Cisco Cisco MDS 9500 Series Supervisor-2 Module Libro blanco
© 2006 Cisco Systems, Inc. All rights reserved.
Important notices, privacy statements, and trademarks of Cisco Systems, Inc. can be found on cisco.com.
Page 7 of 11
Table 3.
Mapping of Frames to Virtual Output Queues
Service Level
DSCP
PHB (per FC-FS-2 Standard)
Mapped to Queue
Absolute
46
EF (expedite forwarding)
Priority queue
High
10, 12, 14
AF1 (assured forwarding 1)
DWRR1 queue
Medium
18, 20, 22
AF2 (assured forwarding 2)
DWRR2 queue
26, 28, 30
AF3 (assured forwarding 3)
Low
34, 36, 38
AF4 (assured forwarding 4)
Default
All others
All others
DWRR3 queue
Four virtual output queues are available for each output port, corresponding to four QoS levels per port: one priority queue and three queues
serviced using DWRR scheduling. Frames queued in the priority queue will be scheduled for transmission prior to any frames queued in the DWRR
queues. Frames queued in the three DWRR queues are scheduled using whatever relative weightings are configured. The default relative weightings
are 33:33:33; however, these are configurable items and can be set to any ratio at all.
The amount of ingress buffering per port is also a configurable option, with options varying for buffering for 12 frames up to 4096 frames.
In most cases, there is no need to change ingress buffer levels beyond the default.
OTHER TRAFFIC ENGINEERING FEATURES
VOQ
All Cisco MDS 9000 Family switches use the VOQ queuing technique (Figure 6). This is a standard feature that is always enabled.
VOQ helps mitigates a phenomenon called “head-of-line blocking”. Blocking occurs when congestion on an output port blocks frames from
being sent. Fibre Channel switches provide a guarantee that no frames will ever be dropped, so when an output port is congested, switches have
to block until such time as the frame can be transmitted. Head-of-line blocking occurs when the frame at the head of the queue cannot be sent
due to congestion at its output port. In this situation, frames behind this frame are “blocked” from being sent to their destination, even though
their respective output ports are not congested.
VOQ prevents head-of-line blocking by utilizing multiple virtual output queues. Individual virtual output queues might block, but traffic queued
for different (nonblocked) destinations can continue to flow without getting delayed behind frames waiting for the blocking to clear on a congested
output port.