Developing a New Prediction Method for Grid Environments

Please enable javascript in your browser.

Naddaf Sichany, Babak | 2009

654 Viewed

Type of Document: M.Sc. Thesis
Language: English
Document No: 39295 (52)
University: Sharif University of Technology, International Campus, Kish Island
Department: Science and Engineering
Advisor(s): Habibi, Jafar
Abstract:
In this project we have worked on the new architecture of the auction based resource scheduling, from the bidders point of view. The performance of different bidding strategies for the resources which participate in reverse auction system has been investigated; our main parameter for evaluating different bidding strategies is the amount of the profit gained by resources which follow such strategies. The main historical bidding strategies are created based on two famous predictors ES and AUTO-REGRESSION. In addition a game theory approach has been proposed. We have shown that our bidding algorithm (based on the sequential game model) reaches to an equilibrium point if all the bidders follow our bidding strategy. The performance evaluation of different bidding strategies have been proposed by comparing two main performance parameters (average turn around, job success rate). After that the comparison between reverse auction model and direct auction scheme has been proposed. Finally the convergence property of the greedily bidding strategy has been investigated under different conditions
Keywords:
Grid Computing ; Exponential Smoothing ; Prediction ; Online Auction ; Scheduling ; Vector Autoregressive Model ; Convergence Property

No TOC