From: Petr Lapukhov (petrsoft@gmail.com)
Date: Sat Apr 29 2006 - 04:06:16 GMT-3
Actually, bandwidth configured at interface is used _directly_ to
calculate MQC limits, along with "max-reserved-bandwidth".
Some important things to remember with MQC (effective since 12.2):
1) By default, 25% of interface configured bandwidth is chomped
from the available interface bandwidth [if_bw]. That is,
"max-reserved-bandwidth" [res_bw_prc]=75%*
2) When you issue "bandwidth" [bw(i)] under a class [i], you set ABSOLUTE
value.
This value could NOT exceed [if_bw*res_bw_prc]. Available interface
bandwidth
is decremented accordingly, when you reserve bandwith for a class.
Also, SUM of all bandwidths across all classes, could NOT exceed
[if_bw*res_bw_prc]. That is: sum(bw(i))<=if_bw*res_bw_prc;
3) When you issue "bandwidth percent" [prc(i)] under class [i], you set
ABSOLUTE value again, and it is measured from the [if_bw]. That is, you
reserve
[prc(i)*if_bw] for class "i". Available interface bandwidth is decremented
accordingly to reservation.
Again, [prc(i)*if_bw] could NOT exceed [if_bw*res_bw_prc]. The same rule
holds for
sum of percents [sum(prc(i))*if_bw<=[if_bw*res_bw_prc].
That is, sum(prc(i))<=75% by default.
4) Now, for the "priority" and "priority percent". Those reserve bandwith
for priority queue, and configure policer. Priority chomps available
bandwidth
from interface. Both variants of priority command works with [if_bw] and are
ABSOLUTE.
5) Lastly, "bandwidth remaining percent". We already know that IOS
demands it's tribute for 25% of [if_bw]. Next, we remember
that there are priority classes under CBWFQ. They claim their
bandwidth too: [pri_bw] = either sum(pri_bw(i)) or [if_bw*sum(pri_prc(i))]
where [i] numbers priority classes.
So after we configure "priority" classes, we have remaining bandwidth
[total_rem_bw]=[if_bw*res_bw_prc]-[pri_bw].
Finally, "bandwith remaining percent" divides the [total_rem_bw] in RELATIVE
percent values. That is, [sum(rem_prc(i))<=100%], and remaining bandwidth
reserved for a class "i" is:
[rem_bw(i)] = [total_rem_bw]*[rem_prc(i)].
6) And last, but not least: you are forced to use consistent units for all
classes (e.g. you could not mix "bandwidth percent" and "bandwidth
remaining percent") for non-priority classes.
HTH
Petr
28.04.06, Alexei Monastyrnyi <alexeim@orcsoftware.com> NAPISAL(A):
>
> My understanding is it is relevant if you use "bandwidth percent" in
> your MQC configuration or something similar referring to relative
> bandwidth, being opposed to absolute values... But I might miss
> something here.
>
> I.e. if you need to configure bandwidth 512K in your MQC task, no
> technical need for "bandwith" on interface. Though in real life you'd
> better keep it in sync with real bandwidth just to be as consistent as
> possible with your OSPF or EIGHR domains etc
>
> A.
>
>
> on 28/04/2006 12:19 Shamin wrote:
> > Hi,
> >
> > First of all my thanks to all you guys for your great inputs.
> >
> > This is with reference to the IEWB LAB2 Ver3.0 Task 8.2
> >
> > It is mentioned in the solution guide as follows ..
> >
> > "Output queue calculations for the MQC are based on the configured
> > bandwidth value of the interface and not the speed the interface is
> > clocked at.Be sure to set the appropriate bandwidth value when
> > configuring the MQC on an interface"
> >
> > As per this statement , what I understand is, when configuring MQC on
> > any interface, it is necessary that we configure the bandwidth command
> > on the main interface, even if it is not mentioned in the lab
> > requirements.
> >
> > What confused me was that, for a similar task in LAB4 Ver3.0 Task 8.2,
> > there was no bandwidth configured under the main interface.
> >
> > So i am bit confused here, My question will be, is it mandatory to
> > configure the banwidth command on the main interface for MQC, even if
> > it is not mentioned in the LAB requirements?
> >
> > Regards
> > shamin
> >
> > _______________________________________________________________________
> > Subscription information may be found at:
> > http://www.groupstudy.com/list/CCIELab.html
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Mon May 01 2006 - 11:41:59 GMT-3