ABSTRACT
In a many-core HPC data centers, jobs arrive at different moments of time and they need to be serviced by allocating on the available system cores at run-time. This chapter provides a brief overview of the platform and workload model along with the problem formulation. It proposes value and energy optimizing resource allocation approaches for HPC data centers. The chapter shows that the approaches combine identification of efficient allocation and appropriate voltage/frequency levels to jointly optimize value and energy consumption. Whilst existing approaches focus on methods like server consolidation and DVFS, they do not consider jobs containing dependent tasks. The chapter also shows that the proposed approach is able to significantly reduce energy consumption and improve value while applying DVFS for jobs containing dependent tasks.
