Power Aware Task Matching and Migration in Heterogeneous Processing Environments

摘要

In previous task matching research the power agnostic Min-Min algorithms has been found to be an effective matching algorithm and has been used as a benchmark in numerous studies. Our study has enhanced these algorithms with the addition of the energy delay product (EDP) resulting in an improved Min-Min that takes both power and execution time into account. In addition, the study incorporates the classical Min-Min and an improved Min-Min with EDP as ways to migrate preempted task between system cores. Our results show that it is possible to obtain significant improvements in scheduling related power consumption with acceptable performance degradation.