A Swarm Intelligent Algorithm for Optimizing Cloud Computing

Abstract

Cloud computing became existing in every domain of life, enhancing their functionality and adding new opportunities to it. It is the mechanism of moving the processing effort from the local devices to the data center facilities. Its exponential growth gained it a huge focus towards solving its challenges. Quality of service is one of the main challenges of cloud computing which are known as: 1) security and privacy, 2) portability, 3) reliability and availability and 4) quality of service (QoS). Quality of service is maintaining the proper management of resources in order to fulfill the Service Level Agreements (SLAs), which is the agreement between the cloud providers and the cloud users. Considering the massive demand to handling Cloud Computing challenges, research has been continuously performed in this area especially in load balancing. Load balancing is the process of distributing load over servers to keep the system steady without overloaded or under-loaded ones which maximize resource utilization. The load can be network load, memory or CPU loads. The Load balancing of any cloud system is dependent on its scheduler either task scheduler or resource scheduler. Research on it assists in improving one of these elements: 1) makespan, 2) response time, 3) migration time, 4) energy consumption, 5) throughput or 6) cost. It is branched to two types of work: Static Load Balancing (SLB) and Dynamic Load Balancing (DLB). Static Load Balancing runs from the start with prior knowledge of the system, while Dynamic Load Balancing depends on the progress of the system as it runs when overload state or imbalance occurs. It is considered a NP-hard problem so to solve it many research was done using heuristic and Meta heuristic Algorithms. This thesis proposes the use of selected swarm algorithms: Ant-Lion optimizer (ALO) and Grey wolf optimizer (GWO) in task scheduling of a cloud computing system as they are known for their high avoidance of local optima and high exploration of the search space in comparison to other intelligent algorithms. Upon experimenting with ALO against the traditional algorithm Round Robin (RR) in the small-scale simulation of ten task and five VMs, it outperforms RR in tasks executing time, but it was slow in scheduling because of its random walk of ants in each iteration. As such, this thesis proposes two modification to speed up the random walk of ALO: ALO2 and ALORW.