Scientific Computing in Object-Oriented Parallel Environments

Volume 1343 of the series Lecture Notes in Computer Science pp 283-290


A general resource reservation framework for scientific computing

  • Ravi RamamoorthiAffiliated withCalifornia Institute of Technology
  • , Adam RifkinAffiliated withCalifornia Institute of Technology
  • , Boris DimitrovAffiliated withCalifornia Institute of Technology
  • , K. Mani ChandyAffiliated withCalifornia Institute of Technology

* Final gross prices may vary according to local VAT.

Get Access


We describe three contributions for distributed resource allocation in scientific applications. First, we present an abstract model in which different resources are represented as tokens of different colors; processes acquire resources by acquiring these tokens. Second, we present distributed scheduling algorithms that allow multiple resource managers to determine custom policies to control allocation of the tokens representing their particular resources. These algorithms allow multiple resource managers, each with its own resource management policy, to collaborate in providing resources for the whole system. Third, we present an implementation of a distributed resource scheduling algorithm framework using our abstract model. This implementation uses Infospheres, which are Internet communication packages written in Java, and shows the benefits of distributing the task of resource allocation to multiple resource managers.