resource allocation and scheduling based on network utilization in packet networks

Abstract

The provision of quality-of-service (QoS) is an important and challenging topic in the design of recent packet networks. To achieve this goal, an effective resource allocation and scheduling scheme is very crucial. Moreover, the scheduling scheme needs to maximize network resource utilization because network resources are usually limited, especially for a wireless environment.

First, an optimal bandwidth and buffer allocation scheme is proposed for a switching node in an environment of heterogeneous traffic with multiple QoS. The proposed scheme minimizes the total required bandwidth for traffic classes to satisfy their packet loss rate (PLR) and delay requirements. A constrained optimization problem is formulated in a general form regardless of bandwidth allocation methods so that it can be applied to many bandwidth allocation methods already proposed in literature. Simulation results show that for a given set of flows, the proposed scheme requires smaller bandwidth than the conventional methods, and it also guarantees both the PLR and delay requirements.

Second, a new scheduling scheme, called earliest-virtual-deadline-first (EVDF), is proposed to improve resource utilization in a high-speed packet network. The proposed EVDF scheduling is a generalized form of the earliest-deadline-first (EDF) scheduling, using the concept of virtual delay bounds. It gives us the flexibility of controlling the delay violation probabilities of traffic flows. As a result, the EVDF scheduler can provide the flows with the QoS guarantees that cannot be supported by the EDF scheduler. An iterative algorithm is developed to obtain the virtual delay bounds that achieve ideal service fairness while satisfying the QoS requirements of the flows. The EVDF scheduler utilizes network resources more efficiently than the EDF scheduler.

Finally, two downlink schedulers are proposed for cellular FDMA/TDMA wireless networks. The proposed throughput-maximizing (TM) scheduler focuses on maxim...