From: Joe (groupstudy@comcast.net)
Date: Fri Jan 03 2003 - 00:26:54 GMT-3
This is a lengthy question, but I would really appreciate the time you
might take to help me with this.
When configuring FRTS, the value chosen for Bc is typically CIR/8,
assuming that an interval Tc of 125 ms is acceptable. Assume I have no
real-time traffic such as voice or video.
For the sake of example, suppose I have a circuit that is 4 timeslots,
or 256000 bps and the carrier CIR is 192000 bps. At the other end of
the link I have a T1 with a carrier CIR of 768000 bps and this end also
supports other PVCs. I am going to want to configure FRTS on the T1
side of the link for two reasons: 1) so it can throttle down to the
lower capacity link and 2) so I can reserve bandwidth for the other PVCs
that are being services.
The first question is this: When deciding on Bc, is the value for CIR
taken from the CIR you specify in the FRTS config or is it 1/8 the CIR
from the carrier?
The second question is this: What if any are the benefits of shaping on
the lower speed link as well? Would this use the same parameters for
shaping?
------------------------------------------------------------------------
----------------------------------------------
This is what I believe to be true about FRTS as it applies to this
example:
First of all, I would configure CIR = 256000 and MINCIR = 192000.
Anyone disagree with that?
If I configure Bc to be 192000/8 or 24000, then I will be sending out
24000 bits every time interval. This allows me to specify a Be of
64000, which would then be transmitted during the first time interval if
no congestion is detected. The data transmitted is:
Data transmitted = (Bc+Be) + Bc + Bc + Bc + Bc + Bc + Bc + Bc =
(24000+64000) + 24000 + 24000 + 24000 + 24000 + 24000 + 24000 + 24000 =
256000
As long as I don't experience congestion, this seems like a desirable
flow of traffic, even though some traffic can be marked DE.
On the other hand, if I configure Bc to be 256000/8 or 32000, then I
will be sending out 32000 bits every time interval. If I configure a Be
of anything other than 0, I will attempt to transmit this during the
first time interval. Suppose I set Be to 64000 again in this case. The
data transmitted is:
Data transmitted = (Bc+Be) + Bc + Bc + Bc + Bc + Bc + Bc + Bc =
(32000+64000) + 32000 + 32000 + 32000 + 32000 + 32000 + 32000 + 32000 =
320000
In this case I am attempting to transmit beyond the physical capacity of
the 256000 bps link, so 64000 bits of traffic will be FIFO queued
(unless I specify otherwise.) If I queue and carry over 64000 bits into
the next time interval then I can no longer burst, so I am down to a
rate of 256000 bits again. But this is STILL over the carrier CIR, so
this traffic is marked DE. This seems like it would create the
potential for packet drop/loss.
So which is the better configuration?
Thanks!
Joe
As a follow on to this:
Please take a look at the link below. I can't quite grasp why the
author omits Be on the remote side but includes it on the hub side.
I am also not in agreement with the math used to calculate the byte
increment at the bottom of the link.
Any comments would be appreciated.
Joe
<http://www.cisco.com/warp/public/125/traffic_shaping_6151.html>
http://www.cisco.com/warp/public/125/traffic_shaping_6151.html
.
This archive was generated by hypermail 2.1.4 : Sat Feb 01 2003 - 07:33:39 GMT-3