Thread Clustering and Selection Schemes for Medium-Grained Computation Migration

Process/thread migration is indispensable for resource sharing, cycle stealing and other modes of interaction. Computing tasks need to be scheduled frequently to adapt to the capacity variation of the system resources in distributed computing environments. Local parallel workload should be migrated to remote machines partially or completely. This paper analyzes three different kinds of clustering algorithms to group local threads. One or more thread groups are selected as medium-grained computations which are distributed across multiple machines. Performance analyses and experimental results are provided to demonstrate their effectiveness.