Priorization QOS

[Tiago Eduardo Zacarias]

Hello everybody,

I need a light with a QOS policies. I am implementing QOS with voice prioritization on a network, but I'm finding it hard to implement when dealing packets of EF PHB class, to be priozrizado it does not and prioritized and congestion in the queue for packets of PHB AF31 class. Below an example of a configuration did a simulation to test scoring telnet packets as EF and priorizoando the same but beginning the transfer of a aquivo via ssh realize the slowness in the telnet connection does not have priority in the output. How do I in a moment of congestion on the interface the algorithm prioritize pocotes marked EF?

class-map match-any TELNET
 description CLASS-TELNET - PRIORITARIA - EF
 match ip dscp ef 
class-map match-any SSH-out
 description CLASS-SSH - CRITICAL - DATA 
 match ip dscp af31 
class-map match-any HTTP
 description CLASS-HTTP - CRITICAL DATA
 match ip dscp af33 
class-map match-any TELNET-out
 description CLASS-TELNET - PRIORITARIA - EF
 match ip dscp ef 
class-map match-any SSH
 description CLASS-SSH - CRITICAL - DATA 
 match ip dscp af31 
class-map match-any HTTP-out
 description CLASS-HTTP - CRITICAL DATA
 match ip dscp af33 
!                
policy-map POL-2MB-out
 class TELNET-out
  set ip dscp ef
  priority 128
 class HTTP-out
  set ip dscp af33
  priority 200
 class SSH-out
  set ip dscp af31
  priority 500
 class class-default
  shape average 1024000
  fair-queue
  random-detect dscp-based
policy-map POL-2MB-in
 class TELNET
  set ip dscp ef
 class HTTP
  set ip dscp af33
 class SSH
  set ip dscp af31

Att, 

Tiago Eduardo Zacarias

[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 whatsoever (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 egress is a full E1 (2 Mbps)?

As written, you policy won't prioritize, because all your defined classes are LLQ (which there is only one of) and your class default is shaped to 1 Mbps while the sum of your LLQ classes is less than 1 Mbps.  Also Cisco recommends not to configure more than 1/3 your bandwidth for LLQ.  I would suggest you not use EF for telnet traffic.

You ingress policy make is useless as it matches and marks the same DSCP settings.

(BTW, be careful with SSH, as it also may be SCP.  I also recommend against using RED, unless you're a QoS expert.)

You might try something like:

class-map match-any LLQ
 description CLASS-TELNET - PRIORITARIA - EF
 match ip dscp ef 
class-map match-any critical
 description CLASS-SSH and HTTP - CRITICAL - DATA 
 match ip dscp af31 af33

policy-map POL-2MB-out
 class LLQ
  priority percent 33
 class critical
  bandwidth remaining percent 75
  fair-queue
 class class-default
  bandwidth remaining percent 25
  fair-queue
 

[Tiago Eduardo Zacarias]

Good morning Colleague,

My output and full E1

Then understand the algorithm will not prioritize between classes by appointment f AF31 and due to this same applying LLQ?

The marking ingress policy'm doing to pass the markup for a network node that is behind this router so that the same control the ingress band that actually for him would be the egress.

This scenario of f to telnet marking was only to test it in the lab do not own voice I chose to telnet to better understand how it would prioritize the telnet stream that will later be applied in voice.

I thank you for the help

[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 whatsoever (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

Then understand the algorithm will not prioritize between classes by appointment f AF31 and due to this same applying LLQ?

There were two issues with your policy.  First, on most platforms there's only one LLQ queue.  When you define multiple priority classes, all their traffic goes to the same queue, FIFO.  The only advantage is, when you define multiple priority classes, each will have a different implicit policer.

The second issue is, your only other class, class-default, is shaped less than the combined bandwidth of all your other classes, and as they are LLQ, if there's any congestion, their implicit policers will engage.