ILP models for the allocation of recurrent workloads upon heterogeneous multiprocessors

Abstract

The problem of partitioning systems of independent constrained-deadline sporadic tasks upon heterogeneous multiprocessor platforms is considered. Several different integer linear program (ILP) formulations of this problem, offering different tradeoffs between effectiveness (as quantified by speedup bound) and running time efficiency, are presented. One of the formulations is leveraged to improve the best speedup guarantee known for a polynomial-time partitioning algorithm, from 12.9 to 7.83. Extensive computational results on synthetically generated instances are also provided to establish the effectiveness of the ILP formulations. Keywords task partitioning • sporadic tasks • unrelated machines • speedup bound • ILP rounding 1 Introduction Heterogeneous multicore CPUs-CPUs in which the processing elements differ from one another with respect to functionality or processing speed-are currently widely available and increasingly becoming the

Bounding the execution time of parallel applications on unrelated multiprocessors

“…Next, Baruah et al. ( 2019 ) with the ILP approach for constrained deadlines, independent tasks, unrelated multiprocessors, and partitioned scheduling, a speed-up bound of 7.83 is achieved. Our approach considers a more general application model which can exploit the parallelism that exists in the applications.…”

“…2015 ; Andersson and Raravi 2016 ; Baruah et al. 2019 ) with no dependencies and typed DAGs (Yang et al. 2016 ; Han et al.…”

Workload assignment for global real-time scheduling on unrelated clustered platforms

“…Partitioned scheduling on heterogeneous platforms is a NP-hard problem and has been studied in several works [10,38,13]. Global scheduling on heterogeneous platforms, also known as unrelated multiprocessor platforms, was initiated by the seminal paper [9].…”

Posicionamento quase ideal de vBBU e atribuição de comprimento de onda em Cloud Fog RAN

“…A literatura aborda diversos trabalhos que tratam do tempo de execuc ¸ão e da escalabilidade da PLI utilizando a relaxac ¸ão linear. As soluc ¸ões propostas aplicam diversos tipos de abordagens de relaxac ¸ão para obter valores próximos ao ótimo em uma infinidade de aplicac ¸ão de redes e de logística [Noor-E-Alam and Doucette 2012, Zaky Kasem et al 2012, Baruah et al 2019. Essas soluc ¸ões mostram a busca por alternativas ou outras abordagens para resolver o problema de escalabilidade do PLI sem grande disparidade em comparac ¸ão ao resultado ótimo.…”