Hello , i have a few QoS questions about the 4500 switch which were not satisfied in the 3com_Switch4500_Configuration_guide.pdf ,
and would greatly appreciate anyone who can answer any of the following questions :
A few questions :
1) .What is the , "weight" , parameter in "queue-scheduler wrr <queue-1weight> <queue2-weight>
... "
Is it the number of packets allowed to be dequeued on each pass the WRR algorithm does around the queues ?
2 ) What is the "bandwidth" parameter in - "queue-scheduler wfq <queue0-width> <queue1-width <queue2-width> ..."
Does the algorithm pass around the queues dequeing the specific amount of bandwidth from each queue
, or does "bandwidth" modify some other unknown value inside the queues?
I also noticed in system-view mode , that the maximum band width value is in the range of INTEGER<0-99968> :
queue-scheduler wfq ?
INTEGER<0-99968> Band width of queue 0 (Kbps). It must be a multiple of 64 in this port.
that is aprox 100mbit , why are 32bit/4bytes spared ?
3) The INTEGER<0-99968> range for wfq is not enouth for the dual personality gigabit ports
Intersting though when i try the command under a gigabit port i see :
-GigabitEthernet1/0/49]queue-scheduler wfq ?
INTEGER<0-1000000> Band width of queue 0 (Kbps). It must be a multiple of 64 in this port.
which is 1000mbit or 1Gbit , and no bytes spared like in <0-99968> range , does this also mean i have to go to each gigabit port and configure 'queue-scheduler wfq' bandwidth with values higher than allowed in system-view mode to take full utilization of the gigabit port?
4) What are the buffer space size of the 8 queues ? if all my traffic has Cos0 and cos-to-local
mapping says Cos0 goes to queue number 1 , then the 7 other queues have buffer space wasted that isn't being used at all?
Wouldn't it be more efficient to have only 1 standard queue but use 7 drop Thresholds instead ,
like in this example :
Drop thresholds define the amount of the total Layer 2 buffer use that must be reached before a specified class of traffic is dropped. In other words, this is how much the bucket needs to fill before a decision is made to begin dropping traffic of a specific class. Some switches have one
priority queue and one standard queue, for instance, with four drop thesholds on the nonpriority queue, refered to as 1p1q4t. For the traffic placed into the nonpriority queue, the four drop thresholds indicate which class of traffic, based on CoS, should be dropped most agresssively in the event that a percentage of the bucket fills.
After the queue has reached 50 percent of capacity, any traffic classified with CoS of 0 or 1 becomes drop candidates to avoid congestion. If the queue continues to fill in spite of the drops, at 60 percent of capacity any traffic classified with a CoS of 0, 1, 2, or 3 becomes drop
candidates to avoid congestion. If the queue still continues to fill in spite of the drops, at 80 percent of capacity any traffic classified with a CoS of 0, 1, 2, 3, 4, or 5 becomes drop candidates to avoid congestion. At 100 percent of capacity, all traffic, regardless of classification, becomes drop candidates.
Thesholds provide an alternative to dividing the buffer space to add more standard queues, eliminating the risk of starving one queue while flooding another. Setting a drop theshold allows the entire buffer space to be used, decreasing the potential for instaneous buffer overrun for
higher-priority traffic in the standard queue.
5) what COS value do BPDU frames and RIP protocol get on the switch ,to what queue are they placed?
Thank you very much.
Marcelo.
and would greatly appreciate anyone who can answer any of the following questions :
A few questions :
1) .What is the , "weight" , parameter in "queue-scheduler wrr <queue-1weight> <queue2-weight>
... "
Is it the number of packets allowed to be dequeued on each pass the WRR algorithm does around the queues ?
2 ) What is the "bandwidth" parameter in - "queue-scheduler wfq <queue0-width> <queue1-width <queue2-width> ..."
Does the algorithm pass around the queues dequeing the specific amount of bandwidth from each queue
, or does "bandwidth" modify some other unknown value inside the queues?
I also noticed in system-view mode , that the maximum band width value is in the range of INTEGER<0-99968> :
queue-scheduler wfq ?
INTEGER<0-99968> Band width of queue 0 (Kbps). It must be a multiple of 64 in this port.
that is aprox 100mbit , why are 32bit/4bytes spared ?
3) The INTEGER<0-99968> range for wfq is not enouth for the dual personality gigabit ports
Intersting though when i try the command under a gigabit port i see :
-GigabitEthernet1/0/49]queue-scheduler wfq ?
INTEGER<0-1000000> Band width of queue 0 (Kbps). It must be a multiple of 64 in this port.
which is 1000mbit or 1Gbit , and no bytes spared like in <0-99968> range , does this also mean i have to go to each gigabit port and configure 'queue-scheduler wfq' bandwidth with values higher than allowed in system-view mode to take full utilization of the gigabit port?
4) What are the buffer space size of the 8 queues ? if all my traffic has Cos0 and cos-to-local
mapping says Cos0 goes to queue number 1 , then the 7 other queues have buffer space wasted that isn't being used at all?
Wouldn't it be more efficient to have only 1 standard queue but use 7 drop Thresholds instead ,
like in this example :
Drop thresholds define the amount of the total Layer 2 buffer use that must be reached before a specified class of traffic is dropped. In other words, this is how much the bucket needs to fill before a decision is made to begin dropping traffic of a specific class. Some switches have one
priority queue and one standard queue, for instance, with four drop thesholds on the nonpriority queue, refered to as 1p1q4t. For the traffic placed into the nonpriority queue, the four drop thresholds indicate which class of traffic, based on CoS, should be dropped most agresssively in the event that a percentage of the bucket fills.
After the queue has reached 50 percent of capacity, any traffic classified with CoS of 0 or 1 becomes drop candidates to avoid congestion. If the queue continues to fill in spite of the drops, at 60 percent of capacity any traffic classified with a CoS of 0, 1, 2, or 3 becomes drop
candidates to avoid congestion. If the queue still continues to fill in spite of the drops, at 80 percent of capacity any traffic classified with a CoS of 0, 1, 2, 3, 4, or 5 becomes drop candidates to avoid congestion. At 100 percent of capacity, all traffic, regardless of classification, becomes drop candidates.
Thesholds provide an alternative to dividing the buffer space to add more standard queues, eliminating the risk of starving one queue while flooding another. Setting a drop theshold allows the entire buffer space to be used, decreasing the potential for instaneous buffer overrun for
higher-priority traffic in the standard queue.
5) what COS value do BPDU frames and RIP protocol get on the switch ,to what queue are they placed?
Thank you very much.
Marcelo.