Experimental Physics and
Industrial Control System

[email protected] (Jeff Hill) · Fri, 10 Dec 1999 10:58:53 -0700

> On Wed, 8 Dec 1999, Peregrine M. McGehee wrote:
>
> > Here at the Low Energy Demonstration Accelerator we are seeing
> > intermittent significant delays in the receipt of CA monitors. As these
> > delays are large enough to trigger a hardware protection system that
> > places the accelerator in a safe mode (e.g. turn the beam off) we have
> > spent considerable time over the past months to understand this.
> >
> > We don't yet understand the root cause and are asking if anyone has seen
> > a similar effect. We realize that TCP is non-realtime but are concerned
> > since these delays can as large as 3 to 10 seconds.
> >
>
> 	.
> 	.
> 	.
>
> I believe this is to be expected. Jeff Hill explained to me at one time
> that the channel access flow-control behavior can produce just these kind
> of results. Channel access favors getting the most recent information to
> an application over sending all intermediate information and this policy
> is a very good one. Heartbeat applications, on the other hand, are not
> well served by this behavior. It is easy to imagine that for a certain
> scenario, with the right amount of network loading, the client would miss
> the event when the value is zero and receive the event when the value is
> one (which would be the most recent info for this imagined case). This
> behavior wouldn't last forever, but on rare occasions it might follow
> this sequence long enough to trip the watchdog.
>

Actually, this (CA monitor flow control induced loss of intermediate
state changes, but not the final value) was not Peregrine's problem
He was seeing a situation where monitors were occasionally late arriving
by several seconds, but no intermediate monitors (as determined by
timestamps) were discarded.

Note however that some 90% of Peregrine's problems disappeared when our
ancient GTA era cabletron hub based Ethernet network was partly upgraded
to a switched Ethernet network, and I fear that the remaining 10%
has not been recharacterized, so we need to be careful about assuming
that our current problems behave the same as our other 90% did.

Jeff

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System