Prioritized Task Offloading Mechanism in Cloud-Fog Computing Using Improved Asynchronous Advantage Actor Critic Algorithm
Open Access
- 17 September 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 136628-136656
Abstract
Task scheduling is a crucial aspect in cloud computing paradigm as tasks with different processing capacities arise from heterogeneous resources and scheduling these different tasks to precise virtual machines (VM) in cloud datacenters is a crucial job for the Cloud Service Provider (CSP) as it is an NP-hard problem. Ineffective scheduling of tasks in cloud paradigm leads to increase in execution time in turn it effects quality of service which is a huge challenge in cloud paradigm. Scheduling tasks to VMs in datacenters which are located at far distances also plays a major part in quality of service of CSP especially in the case of time sensitive tasks. Many of the earlier authors proposed task offloading mechanisms using metaheuristic techniques and machine learning approaches but still there is a chance to improve offloading process by prioritizing tasks based on their runtime capacities, sizes onto fog nodes and schedule them onto precise VMs while addressing important concerns in cloud/fog computing paradigm such as makespan, consumption of energy. In this research, we proposed a task offloading mechanism (PTOMCFIA3C) which considers priorities of tasks and based on the priorities induced into the scheduling offloading mechanism will decide whether to offload tasks to fog/cloud nodes. PTOMCFIA3C modeled using improved Asynchronous Advantage Actor Critic algorithm a deep reinforcement learning technique which is improved by inducing RCNN which can accelerate its learning capability and optimizes scheduling process by offloading tasks to Fog/cloud nodes. Extensive simulations are conducted using simpy by taking the real-time computing worklogs. Simulations conducted in two phases. In the first phase, parameters evaluated with fixed nodes while in second phase parameters evaluated with variable nodes. Proposed PTOMCFIA3C evaluated over existing baseline approaches RATS-HM, MOABCQ, FOG-AMOSM and our proposed task offloading mechanism PTOMCFIA3C generated optimized schedules while improving makespan by 81.09%, consumption of energy by 79.33% significantly over compared approaches.Keywords
Funding Information
- King Saud University, Riyadh, Saudi Arabia (RSPD2024R576)
This publication has 60 references indexed in Scilit:
- A heuristic multi-objective task scheduling framework for container-based clouds via actor-critic reinforcement learningNeural Computing and Applications, 2023
- Task scheduling in edge-fog-cloud architecture: a multi-objective load balancing approach using reinforcement learning algorithmComputing, 2023
- Deep reinforcement learning for fault-tolerant workflow scheduling in cloud environmentApplied Intelligence, 2022
- Optimized task scheduling in cloud computing using improved multi-verse optimizerCluster Computing, 2022
- A gradient-based optimization approach for task scheduling problem in cloud computingCluster Computing, 2022
- Multi-objective workflow scheduling in cloud computing: trade-off between makespan and costCluster Computing, 2021
- Multi-objective hybrid genetic algorithm for task scheduling problem in cloud computingNeural Computing and Applications, 2021
- An improved Henry gas solubility optimization algorithm for task scheduling in cloud computingArtificial Intelligence Review, 2020
- QL-HEFT: a novel machine learning scheduling scheme base on cloud computing environmentNeural Computing and Applications, 2019
- CloudSim: a toolkit for modeling and simulation of cloud computing environments and evaluation of resource provisioning algorithmsSoftware: Practice and Experience, 2010