Service classes Sample Clauses

Service classes. In the simulations that follow, we assume 3 service classes set up to provide Gold, Silver and Bronze services. Gold Service is targeted for real time applications like voice which assume little queuing delay. It is constructed based on limiting the amount of Gold traffic accepted into the network so that the sum of all Gold flows rates (assuming CBR flows) is less than the guaranteed service rate for the Gold aggregate. In DiffServ for example, the Gold service would be EF which is usually setup to emulate circuit switching. Figure 10 shows a typical gold queue size along with the calculated RED average queue size versus time. Figure 10 Typical Gold Queue Size‌ Silver service is targeted for non real time applications that are tolerant to loss but require better than best effort service (e.g. video). It is constructed based on an assured service rate for the Silver aggregate and a buffer management algorithm, RED, which is set up to drop packets whenever the average queue size exceeds a pre-configured threshold. Silver traffic assumes some queuing but smaller queuing than the bronze traffic (better than best effort). Figure 11 shows a typical Silver queue size along with the calculated RED average queue size (assumes a RED max threshold at 20 packets) versus time when all Silver flows are using the maximum allowed rates (worst case condition). Typically the Silver traffic is over-booked (Silver service could be setup so that the sum of all silver flows exceeds 100% of the reserved link capacity ratio). New clients are allowed into the network as long as the aggregate service provided is still within the advertised QoS of that class. Typically, providers use a CAC algorithm in order to maximize the benefit (i.e. dollars) without affecting the “promised” service for the aggregate i.e. as long as the new incoming flow does not affect the “promised” service offered by the Silver class, the flow is accepted even if the sum of the rates of all flows exceed the reserved rate for the aggregate (it builds on the dynamic behavior of the flows and some feedback from the network on whether the new flow would badly affect the Silver Service if accepted). Figure 11 Typical Silver Queue Size As for the Bronze service class, being a best effort service, it is usually overbooked with higher over-booking percentage than the Silver service (since there are no service guarantees associated with the best effort service class). A buffer management algorithm (e.g. RED) keeps the...