Configuring Class of Service

Unanswered Question
Nov 4th, 2009
User Badges:

Hi,


I am new to this subject matter and have done a bit of reading up which seems to me that CoS is nearly same as QoS.


However, I will appreciate if I can get pointers on how o go about configuring CoS on a network. The client simply wants to achieve the ability to split traffic on their network over two separate WAN links.


Any help will be greatly appreciated.


Regards,


Femi

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 3 (1 ratings)
Loading.
Nagendra Kumar ... Wed, 11/04/2009 - 22:56
User Badges:
  • Cisco Employee,

Hi,


QoS is a feature which classifies and prioritize traffic based on requirement whereas, COS is a 3 bit field in 802.1q header which will be used as part of QoS configuration in layer2 network.


Normally, Service providers use COS term to intimate QoS.


From your requirement in above question, Am I right understanding that you need to load balance outgoing traffic via 2 WAN link?.


If so, you can use GLBP or multi group HSRP in case of Layer2 network or routing protocol tuning in case of Layer3 network.


Let us know if your requriement is different.


HTH,

Nagendra

Nagendra Kumar ... Wed, 11/04/2009 - 23:51
User Badges:
  • Cisco Employee,

Hi,


QoS is a feature which classifies and prioritize traffic based on requirement whereas, COS is a 3 bit field in 802.1q header which will be used as part of QoS configuration in layer2 network.


Normally, Service providers use COS term to intimate QoS.


From your requirement in above question, Am I right understanding that you need to load balance outgoing traffic via 2 WAN link?.


If so, you can use GLBP or multi group HSRP in case of Layer2 network or routing protocol tuning in case of Layer3 network.


Let us know if your requriement is different.


HTH,

Nagendra

femi.agboade Thu, 11/05/2009 - 13:42
User Badges:

Hi Nagendra,


Apologies for responding rather late on.


You are kinda right. The client has 2 WAN links, a primary and a secondary failover that all traffic is cut over to when link 1 fails.


Now a new service is to be deployed, VoIP over the WAN link and the client has decided to have the voice traffic routed only through the 2nd WAN link which of course sounds like load balancing to me also. How CoS comes about is what aint clear to me...


Regards,


Femi

Joseph W. Doherty Thu, 11/05/2009 - 19:11
User Badges:
  • Super Bronze, 10000 points or more

The intention might be to use VoIP frame/packet marking to send to 2nd link. (Often VoIP devices can mark L2 and/or L3 ToS.)


Perhaps a better deployment would be to utilize both links but with VoIP prioritized on both.

femi.agboade Thu, 11/05/2009 - 23:06
User Badges:

Hi,


The reason for directing VoIP traffic on the 2nd link is because the primary link is heavily utilized and bandwidth will be an issue. But since the secondary link sits idle most of the time, we are of the opinion that moving the VoIP traffic out through that link will be fine.


Prioritizing VoIP over other network traffic may impact other business processes.


Tagging may be a way to go, will investigate the possibility.


Regards,


Femi

Joseph W. Doherty Fri, 11/06/2009 - 05:10
User Badges:
  • Super Bronze, 10000 points or more

"The reason for directing VoIP traffic on the 2nd link is because the primary link is heavily utilized and bandwidth will be an issue. But since the secondary link sits idle most of the time, we are of the opinion that moving the VoIP traffic out through that link will be fine.


Prioritizing VoIP over other network traffic may impact other business processes. "


Yes, but consider if both link are utilized for all traffic, with QoS for both, you then have the extra bandwidth capacity for all traffic and VoIP works correctly. Further, if either link fails, VoIP and perhaps other most critical traffic continues to perform as necessary but only then does other traffic suffers from loss of the failed link's bandwidth.


The foregoing assumes both WAN links are of about similar capacity. If they aren't, then it might make sense to manually direct traffic, but if the devices support OER/PfR, or perhaps even EIGRP with variable cost routing, I think, are a better approaches.

Nagendra Kumar ... Thu, 11/05/2009 - 21:18
User Badges:
  • Cisco Employee,

Femi,


VOIP-----L2network----WAN_Router


If your L2 network is choked or highly utilized, VoIP traffic over this network definitely needs L2QoS.


If this is your case, you may have to consider Class of Service.


P.S : If L2 network doesn't have multiple switches involved (with trunking), CoS may not help you much.


HTH,

Nagendra

femi.agboade Thu, 11/05/2009 - 22:56
User Badges:

Hi,


The client has complained about bandwidth congestion on the primary link, so I assume that yes, the L2 network might be highly utilized at the moment.


As this seems to be the case, any advice on how to go about setting up CoS?


Regards,


Femi

Nagendra Kumar ... Thu, 11/05/2009 - 23:13
User Badges:
  • Cisco Employee,

Femi,


Primary WAN link congestion doesn't mean that L2 network will be congested.


Check if your L2 network is really congested. If so, CoS may help you in getting voice traffic prioritized.


Below is a link for QoS configuration on 2900 switches.


https://www.cisco.com/en/US/docs/switches/lan/catalyst2950/software/release/12.1_20_ea2/configuration/guide/swqos.html


As said earlier, if your L2 network only has one switch or doesn't have trunking, it may not help you much as CoS marking is done on 802.1q header which is part of trunking.


HTH,

Nagendra

femi.agboade Fri, 11/06/2009 - 03:39
User Badges:

Hi,


Thanks for this. I will go through it and update you.


Regards,


Femi

femi.agboade Wed, 11/11/2009 - 04:36
User Badges:

Hello,

I apologize for getting back a bit later than I initially planned. After meeting with the client, what they want to do is quite simple and straight forward (I guess?).

In their words, “Please help to separate traffic to a redundant link. Simply looking for separation of port 443 traffic on primary link from port 80 traffic on the redundant link. The architecture should simply be EIGRP, and NO static route definitions.”

I have a fair idea what to do, but pointers will be appreciated.

Regards,

Femi


Joseph W. Doherty Thu, 11/05/2009 - 05:49
User Badges:
  • Super Bronze, 10000 points or more

CoS (class of service) is often a subset of QoS (quality of service) and/or refers to L2 CoS priority bits. QoS might include the whole methodology of traffic treatment beyond just best effort, or might refer to L3 ToS (type of service) byte's usage of IP Precedence and/or DSCP.


Provide pointers CoS configuration depends both on what you're trying to accomplish and what your equipment is capable of. You mention you want to split traffi across two separate WAN links. L2 CoS or L3 ToS markings might assist, or they might be overly complex. Again, depends on exactly what you're trying to do. There are other methods for directing traffic (e.g. PBR) to specific WAN links and/or load balance/share multiple WAN links that wouldn't require using CoS or ToS markings.

Mohamed Sobair Wed, 11/11/2009 - 08:23
User Badges:
  • Gold, 750 points or more


Femi,


L2 CoS in QoS has different objectives than what you have described.


I think the best approach for you as per your client requirment is to implement Policy based routing PBR to have port443 traverse one link and HTTP through the second link.


QoS comes into play if you want to apply shaping/policing congestion managment and avoidance mecahnisms. Or If you want to classify your traffic using NBAR, but all of those has to do with classification/marking , and applying your QoS policies accordingly, which is not desirable in your case.



HTH

Mohamed

femi.agboade Thu, 11/12/2009 - 23:26
User Badges:

Hi Mohamed,


Thanks for your comment. Based on PBR as you suggested, below is a config i think should work:


! #identify packets on port 80

!

ip access-list extended port80

permit 80 any any

!

! #identify packets on port 443

!

ip access-list extended port443

permit 443 any any

!

! #Define route maps to control where packets are output, match access list and set output interface for the packets

!

route-map map-port permit 10

match address port80

set interface fastEthernet 2

!

route-map map-port permit 20

match address port443

set interface fastEthernet 3

!

! #Specify the route map to use on the inbound interface

!

interface fastEthernet 1

ip policy route-map map-port

!


What do you think? Am I missing anything? Though I feel this seems like static routing which the client is trying to guard against...


Regards,


Femi

Actions

This Discussion