Enterasys 1G58x-09, Руководство по настройке

Страница: 420 / 808

Port Priority and Classification Configuration Summary
Priority Queueing Modes (Algorithms)
9-2 Matrix E1 Series (1G58x-09 and 1H582-xx) Configuration Guide
9.1.2 Priority Queueing Modes (Algorithms)
The transmit queues for each port on the device can be configured with different queueing
algorithms, as described in the following subsections.
Strict Priority Queueing (SP)
SP queueing provides higher priority queues with absolute preferential treatment over low priority
queues, which minimizes the queueing delay of frames from the higher queues. The transmit port
does not serve a transmit queue unless all higher priority queues are empty.
Weighted Round Robin (WRR)
The frames are emptied out of the four queues with a weighted priority expressed in a percentage
of total traffic for each queue. The weighted queues are served in round-robin order with a
configured weight for each queue. The four queue settings must add up to 100 percent. This
guarantees minimum bandwidth for each queue in all cases, and can ensure that no queue reaches
more than a predetermined proportion of the overall capacity (Guarantee Maximum Bandwidth)
under stress.
Hybrid Queueing
There are two modes of hybrid queueing:
Mode 1
In mode 1, the highest queue (Q3) has a strict priority over all other three queues, so that the frames
in the lower three transmit queues are not served until there are no frames in the highest queue.
When there are no frames in the highest queue, the other three queues are served using the WRR
algorithm according to weighted queues 0, 1, and 2.
Mode 2
In mode 2, the highest two queues (Q3 and Q2) have a strict priority over the other two queues, so
that the lower two transmit queues are not served until the highest two queues (first Q3, then Q2)
are empty. When there are no more frames in the highest two queues, the frames in Q1 and Q0 are
served using the WRR algorithm according to the weighted queues 1 and 0.
A typical implementation of hybrid queueing is to redirect “Expected Service” traffic to the highest
priority queue, which will have strict priority over all other queues. Therefore, when combined with
the appropriate admission control, it can have a guaranteed delay for the frames it holds while the
other 3 queues run in WRR for “assured bandwidth” traffic.