Browse Theses

Fiber optics technology increases network bandwidth enormously, and new applications require a variety of quality of services. Real-time applications require performance guarantees on throughput, delay bounds, and loss rate. A new traffic control discipline is needed: it should generate proper packet orderings for each flow and schedule packet transmissions to meet both minimum and maximum delay bounds. Real-time flows must be protected from flows of misbehaved users and from network load fluctuations caused by packet clustering. Current network traffic mechanisms do not provide both protections.

An integrated network traffic control system is designed for high-speed packet-switched wide-area networks. The traffic control is a rate-based control system with explicit resource reservation under a connection-oriented network architecture. It provides a guaranteed reserved service to real-time flows and a best-effort service to non-real-time flows. The system consists of four mechanisms: flow set-up, packet admission control, packet scheduling, and flow monitoring. The inter-relationship between the mechanisms is studied within a basic framework for each mechanism.

A new packet scheduling mechanism, the Real-Time VirtualClock (RTVC) algorithm, is developed. The RTVC algorithm provides a guaranteed real-time service meeting maximum and minimum delay bounds and supports bursty flows based on peak rates without causing serious disturbances to continuous flows. Different classes of flows are statistically multiplexed to maintain high utilization of network resources. To prevent unpredictable network performance, misbehaving flows are isolated, and network load fluctuations caused by the packet clustering are eliminated.

Simulation is used to validate the integrated traffic control system. Packet scheduling with the RTVC algorithm is evaluated, and comparisons with current packet scheduling mechanisms are made. The simulation results confirm that the traffic control system provides stable network performance under a variety of traffic conditions.