Cell Relay Archive

["Peter E. Fry" <pfry@swbell.net>]

cburkhardt2@home.com wrote:

  Ah!  Now I see.  (I'd still like to know why you'd run VPCs through a
6400 to a RedBack...)

[...]
> Can some ATM switches leave a PVC "nailed-up" when a Customer Premise
> (ADSL modem) equipment is In-Active?  (Hence, good "Segment" F5 Loopback
> but failed "End-to-End" F5 Loopbacks).  This condition is seen approx
> 50% of the time we test.

  This is the only condition I've seen (given a connected path -- VCC in
the vast majority of cases), and I've seen a few while working at SBC,
DSL and otherwise.

> Can some ATM switches "dynamically" activate a PVC ONLY when the
> Customer Premise (ADSL modem) equipment is Active?  (Hence, failed
> "Segment" & "End-to-End" F5 Loopbacks).  This condition is seen approx
> 50% of the time we test.  I know this "dynamic" type of PVC setup sounds
> self defeating for F5 Loopback tests, but I have not found a positive
> answer yet.

  I can conceive the possibility, but have never heard of this type of
configuration.  I have seen lots of equipment that simply does not
support OAM (i.e. old), but I've never seen such deployed by a carrier. 
Note that last I heard the Alcatel DSLAM does not respond to segment F5s
from CPE.  Can't say about the carrier end.
  I don't know of a method to test particular (i.e. not adjacent)
segments using F5.  Also, with a VPC I would expect F5 to behave as
though all possible VCIs are configured/present -- that's been my
(limited) observation (one customer with an IGX switch).  I could run
some tests if you like by hijacking equipment that's not mine...
  Among the odd things I have seen:  Psychotic PVCs that won't pass
traffic and behave inconsistently when the VPI/VCI is changed, but
function just fine when terminated on a different circuit; Paths that
connect in one direction only; etc.  Then again, these are extremely
rare and should not account for any significant portion of what you're
seeing.  Any time I'd try to segment ping a PVC and fail I'd feel
justified in asking the broadband folks to fix the path -- but that's
with the single provider with which I have any experience: SBC.  YMMV.

Peter E. Fry