Cisco 6509 QoS question...

[jonathanaxford]

HI experts,

I have a Cisco 6509E running advipservicesk9_wan-mz.122-18.SXF8.

I have quite a few ports on this device configured with Quality of Service to prioritise VoIP traffic. One of the modules is a WS-X6748-GE-TX that connects a number of remote sites, as well as a couple of directly connected servers.

One of the server ports is reporting output drops, but I haven't enabled any QoS configuration on this interface. The port is configured as follows:

interface GigabitEthernet4/1
description Connected to NPESX-LYAT1-Test
switchport
switchport access vlan 8
no ip address
speed 100
duplex full
spanning-tree portfast
end

When I run the show queueing interface gi 4/1 command I can see the following in the output:

Interface GigabitEthernet4/1 queueing strategy:  Weighted Round-Robin
  Port QoS is enabled
  Port is untrusted
  Extend trust state: not trusted [COS = 0]
  Default COS is 0
    Queueing Mode In Tx direction: mode-cos
    Transmit queues [type = 1p3q8t]:
    Queue Id    Scheduling  Num of thresholds

***REMOVED FOR CLARITY***

  Packets dropped on Transmit:

    queue     dropped  [cos-map]
    ---------------------------------------------
    1                      267  [0 1 ]
    2                        0  [2 3 4 ]
    3                        0  [6 7 ]
    4                        0  [5 ]

  Packets dropped on Receive:
    BPDU packets:  0

    queue              dropped  [cos-map]
    ---------------------------------------------
    1                        2  [0 1 2 3 4 5 6 7 ]

Now, I am not overly concerned about the number of output drops, but my question is - is Quality of service actually enabled on this port? I am assuming not as the port state is listed as "Untrusted" so all traffic will be re-written to CoS 0, correct?

Assuming that QoS is not enabled, then all of the available output buffer memory will be available to the default traffic on that port?

Many thanks for any help...

Jonathan

[Jon Marshall · Accepted solution]

Jonathan

As soon as you enable QOS globally on the 6500 the transmit queues are enabled so your port is indeed using the 1p3q8t architecture. This means that each of the 3q part gets allocated a certain amount of bandwidth to use.

Having said that because you have not configured QOS specifically on that port it is using WRR which basically means any unused bandwidth from the remaining queues ie. in your case any queue that is not queue 1, can be used by queue 1 so you should be okay.

And yes, without any explicit configuration all CoS will be set to 0 and you can see from the output that all packets are being placed into Q1.

Note also that from the configuration docs it is recommended that you configure all the ports on a linecard that share the same asic to have the same QOS settings or this can lead to inconsistencies. From the PFC QOS config guide -

General Guidelines

•Configure the same trust mode on all ports supported by an ASIC. Mismatched trust modes cause inconsistent bandwidth and queue-limit ratios.

here is the full link which also explains about WRR, queue structures etc. -

http://www.cisco.com/en/US/docs/switches/lan/catalyst6500/ios/12.2SXF/native/configuration/guide/qos.html

Jon

[jonathanaxford]

Hi Jon,

Many thanks for your help, I shall take a look through that doco with a view to standardising my config,

Cheers

Jonathan

[jonathanaxford]

Hi there,

If I am reading this correctly, the recommendation for me would be to simple enable the mls qos trust dscp command on all of the pther switchports sharing the same ASIC as my QoS configured ports, but simply leave all other QoS related config as default..? This would then ensure that the port trust state is idenitcal across all f the ports....

Many thanks

Jonathan