SRR input scheduler on Cisco 3560

[Volodymyr Khrystenko]

Hi.

By default there 2 input queues. 2nd is priority queue and has 10% of link assigned. Weights for two queues are 4 and 4. For example we have 1Mbps link. ANd we also have saturated link with voice and other traffic. Will voice get 550k bandwidth? As I undestood 100k is assigned to voice strictly. Remaining 900k is evenly divided by two queues according to 4 and for ratio. Am I right?

[paul driver]

Hello

The srr bw weights are equal. After the PQ is serviced, any remainig BW will be shared between the to queues.

The default ratios could have been anything, it could have been "10 10". or "1 1" is the same, so this ratio identifies the ratio of the  frequency in which the srr scheduler sends packets from each queue. Therefore if we have Queue2 to be serviced twice as much a queue1, we could set the ratios to be "1  2"

By default priority as stated by you, the input queue 2 is allocated upto 10% of BW (this can be change to be queue 1

if need be)

The Weighted Tail Drop thresholds (WTD) are all 100% by default, indicating that packets will not be

dropped until  the buffers are completely full.

Queue     :       1       2

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

buffers   :      90      10

bandwidth :    4       4

priority  :       0     10

threshold1:   100     100

threshold2:   100     100

hope this helps

res

Paul

Please don't forget to rate any posts that have been helpful.

Thanks.

[Joseph W. Doherty]

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

The ingress queues on a 3560/3750 are for packets entering the fabric/ring.  As there's (relatively speaking) much bandwidth, it's unusual you need to modify the default queue settings.  Since your example uses a 1 Mbps link, could you clarify what you believe your need is?  (Also from your description, mixing voice and data on a saturated link, we normally look toward egress queuing support.)

Or, are you just trying to understand a 3560/3750 ingress queuing features?

[Volodymyr Khrystenko]

Im just trying to understand ingress QoS))) Since it is unusual for me to have ingress QoS.

[Joseph W. Doherty]

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

Ok, well in that case, the 3560's "ingress" QoS is really "egress" QoS from all the ingress ports to the fabric/ring.  It's very similar to a 3560's port egress QoS.  Instead of four queues you only have two.  You may select which traffic maps into which queue.  Each queue only has two WTD values.  These allow you to drop some traffic in a queue before other traffic.  You can allocate the amount of buffer resources per queue and you can also allocate their bandwidth allocations.  Lastly, you can also set the one queue to operate as a PQ, but unlike the port egress PQ you also set a bandwidth allocation (which makes this PQ more like a router's CBWFQ LLQ).

Why and what values you would set a 3560's "ingress" QoS to depends on what you're trying to accomplish.  Basically, as there are only two queues, it allows you to favor some traffic over other traffic.  Normally you might only favor real time traffic, such a VoIP bearer, or other extremely timing or drop sensitive traffic.

If you have an original series 3560, as the fabric is 32 Mbps, and you might have over 52 gig ports, it possible aggregate ingress bandwidth will exceed fabric capacity, so the "ingress" QoS features would allow you some action treating congestion.

With the later -E and -X series, the switch's fabric, I believe, can handle all possible ports at full bandwidth.  I.e.  there should never be ingress queuing.

With any 3570 model, the ring may be oversubscribed, so ingress QoS might be important for those switches.

[Volodymyr Khrystenko]

Ok. Interesting. But back to initial question. As I understtod I just asked incorrectly. So I rephrase:

I have default settings:

Cisco-3750-stack-1#sh mls qos input-queue

Queue     :       1       2

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

buffers   :      90      10

bandwidth :    4       4

priority  :         0      10

threshold1:     100     100

threshold2:     100     100

My fabric is 1Mbps and Im getting saturation from incoming links. How much bandwidth does second queue get and first?And what buffers mean here? Buffers==switch fabric? I dont getting "buffer" concept.

[paul driver]

Hello

At present you have priority queuing enabled for Q2  in the ingress queue to 10% of the line rate - Which will be serviced first.

After this you then have the default srr bw weight set  for both queues, so this means any remaining Bw , The srr scheduler will sends packets to each queue with a ratio of 1-1  so basically being serviced equally.

As for the buffers , My understanding is that they hold the packets to be queued,  so in this case Q1 has 90% buffer allocation with a 100% threshold limit meaning that when the whole 90% of this buffer allocation is reached for Q1 will begin to drop packets.

As for queue 2 the same apply's,  10% of buffer allocation with a threshold limit of 100%

res

Paul

Please don't forget to rate any posts that have been helpful.

Thanks.

[Joseph W. Doherty]

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

Paul's follow on reply answers your question although the ingress "line" rate would be fabric bandwidth.

In your question, it would all be percentages against your 1 Mbps (although actual fabric bandwidth is much, much more).

As the 3560 QoS features are the same as the 3750, you might look at 3750 documentation for additional QoS information, for example:

http://www.cisco.com/en/US/products/hw/switches/ps5023/products_tech_note09186a0080883f9e.shtml#cg24