Seite - 180 - in Algorithms for Scheduling Problems

algorithms Article PHEFT:Pessimistic ImageProcessingWorkflow SchedulingforDSPClusters AlexanderYu.Drozdov1,AndreiTchernykh2,3,*,SergeyV.Novikov1,VictorE.Vladislavlev1 andRaulRivera-Rodriguez2 1 MoscowInstituteofPhysicsandTechnology,Moscow141701,Russia; alexander.y.drozdov@gmail.com(A.Y.D.); serg.v.novikov@gmail.com(S.V.N.); victor.vladislavlev@gmail.com(V.E.V.) 2 ComputerScienceDepartment,CICESEResearchCenter,22860Ensenada,BajaCalifornia,Mexico; rrivera@cicese.mx 3 SchoolofElectricalEngineeringandComputerScience,SouthUralStateUniversity, Chelyabinsk454080,Russia * Correspondence: chernykh@cicese.mxorchernykhan@susu.ru;Tel.:+52-646-178-6994 Received: 27February2018;Accepted: 9April2018;Published: 22May2018 Abstract:Weaddress imageprocessingworkflowschedulingproblemsonamulticoredigital signal processor cluster. We present an experimental study of scheduling strategies that include task labeling,prioritization, resourceselection,anddigital signalprocessorscheduling.Weapply these strategies in the context of executing the Ligo andMontage applications. To provide effective guidance in choosingagoodstrategy,wepresent a joint analysis of three conflictinggoalsbased onperformance degradation. A case study is given, and experimental results demonstrate that apessimistic schedulingapproachprovides thebestoptimizationcriteria trade-offs. ThePessimistic HeterogeneousEarliestFinishTimeschedulingalgorithmperformswell indifferentscenarioswith avarietyofworkloadsandclusterconfigurations. Keywords: DSP microprocessor; multicore; multiprocessors; scheduling; workflow; resource management; joballocation 1. Introduction In thispaper,weaddress themulti-criteriaanalysisof imageprocessingwithcommunication workflow scheduling algorithms and study the applicability of Digital Signal Processor (DSP) clusterarchitectures. The problem of scheduling jobs with precedence constraints is a fundamental problem in schedulingtheory[1,2]. Itarises inmanyindustrialandscientificapplications,particularly, in image and signal processing, and has been extensively studied. It has been shown to beNP-hard and includessolvingacomplex taskallocationproblemthatdependsnotonlyonworkflowpropertiesand constraints,butalsoonthenatureof the infrastructure. In thispaper,weconsideraDSPcompatiblewithTigerSHARCTS201S[3,4]. Thisprocessorwas designedinresponse to thegrowingdemandsof industrial signalprocessingsystemsfor real-time processingof real-worlddata,performingthehigh-speednumericcalculationsnecessary toenable a broad range of applications. It is optimized for both floating point andfixedpoint operations. It providesultra-highperformance; static superscalarprocessingoptimized formemory-intensive digitalsignalprocessingalgorithmsfromfully implemented5Gstations; three-dimensionalultrasound scanners and other medical imaging systems; radio and sonar; industrial measurement; and control systems. Algorithms 2018,11, 76;doi:10.3390/a11050076 www.mdpi.com/journal/algorithms180