Kewl...... Thanks!
Mike
"Hundley, Kent" wrote in message
[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> Michael,
>
> My comments are inline with **:
>
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of
> Michael L. Williams
> Sent: Monday, June 25, 2001 9:02 PM
> To: [EMAIL PROTECTED]
> Subject: Re: Frame Relay acceptable DE packets [7:9746]
>
>
> Kent..... I have some questions........ they're inline....
>
> "Kent Hundley" wrote in message
> [EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> > Symon,
> >
> > Unfortunately, its not as simple as looking at the DE packets. Simply
> > looking at DE packets alone doesn't tell you anything really. The
reason
> is
> > that if the FR cloud doesn't experience congestion, those DE packets
will
> > get there just as well as the non-DE packets. Some carriers encourage
> their
> > customers to use 0K CIR because "we don't oversubscribe our network", so
> > _all_ packets are marked DE. (Sprint used to do this, don't know if they
> > still do or not)
>
> This is good...... I often heard that people would suggest 0 CIR so that
> they could oversubscribe the #*&$ out of their network =)
> Then when people's traffic didn't go through they could go "well we agreed
> to carry 0"
>
>
> ** Can't speak to that, the only carrier I have seen that recommended 0
CIR
> is
> ** Sprint, and it seemed to work fairly well at the time, this is circa
> 1997.
> ** They always claimed that they didn't over-subscribe they're network, so
> no
> ** reason to pay for CIR. Course, they're pricing for 0 CIR was inline
with
> buying ** CIR from other carriers, so in the end from a cost point it was
a
> bit of a
> ** wash.
>
>
> > If you see the FECN's spike without a corresponding spike in the DE,
that
> > means your provider is experiencing congestion, but its on a backbone
link
> > and not your link. This means the provider's links are over-subscribed
> and
> > your packets are likely getting dropped without being marked DE.
>
> If this happens, and you sniff both sides and show that you're sending
more
> than you're receiving (i.e. you prove that you have packets being lost
> without being marked DE), isn't that a violation of your CIR agreement
> (assuming it's > 0)? Since nothing should get marked DE except for
packets
> over CIR, I can see how the logic makes sense, but does this happen often?
>
> ** Yes, and no. It depends on how the carriers SLA's are written. A lot
of
>
> ** SLA's are written such that you could lose quite a bit of traffic for
> very
> ** short periods of time and they would still be well within their SLA's.
> For
> ** example, if they guarantee %99.99 packet delivery and you transfer 1
> Terabyte
> ** of data per month, they could still drop 10 Mbytes of traffic, which
> could
> ** cause problems depending on how much was dropped in a given period.
> ** CIR is "guaranteed" only if the subscriber is not over-subscribed, but
> even
> ** then if the carrier has an outage its probable that they will be
> ** over-subscribed for the duration of the outage. When the bits fill the
> ** queues, a switch must drop traffic, DE or not.
> **
> ** IME, you usually only see consistent congestion on international links.
> ** Due to cost, it seems that every carrier is over-subscribed on the
> ** international links and some even mark every packet going across the
link
> ** as DE regardless of CIR. (yes, I've seen it happen) This sort of thing
> ** is obviously not in keeping with the spirit of the CIR agreement, but
> again
> ** one needs to read SLA contracts very carefully as there are usually a
lot
> ** of "weasel words" that benefit the carrier.
> **
> ** In practice, FR seems to continue to work well and be cost effective,
but
>
> ** it needs to be understood that sometimes packets will get dropped even
if
> ** you don't exceed CIR. If this happens consistently, its definitely an
> issue
> ** to take up with the carrier. With the exception of outages, its
> definitely
> ** violating the spirit of agreements for a provider to oversubscribe
their
> CIR.
> **
> ** Regards,
> ** Kent
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=10219&t=9746
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
