Streaming Workload Characterization and Multi-level Client Clustering

This project is part of a larger research project which investigates the applicability of the peer-to-peer (P2P) computing paradigm in designing large-scale content distribution systems. To develop an efficient content distribution system, it is essential to understand the workload that will be distributed, the behavior of content consumers and the environment in which the system will operate. In this project, the intern seeks to understand and analytically model the characteristics of the streaming workload imposed on servers of large content providers and its impact on the underlying network. Analytic models describing the distribution of objects (sizes, relative popularity, and types), clients (geographical distribution, arrival rate, request rate, session duration, and capacity), and network load (traffic per client, per network, and per ISP) will be developed. Based on the analysis of the workload, the intern will develop multi-level clustering schemes to facilitate content sharing and optimize client-perceived quality.

Osama Saleh
Faculty Supervisor: 
Dr. Mohamed Hefeeda
British Columbia