• home
• current students
• graduate
• events

Presenter: Mohamed Fayed
Supervisor:

Date: Mon, December 9, 2002
Time: 13:30:00 - 14:30:00
Place: EOW 430

ABSTRACT

Abstract:

A hierarchical parallel scheduling (HPS) algorithm is studied. This algorithm has a high level of simplicity which makes it more practical for implementation in packet switched networks. The simplicity of the HPS comes from the formulation of scheduling tasks as a linear programming problem that is solved using a simple greedy algorithm. The proposed scheduling algorithm is capable of independently controlling the bandwidth, delay, and delay jitter allocations for different service classes. Complete control of bandwidth-critical, delay-critical, and jitter-critical traffics is assured through the use of two optimization techniques based on the transportation problem and the 0-1 knapsack problem.

We present the theory of operation, design, analysis, and performance evaluation of the HPS scheduler. An OPNET network model is built to verify the functionality and performance of our proposal. OPNET simulations enabled accurate probing in the system operation.

HPS is analyzed to evaluate its packet delay, packet loss probability, and packet transfer efficiency. Its performance is then compared to three other scheduling algorithms: static periority (SP),first-come first-serve (FCFS), and weighted round robin (WRR). Performance parameters are measured while all sessions are transmitting according to precalculated shares and when one of the sessions is misbehaving. Shares are calculated at the start of each frame depending on the type of the session, part of sending the data, and the amount of the backlogged data. The performance analysis shows that the HPS scheduling algorithm satisfies a high level of QoS that suffices for real time applications.

Free and open to the public.

For further information please contact:
Mohamed Fayed (mfayed@engr.uvic.ca)