[19191] in North American Network Operators' Group
Re: FR Tidbit [was: 56k modems/frame/ptp revisited (LONG)]
daemon@ATHENA.MIT.EDU (Dan Jones)
Wed Sep 2 10:42:37 1998
Date: Wed, 02 Sep 1998 00:12:01 -0400
To: scott w <scott@digisle.net>
From: Dan Jones <danjones@bbn.com>
Cc: nanog@merit.edu
In-Reply-To: <Pine.GSO.3.96.980901083913.22515C-100000@guinness.digisle.
net>
Scott,
I'm curious. Is your assertion that "CIR is guaranteed only under
"normal" conditions" based on the Frame Relay spec or a paticular vendor(s)
implementation? I ask because I have been in that very debate for some
time (usually with a FR vendor justifying lousey performance).
My tendancy is to take the "Comitted" in CIR literally. It's my
understanding that the FR congestion management routines (at least the
Stratacom one - Foresight?) will only back off bursty PVC to the point
where it's <= CIR. The implication then is that you'll always get at least
CIR - which makes sense since that's what you're paying for.
Dan
At 08:45 AM 9/1/98 -1000, scott w wrote:
>
>
>
>On Tue, 1 Sep 1998, Scott, Dan wrote:
>
>[snip]
>> line and port, but the important parameter is the committed information
rate
>> you purchase from the carrier. Most of Sprint's frame relay business may
>[snip]
>>from 0-56kbps. Can purchase 56kbps CIR, but it costs extra, would need
>[snip]
>
>CIR is guaranteed only under "normal" conditions. You won't get it under
>heavy congestion. Find out the MIR configured for your PVC. I don't know
>if they will tell you though...
>
>scott
>
>
>
