QOS bandwidth allocation

[mustansirt]

If I have the below policing configured on the input of the interface of the ISP port 

class realtime_dscp

  police cir 1952000 bc 242000 be 486000 conform-action set-mpls-exp-imposition-transmit 5 exceed-action drop

class time_critical_dscp

  police cir 976000 bc 120000 be 243000 conform-action set-mpls-exp-imposition-transmit 4 exceed-action set-mpls-exp-imposition-transmit 0

class besteffort_dscp

  police cir 936000 bc 116000 be 233000 conform-action transmit  exceed-action transmit

so if there is no real time traffic and only best effort traffic then the best effort traffic can use the available bandwidth of  real-time.

2nd question 

if the above is true, then as the best effort traffic is already using the available bandwidth then if real time traffic starts to come in then how will it be treated. Will it have to fight for the bandwidth with other class or real time will get the bandwidth which is specified in the police cir i.e. 1952000.

[Joseph W. Doherty]

Explicit policers always restrict traffic to the rate defined.

[mustansirt]

Thank you for the reply.but we have mentioned that  "conform-action set-mpls-exp-imposition-transmit 4 exceed-action set-mpls-exp-imposition-transmit 0" so the exceed action allows the classes to use the available bandwidth.

[Joseph W. Doherty · Accepted solution]

Touche, my bad, I didn't read fully read all your police statements and missed the exceed-action transmit on two of your policers.

That noted, ingress policy classes don't manage bandwidth, like egress policy classes.  Your 3 classes would obtain bandwidth share on a first come first serve basis except for the policer that would drop over CIR traffic.  Bandwidth is still first come, first served, unless dropped.

To specifically answer your original questions:

so if there is no real time traffic and only best effort traffic then the best effort traffic can use the available bandwidth of  real-time.

Yes, but that's because its policer isn't doing any effective traffic management, not is it marking based on observed data rate.  It does, though, provide you some stats on how traffic exceeds the CIR rate, or not.

if the above is true, then as the best effort traffic is already using the available bandwidth then if real time traffic starts to come in then how will it be treated. Will it have to fight for the bandwidth with other class or real time will get the bandwidth which is specified in the police cir i.e. 1952000.

It will fight for (ingress) bandwidth although realtime_dscp is limited to how much it can have.  Police statements don't directly guarantee obtaining bandwidth, but they can limit the amount of bandwidth that can be obtained.  Police statement can indirectly guarantee bandwidth by precluding or limiting what some traffic can obtain.  So, for example, if class A is limited to 50%, class B (or all other classes) is indirectly guaranteed to be able to obtain the other 50%.

PS:

Oh, and for my class A and B example, although class A is limited to 50%, it would still "fight" for that 50% against other classes.  Given just the two classes, both wanting 100% of the link, and class A policed to 50%, class B might still only obtain about 50%.

[mustansirt]

Thanks Joseph for explaining.

So going with your example of 2 classes A and B. We are limiting Class A by putting a policer of 50% of total bandwidth but Class B is not limited.

So if there is no congestion and no Class A traffic , Class B can utilise the available bandwidth.

And during this time if Class A traffic starts to come in ,it will need to fight for its confirmed 50% bandwidth with Class B.

And in another scenario if there is no Class B traffic , Class A can only use to 50% bandwidth as it is policed to 50%. even though there is 50% bandwidth available .

[Joseph W. Doherty]

Yup - exactly.

[mustansirt]

Dear Joseph ..thank you for conforming.

I have below policy-map 

policy-map XYZ

class real_time

  police rate percent 25

   exceed-action drop

  !

!

class business

  bandwidth percent 20

 !

class mgmt

  bandwidth percent 2

!

class routing

  bandwidth percent 2

!

class time_critical

  bandwidth percent 15

 !

class besteffort

  bandwidth percent 35

!

class scavenger

  bandwidth percent 1

!

class class-default

1) Is the above policy-map prioritizing the real-time traffic  at the time of congestion or no-congestion.

..I believe not as there is no priority command specified for real-time traffic class

if I apply priority and  don't police real-time traffic then the exceeding realtime traffic, if there is no congestion..will the exceeding traffic be sent as priority traffic or remarked as drop eligible traffic I mean will the exceeding traffic will still have more priority than other traffic or less

Thank you in advance for your clarification

[Joseph W. Doherty]

1) Is the above policy-map prioritizing the real-time traffic  at the time of congestion or no-congestion.

It's doesn't appear to have any explicit egress bandwidth management, and it's not configure as LLQ.

if I apply priority and  don't police real-time traffic then the exceeding realtime traffic, if there is no congestion..will the exceeding traffic be sent as priority traffic or remarked as drop eligible traffic I mean will the exceeding traffic will still have more priority than other traffic or less

Setting (LLQ) priority also sets an implicit policer, which drops traffic when it's overrate, but on many platforms, that's only when traffic is actually queued.

W/o a bandwidth/prioirty statement, I'm unsure how that class is treated.

[mustansirt]

Hi joseph..I also wanted to know that when the link is not congested can the priority traffic exceed it defined bandwidth and how is the exceeded traffic treated.

Same as priority traffic or as best effort i.e. dscp 0

[Joseph W. Doherty]

If there's no congestion, there's no congestion.  So, FIFO, but w/o queuing.

On most platforms, it appears LLQ can exceed its implicit policer.  I.e. it would not be dropped and treated normally (i.e. like within rate traffic).

If you have an explicit policer, it's subjected to how that policer is configured.

[mustansirt]

Joseph Can you suggest some books for QOS designing.

[Joseph W. Doherty]

None that I've read that I really like, when it comes to QoS design.

Almost any QoS design book should address how to treat something like VoIP using LLQ/PQ for its bearer traffic and insuring adequate bandwidth for its control traffic.  Beyond that, your mileage will really vary, and I rarely agree with what might even be considered the "best" QoS books.  (Before VoIP, how much QoS was there?  [I thought the subject sorely lacking then, and not much better dealt with today.])

That said, you can still learn much from QoS books, most QoS books from "name" publishers will be good.  Cisco's QoS design guide is worth a review too.  Just, again, unsure those will really help you with good QoS designs, beyond supporting VoIP.

[mustansirt]

Hi Joseph..There was a doubt that when we mark traffic as EF then what is the need to specify priority too in policy map as only EF will be given first preference over other DSCp marking i.e AF4X then AF3X then AF2X..AF1X and last best effort with no marking.

So Why specify priority , shouldn't EF be enough.

As explained that Priority packets are forwarded first at times of congestion but EF packets are also forwarded first at time of congestion right.

[Joseph W. Doherty]

Generally on routers, without explicit configuration, ToS tagged packets do not receive special/different treatment.  (One exception that comes to mind is interfaces running WFQ [not CBWFQ].)

Also, you don't need ToS tags to provide packets special/different treatment.

The purpose of a ToS tag, is to make it easier to identify traffic for special/different treatment.  Again, the tags themselves, don't guarantee anything.

[mustansirt]

Hi Joseph.wishing u a merry christmas.

Just wanted to know Like Cisco has an ideal QOS design for enterprise. do they suggest one for Service provider i.e how much should be reserved for real-time and for data and best effort.

Or we can follow the same standard as defined for enterprise.can you suggest any document if one Thanks.