Publication Details
Abstract
cloud computing scheduling is a proven NP-hard optimisation problem with the challenge that is created by time-variant and varying nature of cloud workload. Even though there are plenty of metaheuristic algorithms that have been created, most of the algorithms fail while obtaining an appropriate balance between local node exploitation and global node exploration thus creating non-optimal scheduling performance. The paper proposes "HybridPSOGWO" as a new hybrid scheduling algorithm that integrates the exploratory nature of the Grey Wolf Optimizer (GWO) and the exploitive capability of Particle Swarm Optimization (PSO) in an effort to solve these challenges. Such integration is created in an effort optimise the convergence rate, adaptability, and efficiency of the schedulesIt is rigorously tested against several advanced algorithms, such as "MPSOSA", "RL-GWO", "CCGP", and "HybridPSOMinMin", based on key performance indicators such as makespan, throughput, and load balancing. Testing was performed through executing an enormous number of simulations on the simulator CloudSim Plus using workload traces quantified in the real world. Experimental results show that "HybridPSOGWO" is the one that outperforms comprehensively the other competing algorithms and report makespan gain up to 15 percent and corresponding throughput gain up to 10 percent at least and as well provide a better even distribution mechanism of the tasks between the virtual machines. Moreover, it is found that the algorithm introduced is fast convergent and is stable and thus indicating an algorithm robustness as well as an optimal algorithm selection in adaptive task scheduling in large-scale cloud computing system.Cloud computing has become one of the revolutionary paradigms in the global information technology, whereby multiple computing resources including servers, storage, applications, and services, can be made available to an end user through on-demand and a shared pool of configurable computing resources . It allows service providers to dynamically allocate resources so as to fulfill the requests of the user further enhancing a broad range of applications in many industries such as healthcare, finance, education and entertainment markets. Namely, the effective allocation of computational workloads to virtual machines (VMs) is one of the strategic difficulties in the sphere of cloud computing and has a direct impact on the performance of the system, resources, power consumption and customer satisfaction .Task scheduling on cloud systems entails the allocation of a set of independent or dependent tasks to a set of heterogeneous virtual machines (VMs) with the objective of optimizing one or more performance objectives, e.g. makespan (total completion time), cost, throughput, load balancing and energy efficiency . The problem of task scheduling is however ranked as an NP-hard optimization problem due to combinatorial nature of the scheduling problem as well as the unpredictable variation of the intensity of work . This means that it is infeasible to find a solution that is optimal within a reasonable amount of time, particularly in large-scale, real-time cloud systems and hence it requires the implementation of heuristic and metaheuristic solutions to gain near-optimal solutions in an efficient manner.