2021/12/14 by Shreshth Tuli, Tuli, Shreshth, Giuliano Casale +3
Computer Science · #Age of Information Optimization #Artificial Intelligence (cs.AI) #Cloud Computing and Resource Management #Distributed #FOS: Computer and information sciences #IoT and Edge/Fog Computing #Parallel #Performance (cs.PF) #and Cluster Computing (cs.DC)
paper · pdf · doi:10.48550/arxiv.2112.07269
openalex publication_date 2021/12/14 · openalex created_date 2022/11/21 · openalex updated_date 2026/07/28
Workflow scheduling is a long-studied problem in parallel and distributed\ncomputing (PDC), aiming to efficiently utilize compute resources to meet user's\nservice requirements. Recently proposed scheduling methods leverage the low\nresponse times of edge computing platforms to optimize application Quality of\nService (QoS). However, scheduling workflow applications in mobile edge-cloud\nsystems is challenging due to computational heterogeneity, changing latencies\nof mobile devices and the volatile nature of workload resource requirements. To\novercome these difficulties, it is essential, but at the same time challenging,\nto develop a long-sighted optimization scheme that efficiently models the QoS\nobjectives. In this work, we propose MCDS: Monte Carlo Learning using Deep\nSurrogate Models to efficiently schedule workflow applications in mobile\nedge-cloud computing systems. MCDS is an Artificial Intelligence (AI) based\nscheduling approach that uses a tree-based search strategy and a deep neural\nnetwork-based surrogate model to estimate the long-term QoS impact of immediate\nactions for robust optimization of scheduling decisions. Experiments on\nphysical and simulated edge-cloud testbeds show that MCDS can improve over the\nstate-of-the-art methods in terms of energy consumption, response time, SLA\nviolations and cost by at least 6.13, 4.56, 45.09 and 30.71 percent\nrespectively.\n