C. Maple, Y. Wang, and J. Zhang (UK)
Load-balancing, job-scheduling, algorithms, heterogeneous networks.
Recent years have seen a large increase in the use of distributed systems to solve computationally expensive problems. More institutions than ever are now utilising the previously idle time of computer laboratories. This raises the profile of two key issues: the desirability that the normal daily operation of the system is unaffected and that the new use of the system is efficient. The latter is an increasingly complex problem since many laboratories are heterogeneous networks. Maple and Vince, [7], presented a framework and implementation of a distributed network that boots off a floppy disk and does not use the hard disks of the machines in the system. The scheduling algorithm implemented involved assigning jobs to processors based upon the assumption of a uniform workload for each job. This paper addresses the problem when the cost of computation is non constant
Important Links:
Go Back
