[PWOT] RE: draft-danenberg-sonet-ces-mpls-mib-00.txt

["Dave Danenberg" <[e-mail]>]

Dave/Jim,

We've considered your input on CEM statistics reporting and certainly
agree that they are good recommendations. We consider this an open
issue. But instead of immediately trying to solve it via the email list,
we're thinking of raising it as an issue at the PWOT session at the
upcoming IETF meeting. Then we'll gather and consider any more input,
before distributing a proposal for better CEM stats.

It looks like others are interested in TDM-like stats in the IP world.
Did you see the thread on MPLS email under the subject heading "RE:
Standards for IP stats collection? (corrected)" ?

-- Dave Danenberg

-----Original Message-----
From: [e-mail] [mailto:[e-mail]]On Behalf Of Dave
Danenberg
Sent: Wednesday, February 28, 2001 9:16 AM
To: David Zelig; Thomas D. Nadeau; Jim Boyle; Andrew G. Malis (E-mail)
Cc: [e-mail]; tom k. johnson
Subject: RE: draft-danenberg-sonet-ces-mpls-mib-00.txt


David Zelig, Jim, Tom, Andy,

Deriving SONET-like statistics from the CEM function is worth exploring
further. As David Zelig points out, the CEM operation at the ends of the
tunnel can monitor the path's B3 (BIP) byte. Although it may be tempting
to piggyback on existing functionality (B3), this technique would not be
able to easily discern between the errors introduced before the path
reached the CEM (Ac-A in the example) and the errors introduced within
the tunnel (A-Z). The CEM operation (at A) could regenerate B3, but that
would violate the integrity of the path's own error statistics. So I
think the way to look at this problem is by thinking of the CEM
operation as what it really is - an emulated SONET 'line' carrying the
path as its payload. As a 'line', the CEM could generate and terminate
its own BIP. This BIP, could then be used to create the same quality and
format of statistics as folks expect from SONET.

For CEM to support its own payload BIP, a new field would be required in
the CEM header. So this needs to carefully considered. If we were to do
this, we also need to decide if the CEM payload BIP should be calculated
on path or on packet delineations (i.e., calculated on the previous path
frame just as B3 does, or on the CEM header to end of packet).

(We'll forward this issue to the PWOT email list)

Dave Danenberg


-----Original Message-----
From: David Zelig [mailto:[e-mail]]
Sent: Wednesday, February 28, 2001 2:11 AM
To: 'Thomas D. Nadeau'; Dave Danenberg; Jim Boyle
Cc: [e-mail]
Subject: RE: draft-danenberg-sonet-ces-mpls-mib-00.txt


Tom, Dave and Jim,

In order to detect payload defects at the STS level there are BIP
checking
on the path overhead. If an operator want to check errors generated by
the
CEM operation it may be simpler (and more understood) to monitor the B3
byte
for errors in both ends of the tunnel. This can be done on the example
below
on the interface on the IP edge device.

It is very hard to manipulate CEM errors to payload errors, and if the
customer bought SONET service (and sign on SONET SLA) the PM should be
on
SONET "language" and measured at the SONET level.

Sure we must cover all potential CEM errors in order to enable detecting
the
cause of the errors, especially for the case of problems caused by
mis-configuration.I think that we need to deliver interval data for all
PM
related errors, and for events count.

David
-----Original Message-----
From: Thomas D. Nadeau [mailto:[e-mail]]
Sent: Wednesday, February 28, 2001 5:49 AM
To: Dave Danenberg; Jim Boyle
Cc: [e-mail]
Subject: RE: draft-danenberg-sonet-ces-mpls-mib-00.txt


At 08:02 PM 2/27/2001 -0500, Dave Danenberg wrote:
>Jim,
>
>Thanx for your input. It sounds like you'd like to see the same sort of
>statistical information from the emulated SONET that folks are
>accustomed to seeing from SONET networking equipment. I like that too
>(from years of developing telecomm equipment for service providers).
>
>Although I was considering second order counts (errored seconds) or
>trends (15 minute, 24 hour interval data), I decided to start with just
>simple counts - and see what folks think.
>
>Andy's CEM does not detect errors over the CEM payload, but it does
>detect errors within the CEM header (the CEM MIB counts them with
>mplsCemVcPerfCrctHdrErrors and mplsCemVcPerfUncrctHdrErrors). Perhaps
>CEM header error rates can be extrapolated to indicate payload errors
>rates. Due to the nature of CEM, other types of errors are detectable
>(buffer errors, missing packets, etc). Assuming CEM equipment can
>process this raw info into some nice statistics, another set of objects
>can be added to the CEM MIB.
>
>So, I'd like to include the type of statistical info you suggest as
well
>as your "reason for last down" idea. Would you like to see interval
data
>as well?
>
>What do you think Tom? If you agree, I can come up with another set of
>objects and send it to Jim (and mpls email) for comment.

         Yes, as I mentioned earlier, I think these things are a
good idea. Lets work on them and get our ideas back out to the
list shortly.

         --Tom




>Dave
>
>
>-----Original Message-----
>From: [e-mail] [mailto:[e-mail]]On Behalf Of Jim
Boyle
>Sent: Tuesday, February 27, 2001 2:24 PM
>To: Thomas D. Nadeau
>Cc: [e-mail]
>Subject: Re: draft-danenberg-sonet-ces-mpls-mib-00.txt
>
>
>
>pretend you are a service provider, a customer calls in and says he's
>seeing errors on his service (in particular, bip violalations in POH).
>
>He is directly connected to your network via port A.
>On the far end, he connects to a clec, which hands it off to you at
port
>Z.  He connects to a CLEC adm on port Z'.  His path terminating
>equipment
>(maybe a router) is port Ac, and Zc, such that the service looks like:
>
>Ac-A-[IP]-Z-CLEC-Z'-Zc
>
>Looking at lines Ac-A and Z-CLEC shows clean lines.
>
>How does the operator determine if his IP network is to blame, or the
>CLECs network is to blame?
>
>i'd recomend a few things:
>
>o) Andy's draft should have some sort of way to determine if there are
>bit
>errors hapenning across the IP network that affect the service
>payload.  This would provide the same sort of delineation of where a
>problem is occuring by looking at LOHs in more traditional transport
>network.
>
>o) your mib should include analogues to the SONET mib, such as bip-8
>violations, errored seconds, etc... (working off memory here).
>
>o) addition of a "reason for last down" might be nice.
>
>This way, when you, as the service provider, replies that "it's not our
>problem, must be the clec", they can say at least go look at it later
to
>see if they were telling the truth.
>
>As usual, excellent attention paid to the configuration aspects.
>
>regards,
>
>Jim
>
>
>On Fri, 23 Feb 2001, Thomas D. Nadeau wrote:
>
> >
> >       Hi,
> >
> >       FYI a new MIB for managing SONET/SDH Circuit Emulation
> > Service Over MPLS has been posted. It should be available
> > shortly via the IETF's web site. In the meantime, it is available
> > at the following web/ftp link. As usual, comments are welcome.
> >
> >
>ftp://[e-mail]/ftp/tnadeau/draft-danenberg-sonet-ces
-
>mpls-mib-00.txt
> >
> >       --Tom
> >

_______________________________________________
pwot mailing list
[e-mail]
http://www.ietf.org/mailman/listinfo/pwot