Output rate-controlled generalized processor sharing: End-to-end delay bounds calculation

Recently multimedia applications such as music or video streaming, video teleconferencing and IP telephony have been flourishing in packet-switched networks. Applications that generate such real-time data can have very diverse quality-of-service (QoS) requirements. In order to guarantee diverse QoS requirements, the combined use of a packet scheduling algorithm based on Generalized Processor Sharing (GPS) and leaky bucket traffic regulator is the most successful QoS mechanism. GPS can provide a minimum guaranteed service rate for each session and tight delay bounds for leaky bucket constrained sessions. However, the delay bounds for leaky bucket constrained sessions under GPS are unnecessarily large because each session is served according to its associated constant weight until the session buffer is empty. In order to solve the problem, a scheduling policy called Output Rate-Controlled Generalized Processor Sharing (ORC-GPS) has been proposed in [10]. ORC-GPS is a rate-based scheduling like GPS, and controls the service rate in order to lower the delay bounds for leaky bucket constrained sessions. In this paper, we extend the delay bounds calculation for single-node networks introduced in [10] to multi-node networks and compare ORC-GPS with GPS in terms of end-to-end delay bounds through numerical experiments.