HSMDA QoS Model
758 Quality of Service Guide
• Hardware implemented provider style port based scheduling
• Elimination of ingress dual pass queuing at the ingress hardware
• Egress intermediate destination (such as DSLAM) shaping using secondary shapers
• Expanded counters at ingress and egress
• CIR Bypass for color aware policing
• Ingress queue policing mode
• RED queue congestion control
• Egress dot1p remarking per packet based on egress queue scheduling rate
• Per queue packet byte offset for queue stats, queue PIR, queue CIR and queue group
PIR accounting
Queue Scaling
The HSMDA supports 160,000 queues in the ingress and egress directions. Each queue
supports two leaky buckets that perform a PIR shaping function and a CIR marking function
based on scheduled rate out of the queue. Each queue also supports two RED slopes that may
be used for managing queue congestion.
The queues are grouped into sets of eight queues each. 3840 queues are reserved for the
system. A set of eight queues is called a queue group and is internally identified by a queue-
group-id. Each queue within a group is numbered from 1 to 8 represented internally as the
queue-id. The queue-id has an implicit scheduling class association based on the number of
the queue (for example, queue 3 is a member of scheduling class 3). Queue groups are
dynamically mapped to egress ports on an as needed basis.
Individual queues are also mapped to 1 of 64 secondary shapers, each representing an egress
intermediate destination (such as a DSLAM). One is pre-allocated per egress port, so only 54
are user-definable in 10x1G HSMDAs and 63 are user-definable in 1x10G HSMDAs.
Port-Based Scheduling
Forwarding for each egress port on the HSMDA is managed by a port-based scheduler. Each
port-based scheduler maintains a maximum of eight strict forwarding levels (Figure 41).
Strict level 8 is the highest priority while strict level 1 is the lowest. There are also eight
scheduling classes that contain each of the queues assigned to the port scheduler. A queues
membership in a scheduler class is controlled by the queue’s identifier. For example, all
queues with a queue-id equal to 1 are in scheduler class 1 while all queues with queue-id equal
to 2 are in scheduler class 2. By default, each scheduler class is directly mapped to its
corresponding strict scheduling level.
To Next Page
Loading...
