Cisco WS-X6324-100FX-MM Livro Branco - Página 11

Procurar online ou descarregar pdf Livro Branco para Hardware de rede Cisco WS-X6324-100FX-MM. Cisco WS-X6324-100FX-MM 18 páginas. Cisco catalyst x6524: supplementary guide
Também para Cisco WS-X6324-100FX-MM: Ficha de dados (9 páginas)

Imprimir página

Scheduling: DWRR

DWRR is a feature that is used on egress (transmit) queues. Explaining DWRR is best served by using an example. Assume a switch port has three

queues. Queue 1 has been given access to 50 percent of the bandwidth, queue 2 has 30 percent, and queue 3 has 20 percent. If the WRR algorithm

is servicing queue 2, and on this service pass it has used 99.9 percent of its allotted bandwidth, the WRR algorithm will still send out the next packet

in the queue as the queue's allotted bandwidth allocation has not yet been used up. When it sends the next packet, it will exceed the amount of

bandwidth that was configured for this queue. Statically over time, queue 2 may end up using a lot more bandwidth than it was initially configured

for when using WRR. (See Figure 12.)

Figure 12. DWRR

DWRR prevents this problem from occurring. If the queue uses more bandwidth than it was allotted, DWRR keeps a tally of the extra bandwidth

used on that pass. On the next pass through the queues, the DWRR algorithm will subtract the extra bandwidth used on the last pass for this turn.

This means statistically over a period of time that each queue will use bandwidth that is much closer to the configured amount for that queue.

Scheduling: SRR

SRR is a recent addition to the scheduling capabilities of the Cisco Catalyst 6500 Family. At the time of writing this paper, support for SRR is only

available on the uplink ports of the Cisco Catalyst 6500 Series Supervisor Engine 32. SRR is different from WRR in that the SRR algorithm

provides a way to shape outbound traffic to a stated rate. In some respects, it is similar to a policer except that traffic in excess of the rate will be

buffered rather than dropped as with a policer.

The shaper is implemented on a per-queue basis and has the effect of smoothing transient bursts of data that pass through that port. SRR also

modifies the way in which it schedules data when compared to the WRR algorithm. This can be seen in Figure 13, which shows a representation

of the packet scheduling order for both algorithms.

Important notices, privacy statements, and trademarks of Cisco Systems, Inc. can be found on cisco.com.

Page 11 of 18