Scheduling is the process that determines which traffic is forwarded when multiple QoS components are competing for egress bandwidth. The ExtremeXOS software supports the following scheduling methods:
Strict priority queuing: All higher priority queues are serviced before lower priority queues. This ensures that high priority traffic receives access to available network bandwidth immediately, but can result in lower priority traffic being starved. As long as a queued packet remains in a higher-priority queue, any lower-priority queues are not serviced.
Weighted fair queuing: All queues are given access to a relative amount of bandwidth based on the weight assigned to the queue. When you configure a QoS profile with a weight of 4, that queue is serviced four times as frequently as a queue with a weight of 1. The hardware services higher-weighted queues more frequently, but lower-weighted queues continue to be serviced at all times. Initially, the weight for all queues is set to 1, which gives them equal weight. If all queues are set to 4, for example, all queues still have equal weight, but each queue is serviced for a longer period.
Round-robin priority: All queues are given access based on the configured priority level and a round-robin algorithm.
Scheduling takes place on the egress interface and includes consideration for the color-marking of egress frames and packets. Green-marked traffic has the highest priority and is forwarded based on the scheduling method. When multiple queues are competing for bandwidth, yellow-marked traffic might be dropped or remarked red. Red-marked traffic is dropped when no bandwidth is available. If yellow-marked traffic is forwarded to the egress port rate-shaping component, it can be dropped there if the egress port is congested.