
DLSw Implementation Notes
304233-A Rev. 00
2-17
Nonordered Queues
Although Site Manager numbers the DLSw priority queues, the queue number
hierarchy does not determine priority. You configure both the number of queues
for each DLSw peer and the percentage of bandwidth assigned to each queue. For
example, you can assign 50 percent of available bandwidth to Q3 and 25 percent
each to Q1 and Q0.
The Dequeuing Process
The algorithm for DLSw bandwidth allocation is called weighted dequeuing. With
weighted dequeuing, packets at the front of the protocol prioritization queues
enter a dequeue list and receive a weighted score. Packets with the lowest score
are transmitted first.
By distributing the selection of packets from all queues, weighted dequeuing is
more stable than the algorithms used for circuit-level WAN protocol prioritization.
One goal of weighted dequeuing is to send smaller packets ahead of large packets,
without violating the bandwidth of each queue or depriving large packets. The
algorithm accomplishes this by putting smaller packets ahead of larger packets by
simultaneously considering how long the larger packets have been in the dequeue
list. A large packet accumulates credit (lowering its weighted score) as each
smaller packet gets ahead of it, and eventually the larger packet moves to the front
of the dequeue list.
The Dequeue List
Three factors determine a packet’s weighted score:
• Size of the packet
• Percent of bandwidth allocated to the packet’s queue
• Time spent in the dequeue list
The dequeuing algorithm calculates a packet’s dequeue weight using the
following formula:
Weight =
Size of packet
Bandwidth %
- Time in dequeue list
Comentários a estes Manuais