04-01-2005 11:23 AM - edited 03-02-2019 10:20 PM
Hi, folks:
I have a dummy question. In one document, I read that each physical port only have 1 FIFO based hardware queue. The QoS queueing tools like PQ, CQ, WFQ, etc only work on the software queue.
However, another document mentions that, by issuing "show interface type mod#/port# capabilities" command, you can find out the type of queuing on the port. For example:
R# show interface gigabitethernet 6/2 capabilities
........
QOS scheduling: rx-(1p1q4t), tx-(1p2q2t)
........
According to the output, this port has 1 priority queue and 2 WRR queues.
I would like to know the term of "queue" here are referring to the software queue or hardware queue? I feel confused now.
Thanks a lot!
cableip
04-01-2005 08:49 PM
Hi,
It's referring to software queue. Any queuing techniques designed to handle only software queues.
Hardware queues are only FIFO, you can monitor them with "sh controllers
-Ashok.
04-02-2005 02:46 PM
Hi, Ashok:
Thank you very much for your reply. So does this mean, even if I used CQ tool, which supports 16 queues thereotically, I still can only treat the port as 3-queue system?
Because CQ doesn't have the concept of priority queue. Does this mean it will alter the way the physical port schedule the delivery of outbound packets? If I apply CQ on the port, does this mean the original strict priority queue will disappear or disabled automatically?
Thanks!
Shixiong
04-02-2005 11:09 PM
Hi Shixiong,
No, CQ consists of 16 software queues which by default transmits the packets in round robin fashion from queue by queue.
packets -> if congestion, into any of software queue as designed (PQ, CQ, WFQ etc) -> Hardware queue (FIFO)
packets -> if no congestion, directly into Hardware queue (FIFO)
CQ also have one priority queue called as Queue 0 which generally used for management traffic. And the other 16 queues are in user control, which traffic to put into which queue.
You will prioritize even in CQ by defining queue-limit and bytecount parameters. This means for eg you have defined as queue-limit as 10 for Q 1 and 50 packets for Q 2. So, CQ serves Q 1 10 packets and then moves to Q 2 and serve 50 packets (you are allocating more bandwidth for Q2).
Hope you have clarified...
Thanks...
-Ashok.
04-06-2005 02:47 PM
Hi, Ashok:
Thank you very much again. You are correct. CQ has preemptive queue. I already forgot it. 8-)
If I apply CQ tool on this interface (1p2q2t), even if CQ allows me to use up to 16 queues, I still only have 2 regular queues to play with? Am i correct?
Thanks!
Shixiong
04-06-2005 08:52 PM
Hi Shixiong,
Oh ok! I have understood your confusion.
You have been talking about WRR queues available on the Cat 6K modules. Please note that these modules can only support WRR queues and types like 1p2q4t depends upon the module. There is no separate queuing technique like CQ or PQ to apply for them. The WRR queuing techniques are more than sufficient.
The CQ, PQ, CBWFQ queuing techninques are configurable mainly on Cisco routers.
Please see the following links to get clear idea...
http://www.cisco.com/warp/public/473/73.html
http://www.cisco.com/warp/public/473/60.html
Thank you,
Regards...
-Ashok.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide