A scalable QoS guaranteeing scheme for scheduling IP traffic

Abstract

The continuing increase in network speed demands a high speed performance of packet processing functions such as queuing and scheduling. In addition, the co-existence of different traffic with quality of service (QoS) requirements limits the scalability of packet processing functions. Two basic queuing architectures implemented in routers are Output Queuing (OQ) and Input Queuing (IQ). OQ offers optimal throughput and guarantees QoS but it is not scalable. Even though IQ has been identified as the most scalable and the simplest, the optimum throughput achievable is 58.6% due to the Head Of Line (HOL) blocking effect. The Virtual Output Queuing (VOQ), a proffered solution to the HOL blocking achieves a 100% throughput with an effective scheduling algorithm. Existing schedulers for the VOQ are deterministic algorithms which either provide deterministic QoS guarantee or do not guarantee QoS. However, the algorithms that guarantee QoS have high time complexity. This paper proposes a probabilistic scheduling algorithm known as Iterative Probabilistic Scheduling (IPS) for the VOQ. The queuing architecture presented is a multi-stage queuing and scheduling (MQAS) in which VOQ is implemented at the input port and OQ at the output port of the router.