[c-nsp] BFD on ME3600/ME3800/7600s

[James Bensley]

Hi All,

On devices like the ASR9000s and ASR920s we have hardware acceleration
for BFD echo mode and this works as expected. On platforms that do not
have hardware support for BFD echo mode such as the
ME3600X/ME3800X/7600 I am confused as to how I am getting low BFD
interval times.

Enabling echo mode gives poor performance: the BFD echo packets are
originated by the local peer at the configured interval rate from
software, even though the remote peer loops them back in hardware,
upon return they are then checked in software, so all local actions
are happening in software. BFD control packets are sent at the rate of
"slow-timers" which is much slower and of no real concern here since
it is a fraction of the rate the echo packets are sent at. With echo
mode disabled the local (and remote) peer sends BFD control packets at
the rate it would have sent echo packets (the configured BFD interval)
and no echo packets are sent (so control packets are sent at the rate
of echo packets, instead of echo packets).

But we must have hardware accelerated support in order for these
control packets to be checked at the required rate to give fast
failure detection. From the Cisco docs:

http://www.cisco.com/c/en/us/td/docs/switches/metro/me3600x_3800x/software/release/15-3_3_S/configuration/guide/3800x3600xscg/swiprout.html#82479

"BFD is supported on physical interfaces that are configured as
routing interfaces."

I have been searching and only for the ME3600-24CX can I find this
snippet "The switch [ME3600-24CX] supports BFD only in Asynchronous
mode or no echo mode". I am guessing the statement above from the
ME3600X/ME3800X page "means" the same thing as the statement from the
ME3600X/ME3800x page, we do have hardware acceleration for BFD but it
is only in "no  echo mode". Surprise surprise, the documentation is
not clear so I want to make sure I have fully understood this.

Running BFD in echo mode between two ME switches the average interval
is very high and varies a lot:
Rx Count: 72, Rx Interval (ms) min/max/avg: 1/4976/4323 last: 2348 ms ago
Tx Count: 74, Tx Interval (ms) min/max/avg: 1/4968/4217 last: 1436 ms ago

Running BFD in no echo mode (which is what I usually do) we see the
average interval is a lovely 47ms (but I've never had 100%
clarification of why):
Rx Count: 3314443, Rx Interval (ms) min/max/avg: 1/72/47 last: 36 ms ago
Tx Count: 3310865, Tx Interval (ms) min/max/avg: 1/72/47 last: 40 ms ago


Cheers,
James.