QoS issue

[syntaxmonster]

Dear all,

i have an issue here which gives me a lot of headakes.

We have one Catlayst 6509 and one 6506. This two are linked over a 10Mbit/fdx WAN connection.

In both locations we have a Codian Lifesize Videoconference System.

This systems are marking the IP packets

voice: 101110 (EF)

video:100010 (AF41) so that i trust on all ports the dscp value (from Layer 2 to the Layer 3 interfaces). Now i have o both sides following configuration:

class-map match-all CLASSIFY-VOICE

match access-group name CLASSIFY-VOICE

class-map match-all CLASSIFY-OTHER

match access-group name CLASSIFY-OTHER

!

policy-map QoS-10Mbit

class CLASSIFY-VOICE

trust dscp

class CLASSIFY-OTHER

police 7000000 125000 125000 conform-action set-dscp-transmit default exceed-action drop

ip access-list extended CLASSIFY-OTHER

permit ip any any

ip access-list extended CLASSIFY-VOICE

permit ip any any dscp ef

permit ip any any dscp af41

!

Now to test that Qos is working we made a huge filetransfer between this locations and held in the same time a conference. The bandwith was occupying those 7Mbit as in the policy-map. The problem was that the sound and video was distorted.

Now here is the output of sh mls qos ip from both switches:

Switch 1:

sw-20-cd01#sh mls qos ip

QoS Summary [IPv4]: (* - shared aggregates, Mod - switch module)

Int Mod Dir Class-map DSCP Agg Trust Fl AgForward-By AgPoliced-By

Id Id

-------------------------------------------------------------------------------

Gi3/42 5 In CLASSIFY-V 0 1 dscp 0 800333710 0

Gi3/42 5 In CLASSIFY-O 0 2 No 0 29606082264 473387588

All 5 - Default 0 0* No 0 447495963416 0

Switch 2:

sw-10-cd01#sh mls qos ip gig1/43

[In] Policy map is QoS-10Mbit [Out] Default.

QoS Summary [IPv4]: (* - shared aggregates, Mod - switch module)

Int Mod Dir Class-map DSCP Agg Trust Fl AgForward-By AgPoliced-By

Id Id

-----------------------------------------------------------------------------------

Gi1/43 5 In CLASSIFY-V 0 1 dscp 0 1897936817 0

Gi1/43 5 In CLASSIFY-O 0 2 No 0 4534286636 4917207

I don't understand why qos doesn't work in my case. Any help/hint are much appreciated.

Thank you,

Chris

[syntaxmonster]

Edison and marwanshawi, thank you both for your replies.

I think a lightbulb just turned on over my head :D

So i have to figure out how much bandwith video and voip is needed and configure te port accordingly.

So if this hapens am i right to say that the access lists, class-maps and the policy-map isn't needed anymore as the packets have already the right DSCP value while entering the switch?

Thank you guys, you helped me much to see the whole QoS topic on the Cat6500 clearer.

If i succeed (and i will succeed ;) to have the desired qos configuration up and running, i'll post the runnig config.

Chris

[Marwan ALshawi]

yes if you allocated the bandwidth accordigly you dont need the class map

this is per port

and as i mentioned above each cisco cat6500 line card has diffrent queue/throshoulds number this is imortant to know

for example:

WS-X6704-10GE (1P7Q8T)

wrr-queue bandwidth 5 25 20 20 20 5 5

! Sets the WRR weights for 5:25:20:20:20:5:5 (Q1 through Q7)

wrr-queue queue-limit 5 25 10 10 10 5 5

! Allocates 5% to Q1, 25% to Q2, 10% to Q3, 10% to Q4,

! Allocates 10% to Q5, 5% to Q6 and 5% to Q7

while WS-X6408A-GBIC (1P2Q2T)

wrr-queue bandwidth 30 70

! Sets the WRR weights for 30:70 (Q1:Q2) bandwidth servicing

wrr-queue queue-limit 40 30

see the deffrrence in the config based on the line card !!!!

and also put the voice in the priority queue and enable the priority queue in the interface level

priority-queue cos-map 1 5

also cisco 6500 if you apply the above qos config for one port in a line card it will be applied automatically to all the ports in that line card !!

good lcuk

HTH

[syntaxmonster]

Hi marwanshawi,

yes, my linecard has the 1p3q8t:

Interface GigabitEthernet1/46 queueing strategy: Weighted Round-Robin

Port QoS is enabled

Trust state: trust DSCP

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

-----------------------------------------

01 WRR 08

02 WRR 08

03 WRR 08

04 Priority 01

WRR bandwidth ratios: 100[queue 1] 150[queue 2] 200[queue 3]

queue-limit ratios: 50[queue 1] 20[queue 2] 15[queue 3] 15[Pri Queue]

Chris

[syntaxmonster]

OK i've unbound the service policy from the WAN

interfaces with:

no service-policy input QoS-10Mbit on both swithces

The cos to dscp map looks like this on both switches:

mls qos map cos-dscp 0 8 16 26 34 46 48 56

The transmit queue 1p3q8t has the default values on both sides.

As mentioned in this dokument:

http://www.cisco.com/en/US/docs/switches/lan/catalyst6500/ios/12.2SXF/native/configuration/guide/qos.html#wp1478769

the Q1 has default 50% of the bandwith. Now i did a huge filetransfer and the bandwith utilization went to almost 90%, hence almost 10Mbit.

How can this be as traffic from the Q1 queue only gets 50%?

I kind of confused :(

Chris

[Jon Marshall · Accepted solution]

Chris

Remember that queueing only comes into effect at times of congestion. So if there is no other traffic on the link then there is no reason why your'e file transfer cannot use the whole link.

Jon

[syntaxmonster]

Ahhhhh.....:)

Jon, thank you so much for clarifying this.

[Marwan ALshawi]

hi Jon

but i think apply only to policy maps

while when you allocate amount of bandwidth at interface leve using wrr this will limit the traffic for that COS or DSCP regardless the interface conjested or not !!

am i right !!

[Jon Marshall]

Marwan

As with all things to do with Catalyst QOS it is not always that obvious :-).

My understanding is that all traffic will be sent on the link regardless of the queue length UNTIL congestion starts and then and only then will the traffic be limited to it's queue size.

There is a queueing algorithm called Shaped round robin (SRR)-SRR which allows a queue to use only the allocated bandwidth but Chris isn't using this.

Having said all that, this is only my understanding of how it works so happy to be corrected if it is not right. Unfortunately i don't have a 6500 to test with :-)

Jon

[syntaxmonster]

Well, i can test this tomorrow as we have a maintenance window. Then i can start a filetransfer and simultaneously generate voip/video.

Chris

[Marwan ALshawi]

Hi jon

when we assigning amount of bandwidth to each queu here works like police in MQC

which is workiing with conjestion and without as well

[syntaxmonster]

Hi Jon,

well i've tested this configuration and it seems that it worked.

I've made a huge filetransfer while i was holding a videoconference whith a colleague. The normal (DSCP 0) traffic was occupying about 70% from the available bandwith while the DSCP 46 and 34 traffic didn't had any packet loss or quality problems.

So i think this worked.

Thank you once again @all

P.S. is there a way to see how the drop ratios for the queues are and how many packets where droped?