Cisco Inter-Switch QOS problem

[keithja]

Hello,
I'm working on modifying some of our 3560 and 3750 switch configs to better accomodate Nortel IP 200x phones on a Cisco network, per Nortel's "ClientsInterop-cisco-L2" doc and I'm not seeing what I expect from one switch to the next.

Let's pretend a layout where an IP phone is connected directly to the WG switch P1, with no PC in the mix for the moment.

The phone connects to WG-FA/1. From there an uplink port, Gi0/1 connects the WG switch to a backbone switch on P Gi0/2.

When I look at the QOS stats via command Line, (show mls qos int [int] statistics),

I see COS 6 into the switch from the phone (fa/1),
and I see COS 5 and 6 outbound on the uplink (Gi0/1)


I see DSCP 40 and 46 into the switch from the phone (fa/1), and I see DSCP 40 and 46 outbound on the uplink (Gi0/1)

- all well and good. But when I look at stats on the backbone switch for int Gi0/2 I see:

***
DSCP 40 and 46 inbound, which is fine,
but COS 5 Inbound instead of 6.
***

My WG switch Global config (as pertinent to QOS)
mls qos map cos-dscp 0 8 16 24 32 40 46 56
mls qos srr-queue input bandwidth 90 10
mls qos srr-queue input priority-queue 2 bandwidth 30
mls qos

and the Fa/1 int config QOS settings are:
switchport voice vlan 100
priority-queue out
mls qos trust dscp

and the uplink port:
interface GigabitEthernet0/1
switchport trunk encapsulation dot1q
switchport mode trunk
switchport nonegotiate
priority-queue out
mls qos trust dscp
macro description cisco-switch
spanning-tree link-type point-to-point

Since I see pretty much what I expect entering and leaving the WG switch, I assume I'm ok here.

On the BB switch, the global QOS settings are:
mls qos map cos-dscp 0 8 16 24 32 40 46 56
mls qos srr-queue input bandwidth 90 10
mls qos srr-queue input priority-queue 2 bandwidth 30
mls qos

- the same as WG -

the Uplink port settings are:
switchport trunk encapsulation dot1q
switchport mode trunk
priority-queue out
mls qos trust dscp
macro description cisco-switch
spanning-tree link-type point-to-point

I has to be something dumb I'm overlooking. Can anyone see why I am getting CoS 5 into the bb uplink switch instead of CoS 6?

Thx
Tarred&Fractured

 

[North323]

im not a QOS guru but I dont think QOS comes into play until there is congestion in the network. so you may not see what you want because you are only dealing with one phone

 

[Belushi]

Not sure the answer, but one thing I do know. I would strongly suggest not having your phone mark traffic with COS 6. I would generally reccomend marking at COS 5 with a DSCP value of 46 (EF). Marking anything above COS 5 is generally reserved for switch protocol and control plane traffic.

 

[keithja]

I got it.

Since Cisco Catalyst retags the packet at each switch based on the DSCP, once it hit the second switch it was reclassifying based on it's default dscp to cos table - which flagged DSPC 46 as 5. So applying a remap:
mls qos map dscp-cos 46 to 6

to each switch allowed it to maintain the expected CoS 6 value.

However in regard to Belushi's remarks, I haven't heard of this before. Here is an excerpt from Wikipedia:

802.1p defines eight different classes of service which are available, usually expressed through the 3-bit user_priority field in an IEEE 802.1Q header added to the frame. The way traffic is treated when assigned to any particular class is undefined and left to the implementation.


The IEEE however has made some broad recommendations:[1]

User priority Traffic Type
1 Background
2 Spare
0 Best Effort
3 Excellent Effort
4 Controlled Load
5 Video
6 Voice
7 Network Control


Which seems to imply that CoS 6, as well as 5, should be an acceptable marking: and the 6 I am working with is per Nortel recommendations for Cisco switches.

Can anyone shed some light on this issue?

Thx
T&F

 

[keithja]

Or alternately, and probably preferred,
mls qos trust cos
on the incoming interface.

and to add to the confusion, a new question:

I can now see CoS 6 traffic coming into the uplink interface using:

show mls qos int [int] statistics

However when I SPAN the uplink port to a traffic analyzer, it's showing it as CoS 5 instead of CoS 6 that the switch statistics claim.

I have the Span type set to replicate. I would think it should be unnecessary (and undesireable) to have to set any specific QOS commands on the Span destination port. True? Why am I not seeing CoS 6 at the analyzer?

Thx again

 

[keithja]

Did anyone have any ideas on the SPAN issue? Or the CoS 6 vs 5, or my understanding of the solution to the original problem?

 

[VinceWhirlwind]

The generic DSCP values used throughout the world of real-time traffic are AF41 & EF, neither of which normally maps to COS6.
Default queue settings are ideal for voice packets being on COS5, not 6, which is used (as explained above) for something else.

Changing the default settings would seem like pointless effort and increased risk.

I would never even *look* at Wikipedia, let alone rely on it for accurate technical information.