Questions on CBWFQ

[sarahr202]

Hi everybody.

I hope all of you are having good weekend:)

Assume we have 100M CKT, and we are using default setting ( max-reserved 75 under f0/0 ) ;  we are using CBWFQ with three user defined classes.

int f0/0

service-policy LEE out

Policy map LEE

class HTTP

bandwidth percent 20

Class FTP

bandwidth percent 20

Class NETWORK_CONTROL_DSCP_48

bandwidth 35 percent

#############################

1) Let say all the above three classes are using 50M combined, that means 25 M is left out of max-reserve  bandwidth of 75M, will this 25M be shared among these user defined three classes only or will it also be shared with class default ?

2) Let say we have no traffic in class default (which is created by default and reserved 25 % of interface bandwidth in our case it is 25M), will this 25 M be shared with user defined classes ?

3)  Will network control traffic with DSCP 48 originated by this router be matched with " class Network_Control_DSCP_48" when exiting  f0/0?

CASE:2

int f0/0

service-policy LEE out

Policy map LEE

class HTTP

police rate 20

Class FTP

police rate 20

Class NETWORK_CONTROL_DSCP_48

police rate 35t

1)Let say there is no traffic in class HTTP, will the other two classes able to use this additional bandwidth or will they be limited to 20 M each because we are using police rate under them?

Much appreciated and have a nice weekend:)

[Furose M]

1) Let say all the above three classes are using 50M combined, that means 25 M is left out of max-reserve  bandwidth of 75M, will this 25M be shared among these user defined three classes only or will it also be shared with class default ?

class-default can use it, if not used by the 3 classes.

2) Let say we have no traffic in class default (which is created by default and reserved 25 % of interface bandwidth in our case it is 25M), will this 25 M be shared with user defined classes ?

yes.

3)  Will network control traffic with DSCP 48 originated by this router be matched with " class Network_Control_DSCP_48" when exiting  f0/0?

yes (99% sure).

===================

1)Let say there is no traffic in class HTTP, will the other two classes able to use this additional bandwidth or will they be limited to 20 M each because we are using police rate under them?

they will be limited to 20 and 35 respectively. they cant exceed the configured police level.

[sarahr202]

Thanks all for valuable info.

This is how I visualize egress queuing:

traffic coming from fabric-->Output queue--->User defined classes( queues) ----  scheduler--->TX    ring--interface

All user defined queues ( traffic class) exists between  OUTPUT queue and TX ring .

Traffic coming from fabric will be placed into output queue if TX ring is full or there is already traffic in output queue, , if it is not full and there is no traffic in output queue,, traffic is placed on TX ring.  All the traffic coming out from OUTPUT queue are then subject to policy map which places them into different classes ( queues) and scheduler dequeue  them and put them on TX ring.

Please correct me If I am wrong or miss something.

Thanks and have a nice weekend:)

[Carlos Villagran]

Hi!

1) This configuration really matters only when the link is congested. If it is not, any class will be able to take any available bandwidth needed. However in a congested state, they will be able to use only up to the configured bandwidth percent, this can be seen as a 'Reserved' bandwidth.

2) Yes. As stated above, everyone can access to all bandwidth as long as the interface is not congested. 

3) I do not know. Do you have a class-map called NETWORK_CONTROL_DSCP_48 matching dscp 48 marking?

-------------------------------------------

1) Nop. When using policers, the class traffic will NEVER be able to exceed the CIR configured in the policy-map (in bps or pps) no matter if there is available bandwidth.

If you have any other question or concern please let me know, also here is some documentation regarding QoS configuration guidance:

http://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst3750/software/release/12-2_55_se/configuration/guide/scg3750/swqos.html

Hope it helps, best regards!

JC

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

However in a congested state, they will be able to use only up to the configured bandwidth percent, this can be seen as a 'Reserved' bandwidth.

My understanding, and experience, is bandwidth statements actually provide the descheduling weight and do not set a maximum bandwidth allowance.

For example, I've used policies like:

policy example

class hi

 bandwidth percent 89

class lo

 bandwidth percent 1

class class-default

 bandwidth percent 10

and have seen the lo class take 70%, default take 25% and hi 5%.  This while class lo has queued packets.

[Joseph W. Doherty · Accepted solution]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

I don't believe max-reserved works the way you might believe.  Pre-HQF, by default, didn't allow you to allocate more than 75% of the bandwidth to the aggregate of all the non-default classes.  With pre-HQF, again by default, you would not be able to increase any of your non-default bandwidth settings in your first CBWFQ policy (without making other configuration changes).  However, interface and/or class or classes bandwidth usage is not limited to 75%.  I.e. any class or combination of classes can drive the interface to 100% utilization.

With your posted first CBWFQ policy, your four (remember there's always a class default) classes will be processed/dequeued in the ratio of 20:20:35:25, if they all want their maximum bandwidth allocations.  If one or more classes don't want their maximum bandwidth allocation, whatever they don't use is available to other classes.  Ratios are still preserved.  For example, if there was only HTTP and FTP class traffic, and both wanted 100% of the interface, they should split it 20:20 or 1:1.

3)  Will network control traffic with DSCP 48 originated by this router be matched with " class Network_Control_DSCP_48" when exiting  f0/0?

Maybe, maybe not.  Cannot tell since you didn't post the class maps.  If your question is traffic originated on the device treated like transit traffic in the egress CBWFQ policy, I believe the answer is yes.

For Case 2, as the other posters have already noted, policers will restrict bandwidth to what's defined by the policer.  BTW, this is also true of shapers.  However, on many platforms, it's not always true of a LLQ's implicit policer.