| 摘要 |
The HPRR method uses "token bucket" rate classifiers to mark each individual packet as conforming or not conforming to a traffic specification for the flow. Flows are considered to be in a single service class. One such class is distinguished as a default "best effort" service class. Each service class is assigned a weight corresponding to its fraction of bandwidth granted to the class when all classes are active. The HPRR method allows a packet from a flow to be forwarded in one of two ways, either as part of its class's allocated bandwidth or as part of the "best effort" bandwidth. By always providing two paths for a flow to send its packets, a flow is always given its "fair share" of two different classes: its primary or configured class and the best effort class. An overbooked class will have each of its flows compete for the inadequate bandwidth allocated to the class, but because each flow can use the best effort bandwidth it also gets a fair share of that bandwidth to allow the classes bandwidth guarantees to be when other classes are inactive.
|
