Workload clustering for increasing energy savings on embedded MPSOCS
Narayanan, S. H. K.
Zomaya, A. Y.
Lee, Y. C.
549 - 565
Item Usage Stats
MetadataShow full item record
Energy‐efficient distributed computing systems
Our goal in this chapter is to explore a workload (job) clustering scheme that combines voltage scaling with processor shutdown.1 The uniqueness of the proposed unified approach is that it maximizes the opportunities for processor shutdown by carefully assigning workload to processors. It achieves this by clustering the original workload of processors in as few processors as possible. In this chapter, we discuss the technical details of this approach to energy saving in embedded MPSoCs. The proposed approach is based on ILP (integer linear programing); that is, it determines the optimal workload clustering across the processors by formulating the problem using ILP and solving it using a linear solver. In order to check whether this approach brings any energy benefits over pure voltage scaling, pure processor shutdown, or a simple unified scheme, we implemented four different approaches within our linear solver and tested them using a set of eight array/loop-intensive embedded applications. Our simulationbased analysis reveals that the proposed ILP-based approach (i) is very effective in reducing the energy consumptions of the applications tested and (ii) generates much better energy savings than all the alternate schemes tested (including one that combines voltage/frequency scaling and processor shutdown).
Published Version (Please cite this version)https://doi.org/10.1002/9781118342015.ch19