MLS QOS 3750E

Gabriel Hill · ‎11-26-2012

Okay, I am stepping in some completely unfamiliar territory that I need advise on. About a year ago, the company I work for added a new DR site. This site is connected by a 2 1gigabit connections. We had a lot of trouble with performance, and we had a consulting company helping us work through some of the issues. One of the things they had added was a QOS configuration to the 3750E's that the p2p link lands on.

3750E-STACK#sh running-config | include mls

mls qos srr-queue input threshold 1 80 90

mls qos srr-queue input threshold 2 80 90

mls qos srr-queue input buffers 80 20

mls qos srr-queue input cos-map queue 2 threshold 1 3 4

mls qos srr-queue input cos-map queue 2 threshold 2 6 7

mls qos srr-queue input cos-map queue 2 threshold 3 5

mls qos srr-queue input dscp-map queue 1 threshold 2 8 9 10 11 12 13 14 15

mls qos srr-queue input dscp-map queue 1 threshold 2 16 17 18 19 20 21 22 23

mls qos srr-queue input dscp-map queue 1 threshold 3 24 25 26 27 28 29 30 31

mls qos srr-queue input dscp-map queue 1 threshold 3 32 33 34 35 36 37 38 39

mls qos srr-queue input dscp-map queue 2 threshold 1 48 49 50 51 52 53 54 55

mls qos srr-queue input dscp-map queue 2 threshold 2 56 57 58 59 60 61 62 63

mls qos srr-queue input dscp-map queue 2 threshold 3 40 41 42 43 44 45 46 47

mls qos srr-queue output cos-map queue 2 threshold 2 1 2

mls qos srr-queue output cos-map queue 2 threshold 3 3 4

mls qos srr-queue output dscp-map queue 1 threshold 1 0 1 2 3 4 5 6 7

mls qos srr-queue output dscp-map queue 1 threshold 1 48 49 50 51 52 53 54 55

mls qos srr-queue output dscp-map queue 1 threshold 2 56 57 58 59 60 61 62 63

mls qos srr-queue output dscp-map queue 1 threshold 3 40 41 42 43 44 45 46 47

mls qos srr-queue output dscp-map queue 2 threshold 2 8 9 10 11 12 13 14 15

mls qos srr-queue output dscp-map queue 2 threshold 2 16 17 18 19 20 21 22 23

mls qos srr-queue output dscp-map queue 2 threshold 3 24 25 26 27 28 29 30 31

mls qos srr-queue output dscp-map queue 2 threshold 3 32 33 34 35 36 37 38 39

mls qos queue-set output 2 threshold 1 3200 3200 100 3200

mls qos queue-set output 2 threshold 2 200 200 25 400

mls qos queue-set output 2 threshold 3 200 200 25 400

mls qos queue-set output 2 threshold 4 200 200 25 400

mls qos queue-set output 2 buffers 70 10 10 10

mls qos

I understand some of this after reviewing this document:

https://supportforums.cisco.com/docs/DOC-8093

This is what I am confused about. When I look at one of the interfaces that goes over the WAN, I see that almost all of the traffic is tagged with DSCP 0-4 and DSCP 45-49 (sh mls qos int gig2/2/48 statistics). When I do a "sh mls qos int buffers" I can see that all interfaces are mapped to qset #1. Which buffer allocations are 25 25 25 25. In the config I posted, queue-set#2 is allocated 70 10 10 10. I am trying to understand why there is even a configuration for queue-set #2 since no interfaces are tied to it. Wouldn't the allocation for queue-set#2 be appropriate for queue-set#1 since it has most of traffic going to queue#1? Any help understand this would be greatly appreciated.

Joseph W. Doherty · ‎11-26-2012

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

Perhaps queue-set was defined for some special purpose, but for whatever reason, never utilized for that purpose.

If I correctly recall some of the information in the document you reference, having a second queue-set allow some different QoS parameters for a set up ports of a different "kind". A good example would be user edge ports and the aggregation uplink port (your WAN port might qualify for different "kind" treatment).

You didn't provide enough information to suggest what might be good buffer settings for your usage, for example, whether egress queue 1 is being used for PQ.

That noted, the two usual schools of thought on buffer allocations are: allocating buffers resources similar to bandwidth allocations or allocating buffers resources the inverse of bandwidth allocations.

Do your QoS stats show drops? If so, at an unacceptable level?

Gabriel Hill · ‎11-27-2012

Hello Joseph,

You're correct, the document did mentioned different queue-sets for different port "types". The 3750E I am using shouldn't need two different queue-sets with this line of thinking. The switch has two port-channels, one going to another 3750E at a different location, and the other port-channel going to the local core device. All traffic going from DR to PROD or vice versa will transverse both port-channels.

I do show drops, it's "acceptable" but after looking through this I believe it could be greatly improved. As I said, this is completely unfamiliar territory, I've never worked this this kind of QOS configuration before. I am not sure what you mean when you say "whether egress queue 1 is being used for PQ". I am attaching some information below. Thanks for your help.

3750# sh mls qos queue-set

Queueset: 1

Queue : 1 2 3 4

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

buffers : 25 25 25 25

threshold1: 100 200 100 100

threshold2: 100 200 100 100

reserved : 50 50 50 50

maximum : 400 400 400 400

3750# sh platform port-asic stats drop gigabitEthernet 1/0/49

Interface Gi1/0/49 TxQueue Drop Statistics

Queue 0

Weight 0 Frames 0

Weight 1 Frames 0

Weight 2 Frames 0

Queue 1

Weight 0 Frames 72698676

Weight 1 Frames 1

Weight 2 Frames 0

Queue 2

Weight 0 Frames 0

Weight 1 Frames 0

Weight 2 Frames 0

Queue 3

Weight 0 Frames 0

Weight 1 Frames 0

Weight 2 Frames 40

Queue 4

Weight 0 Frames 0

Weight 1 Frames 0

Weight 2 Frames 0

3750#sh mls qos input-queue

Queue : 1 2

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

buffers : 80 20

bandwidth : 4 4

priority : 0 10

threshold1: 80 80

threshold2: 90 90

3750#sh mls qos maps dscp-input-q

Dscp-inputq-threshold map:

d1 :d2 0 1 2 3 4 5 6 7 8 9

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

0 : 01-01 01-01 01-01 01-01 01-01 01-01 01-01 01-01 01-02 01-02

1 : 01-02 01-02 01-02 01-02 01-02 01-02 01-02 01-02 01-02 01-02

2 : 01-02 01-02 01-02 01-02 01-03 01-03 01-03 01-03 01-03 01-03

3 : 01-03 01-03 01-03 01-03 01-03 01-03 01-03 01-03 01-03 01-03

4 : 02-03 02-03 02-03 02-03 02-03 02-03 02-03 02-03 02-01 02-01

5 : 02-01 02-01 02-01 02-01 02-01 02-01 02-02 02-02 02-02 02-02

6 : 02-02 02-02 02-02 02-02

One other question: If every DSCP value is mapped to either queue 1 or queue 2 would buffers even need to be allotted to tx queue 3/4?

Joseph W. Doherty · ‎11-27-2012

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

"Whether egress queue 1 is being used for PQ" refers to whether priority-queue out statement was configured on the interface. If it is, it impacts bandwidth management across the 4 egress queues which often has implications for buffer settings, which is what you were asking about.

So you're saying there's only two active "links", i.e. the two port-channels. If true, perhaps the best tuning change would be to much decrease the buffers reserved per port and much increase the common pool buffers. This should then allow the most buffer space for your two active "links".

If these two links show a large difference in traffic volume per direction (i.e. much more from PROD to DR), there might be an advantage treating the two port-channels as different "kinds".

I don't think it matters much if only using one set of queues for an interface. I don't recall buffers being reserved per queue, although, I believe, they are reserved per interface. However, as you won't be sharing buffers across queues, you should be able to push logical drop limits to their max values for the single queue.

Also, BTW, unlikely you need to worry about the input queue on a 3750E, as the fabric isn't oversubscribed as it can be on the original 3750 series.