Dissertation Defense Details
Scheduling for Fast Turnaround in Peer-based Desktop Grid Systems
| Author: | Dayi Zhou |
|---|---|
| Date: | May 22, 2006 |
| Time: | 14:00 |
| Location: | 220 Deschutes |
| Committee: | Virginia Lo (Chair) Allen Malony Andrzej Proskurowski Patricia A. Gwartney Dr. Sharad Garg |
Abstract
Peer-based desktop grid systems represent a new research area stimulated by the recent success of Internet-based computing projects and rapid developments in peer-to-peer computing. A light-weight, open peer-based desktop grid system allows cycle donors with similar interests to self-organize into a cycle sharing community by joining an overlay network, similar to those used by peer-to-peer file sharing systems.
This talk addresses the development of scheduling strategies that achieve fast turnaround in open, dynamic, and large scale peer-based desktop grid. The challenges are two-fold: How does the scheduler quickly discover idle cycles in the absence of global information about host availability? And how can faster turnaround be achieved within the opportunistic scheduling environment?
I describe WaveGrid, a timezone-aware overlay network for fast resource discovery. Jobs ride the wave of available cycles by migrating from busy day-time hosts to hosts located in idle night-time zones around the globe. Simulations driven by a heterogeneous host model show that WaveGrid consistently outperforms other systems with respect to turnaround, stability and minimal impacts on hosts.