Advance QoS implementation

Unanswered Question
Aug 21st, 2008

Dear all,

I am using QoS in my production network by applying IP Precedence. I have identified four classes. But when the link is congested in peak times, the lower two classes are completely dropped because the upper two eat the entire bandwidth.

Then i saw that i can fix bandwidth using the BANDWIDTH command in class-maps. But lets say if i allocate the bandwidth as 50%, 30%, 10%, 10% to all classes respectively. So, in congestion times the lower classes will still have some bandwidth to consume.

My question is, that when the link is not congested, will my top class still have only 50% of the bandwidth to use or it can use as much as it can get. I hope i have made my question clear.

Thanks in advance.


I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 3.8 (3 ratings)
Marwan ALshawi Thu, 08/21/2008 - 21:32

in case of conjestion the droping if based on ip precedence or dscp

the droping will start from the less to higher regardless of how much that traffic has in the bandwith command becasue i the case of droping means that the link got conjested and oversusecribed so it will start droping also u can control dropin threshoulds through advanced Qos confige on cisco catswithces but u need to be carefull and knowledgeable when do that

each cisco switch platform serience and also switches seriece module has diffrent port thrushould capabilities

for example the 350 has 2 ouput throushould

2970/3560/3750 they have 3T

while the WS-X6748-SFP in CAT6500 has 8T

and about the las question:

if the link not conjested each class will use bandwidth more than the amount u have reserved with bandwidth statement

this amount only incase of conjestion

thank you

please, if helpful rate

jaykishan Fri, 08/22/2008 - 00:52

Thanks you very much. If this is the case then i guess my task is done. I will just implement it in my production network and see if it actually happens.

Thanks again.


brandon.bennett... Fri, 08/22/2008 - 09:45

If you are doing bandwidth in policy-maps then I assume this is software QoS on a Cisco router of some type and not a switch.

First of all applying IP Prec (ToS) is just the first step of QoS and that is classification. You classify the packets into your classes and use PHBs (Per Hop Behavoir) on each router.

When you bandwidth statements under a policy-map you are using a queing mechinism that cisco calls CBWFQ. The actually algorthim of how the queues are servered is a Cisco guarded secret, but what it essentially does is garantee that much bandwidth at times of congestion and times of congestion only.

When the Tx Ring (oubound hardware queue) is not full packets flow normally and queuing is not needed.

So to answer you question: When the link is not congested you are not limited to that bandwidth and those flows will be able to use as much bandwidth as it can.




This Discussion