From: richard.harvey@nbs.nhs.uk
Date: Thu Aug 11 2005 - 07:15:42 GMT-3
What confuses me more is the application of these values in real-world deployments. What type of application or transport would benefit from gaining an extra Be bytes every so often. Would TCP keep enlarging and then shrinking the window size? or would it stick to the configured CIR for the policy (assuming no other traffic present).
If I had to choose I would just go for a zero Be every time and keep everything at the CIR. At least then you know what bandwidth you've got left to work with. In the exam though the requirement to burst would surely be quite clear - I would plump for Be = Bc in that case, unless a specific value is indicated.
Richard
-----Original Message-----
From: MIME :chrlewis@cisco.com Sent: 11 August 2005 10:37
To: kumara.shunmugam@wipro.com; marvin@ccbootcamp.com; andrew.m.edwards@boeing.com; ccielab@groupstudy.com
Subject: RE: FRTS Question [bcc][faked-from][bayes]
I have seen the formula you quote for Be before, but don't know where it
comes from, I do not believe a general formula like that for Be is
useful. I'd like to hear from anyone that uses this formula to configure
shaping and what results are obtained.
The router only bursts (Be) when there are tokens in the token bucket.
The token bucket does not accrue tokens unless the amount of traffic
being sent out is less than the CIR. The router can only burst for the
first Tc, after which the token bucket is empty.
This is illustrated graphically at the following link that may help
http://www.cisco.com/warp/public/125/traffic_shaping_6151.html Remember everything the router will send out will be at T1 rate, if that
is the access circuit rate, it is just that the router will not be able
to send out constantly at that rate with a CIR that is lower than T1
rate. When the access rate is higher than the CIR, the traffic comes out
in bursts at the interface rate and averages down to the CIR over the
course of a second.
To respond to specific statements in your post.....
The "peak rate" is always Bc+Be
Well kinda, that is true in terms of the number of bits the router is
able to send out for the first time interval, if and only if for the
previous period less than the CIR was sent and credit has accrued. Just
because you configure a given Be value does not mean it will ever be
sent, it depends entirely upon the traffic offered.
His circuit should be able to burst up to the interface space when
credits area available
Hmmmm, depends. All bits are clocked on to the serial circuit at T1
rate, a better way to "burst" up to access rate would be to set CIR to
the T1 rate and mincir to the contrated CIR rate.
Chris
-----Original Message-----
From: kumara.shunmugam@wipro.com [mailto:kumara.shunmugam@wipro.com] Sent: Thursday, August 11, 2005 3:44 AM
To: Chris Lewis (chrlewis); marvin@ccbootcamp.com;
andrew.m.edwards@boeing.com; ccielab@groupstudy.com
Subject: RE: FRTS Question [bcc][faked-from][bayes]
I think he require more clarify with his own example .. the "Peak" rate
is always Bc+be. As per his example,his AR is 1.544Mb and the CIR is
512K.His circuit should able to burst up to the interface speed when
credits are available..
Tc=125ms (default), which is 1/8 of 1 Sec (1 sec 1000ms) AR(Access Rate)
= 1544K
CIR=512K
Bc = CIR*TC/1000 = 64000 Bits per interval to achieve
512K.(64000*8)=512000
Be = AR-CIR*Tc/1000 =129000, which is 129000*8 = 1032000 per Sec
Peak Rate in one sec will be = Bc*8+be*8 = 512000+1032000 =1544000
=1544Kb
I think this should be the logic.. Correct me if am wrong..
Shun
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Chris Lewis (chrlewis)
Sent: Thursday, August 11, 2005 12:27 AM
To: marvin greenlee; Edwards, Andrew M; ccielab@groupstudy.com
Subject: RE: FRTS Question [bcc][faked-from][bayes]
That sounds about right.
Perhaps the idea of Bc=Be for shaping is getting crossed with that
practice for configuring CAR where that configuration menas no burst?
Andrew, all I can do is confirm my description of shape average and
shape peak and confirm that frame relay shaping operates like shape
average comes directly from the people coding IOS.
Chris
-----Original Message-----
From: marvin greenlee [mailto:marvin@ccbootcamp.com]
Sent: Wednesday, August 10, 2005 8:29 PM
To: 'Edwards, Andrew M'; Chris Lewis (chrlewis); ccielab@groupstudy.com
Subject: RE: FRTS Question [bcc][faked-from][bayes]
Not quite. With a line rate of 1.544 and a time interval of 1/8, you
only have 192k that can be sent during the time period. If you
configure 64k for Bc, then Be can use the other 128k, but not more.
Marvin Greenlee, CCIE#12237, CCSI# 30483 Network Learning Inc
marvin@ccbootcamp.com www.ccbootcamp.com (Cisco Training)
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Edwards, Andrew M
Sent: Wednesday, August 10, 2005 6:11 PM
To: Chris Lewis (chrlewis); ccielab@groupstudy.com
Subject: RE: FRTS Question [bcc][faked-from][bayes]
Importance: Low
Hmm..
I'll test the water cause I've had a similar question and discussion
with Marvin...
With the idea in mind that Be is only valid on the first time interval
for the second, if I have a 1.544mbps clocked circuit with a CIR of
512Kbps that I want to always utilize and my provider will allow me to
burst up to the full rate (1.544mbps), I should (generally speaking) set
FRTS as follows:
Frame CIR 512000
Frame mincir 512000
Frame Bc 64000
Frame Be = 512000
No frame adaptive-shape becn
This way I always get 512kbps CIR but can crank up to full line rate in
the first interval???
I thought shaping Be was bits per interval such that each interval I set
a maximum burst for that specific interval... Thus, Bc = Be to hit a
peak rate of 1.544Mbps?
andy
-----Original Message-----
From: Chris Lewis (chrlewis) [mailto:chrlewis@cisco.com]
Sent: Wednesday, August 10, 2005 3:55 PM
To: CCIE CCIE; ccielab@groupstudy.com
Subject: RE: FRTS Question
The simple answer is when that is the behavior you want :)
It is interesting though to look at what that actually does though.
Starting off looking at generic shaping configured with MQC.
Shape average has a burst setup that allows the Be amount of bits to be
sent in addition to the Bc amount of bits on the first interval of a
second.
Shape peak allows Be amount of bits to be sent along with Bc amount of
bits at every time interval.
Frame shaping behaves like the shape average, in that Be is only sent on
the first interval.
So really if you want to have an additional burst of say 1000 bits per
second, configure FRTS with a Be of 1000, as this will allow the shaper
to send an additional 1000 bits once per second on the first time
interval.
I have seen solutions to questions that have Be = Bc for FRTS when the
requirement is to have the burst rate equal to the committed rate, but
that I believe is incorrect. As Be is only added on the first time
interval and not on every time interval, that would not work.
Chris
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
CCIE CCIE
Sent: Wednesday, August 10, 2005 5:29 PM
To: ccielab@groupstudy.com
Subject: FRTS Question
If i configure FRTS when can i put Bc=Be?
Regards,
Patrick
---------------------------------
Find your next car at Yahoo! Canada Autos
This archive was generated by hypermail 2.1.4 : Sun Sep 04 2005 - 17:01:19 GMT-3