Re: Variability in one-way latency and jitter

[Michael Meeks <michael ximian com>]

Hi Marc,

On Thu, 2004-07-22 at 15:54 -0400, Marc Siegel wrote:
> I've been doing some benchmarking of open-source CORBA ORBs at work, and
> have come across the odd result that even in the local case,
> asynchronous (one-way) invocations of calls often take much longer than
> the synchronous counterpart, with a huge variability in latency and
> jitter results across test runs.

	I imagine this is because locale or remote oneway calls are really
rather asynchronus ;-) as soon a you send a number of them, so as to
fill the socket buffer; we form a queue that is subsequently emptied
when more space in the socket buffer becomes available - I guess that
discontinuity may cause the jitter; do you notice a step effect in
throughput / latency as you increase emission frequency ?

	Regards,

		Michael.

-- 
 michael ximian com  <><, Pseudo Engineer, itinerant idiot