WAN QoS with multiple levels - assistance needed

[jacob6000]

Hello,

I have one hq site and three remotes sites. I need to configure  a QoS policy with multiple levels of priority. I need to define three levels of priority and  one default queue. Can you recommend what commands are applicable for each level? Any additional feedback is welcome.

The med-priority queue just needs some guaranteed bandwidth so maybe the bandwidth command is sufficient but what about the low priority queue? What would go in there?

Also, it appears that I can add the priority command for two levels for two classes at once. For example, I have the priority command on the High-Voip class and I can add it for the med-priority class. It seems like this would cause issues. Can both classes really be in the Priority bucket? Feedback?

policy-map qoslevels
class High-Priority-Voip
priority 2048
class Medium-Priority
bandwidth ??
class Low-Priority
command ?
class class-default
fair-queue

policy-map WAN-QoS
class class-default
shape average 42500000
service-policy qoslevels

Thank you,

Jay

[jacob6000]

Anyone? I could really uses some guidance. Thank you!

[Joseph W. Doherty]

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

Your QoS requirements are a bit vague.  What are your service level requirements?  Knowing those would determine how to configure your classes.

For a generic CBWFQ, I've found something like:

policy-map Sample

class real-time

priority percent 33

class HiPriority

bandwidth remaining 81 percent

fair-queue

class LoPriority

bandwidth remaining 1 percent

fair-queue

class class-default

bandwidth remaining 9 percent

fair-queue

deals well with many, many QoS prioritization needs.

All traffic uses default unless there's a reason to have traffic in another class.  Real-time is for traffic like VoIP bearer.  HiPriority is for traffic like video or VoIP control.  (NB: on average, HiPriority shouldn't need more than about a third of the bandwidth.)  LoPriority is for bulk data that can wait behind other traffic.

Regarding your question about two priority commands, on most platforms it just another class that shares the single LLQ.  This so different priority classes can have different implicit policers.  However, on ASRs, I believe they support two priority queues.  I'm not very familiar with the ASR implementation, but I think the intent is to allow a priority between priority classes (so you could have VoIP bearer in one, with its small packets, and perhaps real-time video in the other, with its larger packets, and keep the latter causing from jitter issues with the former.)

[jacob6000]

Thank you everyone and especially you Joseph, for assisting me once again.

1) Service levels?
Just that voice always go first to avoid choppy calls and second, that the med queue traffic is at least able to run regardless of any other traffic. The med traffic is clients hitting a database for call screen lookups, etc.

2) Why do the "bandwidth remaining" commands in the last three classes only equal up to 91%?  Why didn't you carve it up into chunks that equal 100%?

Thank you,

[Joseph W. Doherty]

re: #2

Because I'm actually working class bandwidth ratios.  In this case 81:9:1.

re: #1

FQ, alone, often insures light usage flows aren't (too) adversely impacted by bandwidth hogs.  However, FQ alone isn't good enough for something like VoIP.

Once you've got VoIP in LLQ and a default FQ, often you don't need more, but if you do, you can bump traffic up into the HiPriority class or bump it down into the LoPriority class if you need too.  Be careful, though, about putting traffic into HiPriority as any bandwidth hog there is likely to be adverse to your default traffic.

[Paul Bundschuh]

I just looked up multiple priority queues in my Szigeti book. They are allowed, but like Joe D mentioned, there is a single priority queue that they all share.

Here's a Cisco diagram that might help:

http://www.cisco.com/c/dam/en/us/td/i/200001-300000/290001-300000/290001-291000/290239.eps/_jcr_content/renditions/290239.jpg

[Paul Bundschuh]

I believe multiple priority queues can be configured, but one queue does not get priority over another queue, other than the policed value that you specify. So different queues can have different maximum policed rates, corresponding to high/med/low priority. My 15.1-running router will accept this policy map:

policy-map Foo

  class High

    priority 2048

  class Medium

    priority 1024

  class Low

    priority 512

But I didn't apply it to an interface, so am not sure if this works.  Might have to look this up.

[Joseph W. Doherty · Accepted solution]

Yes, it will work, but all 3 class go into one FIFO queue (as you note).  Each is subject to its own class policed rate (as you also note) but that's not a high/medium/low queuing priority.

You would obtain a somewhat similar result with:

policy-map foobar

class high

bandwidth percent 10

police 2048

class medium

bandwidth percent 10

police 1024

class low

bandwidth percent 10

police 512

The difference with this policy, there are 3 queues and we can adjust the priorities between the queues (independent of the policed rates).