Chapter 7 focused on queuing and scheduling concepts and how they operate. This chapter discusses more advanced scenarios involving overprovisioning and guaranteed rates that are used in large-scale queues to differentiate shaping rates. This chapter touches on the debate about the optimal size of the queues and different memory allocation techniques. Finally, the chapter details the famous (or infamous) Random Early Discard (RED) concept.
8.1 Hierarchical Scheduling
The techniques in Chapter 7 described single-level port scheduling. But modern QOS scenarios and implementations need to support multiple levels of scheduling and shaping. A situation when this is common is a subscriber scenario that provisions end users. Consider the topology in Figure 8.1.
The Broadband Service Router (BSR) in Figure 8.1 has a 10-Gbps Gigabit Ethernet link to the switch. The two connections between this switch and the two DSLAM switches are both Gigabit Ethernet links. The subscribers have traffic contracts with different services and rates. To be able to have a functional traffic provisioning model, rates to and from the subscribers need to accommodate service and contracts. Traffic from the subscribers must be shaped to align with the service contract, and traffic to the subscribers must be shaped according ...