same time, it runs the standard deviation on
when exactly the changes occur in the quad tree and when it receives the
changes from the test run. I.e. this is the latency measurement and should
be fairly accurate since the test simulation is fully deterministic.
The actual data involved is a sweep prune system running over the
data points in the quad tree. Each data point is moving, and the size of
its area of influence is changing. The data sent out for the interior data
is notification of when one point enters or leaves another points area of
influence. The edge nodes send out other data messages to their neighbor
but that currently is just eaten up by the test fixture and ignored.
Grand total, there are about 800,000 data points in the system.
Is that the information you wanted?
KB
_______________________________________________
zeromq-dev mailing list
zeromq-dev@lists.zeromq.org
http://lists.zeromq.org/mailman/listinfo/zeromq-dev
