Adaptive Sabotage-Tolerant Scheduling for Peer-to-Peer Grids
Ana Cristina Oliveira, Livia M. R. Sampaio, Stênio Fernandes, Francisco Brasileiro · 2009
Computational grids are an infrastructure to aggregate computing power to support and improve performance of parallel applications. Some of them evolved in the sense of forming free-to-join communities over the Internet and became peer-to-peer (P2P) grids. One of the main problems associated with users freely joining and leaving grid communities is that cheating users may corrupt the final computation. Sabotage tolerance techniques, generally based on replication, tackle this problem by estimating the computation correctness. The use of credibility-based techniques in the task scheduling may promote high confidence levels for the computation results, whilst minimizing replication costs, when compared to the traditional voting technique. This work aims at evaluating the usage of scheduling heuristics that adapt themselves to the machines' confidence level in P2P grids. Three scheduling heuristics were evaluated. They present advantages and disadvantages, leading us to the conclusion that the performance of the scheduling heuristics is deeply influenced by the grid environment.