Reputation-Based Resource Allocation in Market-Oriented Distributed Systems

The scale of the parallel and distributed systems (PDSs), such as grids and clouds, and the diversity of applications running on them put reliability a high priority performance metric. This paper presents a reputation-based resource allocation strategy for PDSs with a market model. Resource reputation is determined by availability and reliable execution. The market model helps in defining a trust interaction between provider and consumer that leverages dependable computing. We also have explicitly taken into account data staging and its delay when making the decisions. Results demonstrate that our approach significantly increases successful execution, while exploiting diversity in tasks and resources.