Cell Relay Archive

[manfredi@arl.bna.boeing.com]

In article <5r6ppv$rpn$1@dismay.ucs.indiana.edu>,
  Patrick Hurley <patrick.hurley@wco.com> wrote:
>
> marc langston wrote:
> >
> > I have recently been in a discussion concerning how ATM would handle the
> > case of a PVC/VCC reconnect in the event of an intermediate switch
> > outage.  For example, if the PVC travelled from access switch A to
> > intermediate switch B to intermediate switch C to access switch D, and
> > switch B were to fail, what would happen to the PVC?  Would it "regrow,"
> > or is that PVC just plain down?
> >
> > I have always been under the impression that ATM "the protocol" does not
> > have any way to "rebuild" the PVC in the event of a switch outage.  (If
> > there was a trunk outage, SONET would re-route to the switch.)  But my
> > recent discussions have brought up points to the tune of "How can you
> > have a modern transport protocol that isn't able to recover from that
> > sort of outage?"
> >
> > I haven't been able to find any information in ATM books about what
> > happens in outage situations, so thought of asking the question here.
> > Any thoughts?
> >
> > Thanks,
> > marc
>
> Soft PVCs will reroute with an intermediate switch failure.  Cisco
> Stratacom switches have this functionality.   Also, PNNI will support
> the same type of rerouting with crankback.

I'm sure that different, more or less proprietary approaches can be
developed to monitor ATM VCs periodically, then re-establish them if a
link breaks. But the more fundamental reply to Marc's point has to be
that ATM was not intended to be used like SONET. Permament or very
long-lived ATM pipes are an anomaly.

ATM, invented by the telco community (Bellcore), works on the same model
as telephony. If a link breaks while an unfortunate VC was using it, and
the physical layer cannot repair it transparently, then ATM would assume
that the VC would be re-established by the end users. The PNNI would find
a working path from A to B, assuming one exists, but for this new route
to take effect the end users have to go through the signaling procedure.
Just as you would for a telephone call that suddenly quits in
mid-sentence.

If VCs are short-lived, this model works just fine. In this case, the
shorter- lived the better. Ultimately, a VC could conceptually be
established for only one data frame at a time (frame, not cell), at which
point ATM's circuit switched nature would collapse into something similar
to packet switching.

Bert
manfredi@arl.bna.boeing.com

-------------------==== Posted via Deja News ====-----------------------
      http://www.dejanews.com/     Search, Read, Post to Usenet