The IP over ATM Mailing List Archive by date

Cell Relay Retreat>List Archive>month:1995-May> msg00160



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Resilient ATM

  • From: "Gerhold, Mark L [TR]" <MLG2@trpo6.Tredydev.Unisys.com>
  • Date: Thu, 18 May 95 10:31:00 EDT
  • CC: "Erickson, Richard A [RV]" <rae1@PO5.RV.unisys.com>, Gerhold at ecdcsvr <mlg1@ecdcsvr.tredydev.unisys.com>


Here's a diversion from multicast....

I have been trying to decide if a resilient ATM endsystem is
workable, or an easy upgrade to RFC 1577.  In TCP/IP, define
resilient ATM as preserving TCP dialogs even when there is an
underlying hardware failure.

Pitch. I see that the ARP server function has resiliency as a
work item.  What's good for the server is good for the endsystem.
ATM should try to compete with FDDI and allow fault tolerant
operation of the entire network, including the endsystem.

Here is a strawman for a redundant ATM endsystem.

Endsystem configuration:  One IP address.  Two ATM NICs, with two
ATM addresses - One active, one standby.

1. One NIC fails, or someone pulls the cable to the NIC.  Detect
the failure.  Close all VCs from the failed NIC.

2. Move the IP address to the backup NIC.

3. From the backup NIC, initiate the VC to the ARP Server, if not
already setup.

4. Transmit and respond to ARP_Requests and InARP_Requests, as in
RFC 1577.

5. The remote endsystems must also close the failed VCs,
invalidate the ATMARP entries, and periodically transmit
ARP_Requests to the server.

One obvious question with this simple scenario.  How do you
"detect the failure"?  There are a few partial failure
recognition methods.  First, signaling can detect the network
loss of SVCs.  Second, SONET can detect the loss of a link.
And third, there are F5 flows that send periodic wraparound cells
on a particular VC.  I'm not sure that any or all of these can
detect the loss of a connection under all failure scenarios, but
F5 flows offer the best hope.  It's not clear that F5 flows are
being universally implemented, and what happens if only one end
of a VC originates F5 cells.

So, to recap:

Has anyone else attempted to define a resilient ATM endsystem as
an addon to RFC 1577?

What mechanisms will be used to detect a ATM VC failure, and to
close the VCs?


Mark Gerhold
Unisys
mlg1@ecdcsvr.tredydev.unisys.com