Re: Multicast delays and high iowait

[Matt Garman]

On Tue, Apr 01, 2008 at 10:24:19PM +0200, H. Willstrand wrote:
> On Tue, Apr 1, 2008 at 6:05 PM, Matt Garman <matthew.garman@xxxxxxxxx> wrote:
> > We're using multicast basically for some inter-processs
> > communication.
> 
> Which protocol(-s) are in use? (UDP, IGMP, ...)

UDP.  FWIW, we're using the ACE_SOCK_Dgram and ACE_SOCK_Dgram_Mcast
objects from the ACE libraries (which, as far as I can tell, at
least in this situation, are just wrappers around system calls).

    http://www.cs.wustl.edu/~schmidt/ACE.html

> >  We timestamp (and log, in a separate thread) all of our sends
> >  and receives, and do analysis on the logs.
> 
> Are timestamps sent in the broadcast? If so, can the timestamps be
> out of sync generating the "delays"?

No, the timestamps aren't sent in the broadcast.  I just send a
sequence number.  So in my logging thread, I basically log something
like this: "sequence number n sent/received at time t".

I wrote a simple Perl script to compare the send and recv timestamps
logged for each sequence number.

Note that when I perform this test across two different machines,
the machines are sync'ed via NTP; we actually monitor clock
synchronization, and I can say with confidence that the clocks are
never more than a couple milliseconds off.  But also, I'm performing
these tests on the same machine (i.e. sender == receiver), so clock
sync issues don't apply.

Thanks again!
Matt

--
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html