Article Preview
Top1. Introduction
Cloud computing is a shared pool of geographically distributed computational resources. Its primary aim is to provide reliability and scalability computing environment for cloud application and services.
The cloud computing technology provides the hardware at a datacentre. The primary objective includes efficiency improvement. Cloud service providers provide resources, which are allocated to the client, dynamically when required under overloaded condition. The user’s requests and resource requirement are the deriving parameters which define the performance of various services provided by the cloud. The significant concerns are related to the efficiency of the Infrastructure as a service (IaaS). In the scalable cloud aura, resource provisioning is a way of distributing the data centre configurations among the Software as a service (SaaS) modeller. The resource mapping (allocation of cloudlets on a virtual machine) achieves the performance metrics of the SaaS modeller. The performance metrics cover the requirement of developers, client, and datacentre owner. The data centre owner assures the quality of service using service level agreement.
Primarily, scheduling focuses on performance improvement of hardware configuration at a datacentre. Scheduling is essential for various category of users across the globe Achar et al., (2012). The challenging concerns are monitored by efficient provisioning techniques (Manvi & Shyam, 2014). The review work focused on various features of cloud-like scalability and mobility in a cloud (Dinh et al., 2013), resource allocation (Ma et al., 2014; Madni et al., 2016), load balancing (Gabi et al., 2015; Kansal & Chana, 2012), task scheduling (Kalra & Singh, 2015; Tsai & Rodrigues, 2013; Zhang & Su, 2014) and power efficiency in a cloud Gabi et al., (2015). The various cloud task scheduling schemes are being proposed of cloud services which cover various issues in the cloud. Some of the performance metrics like execution time, reliability and failure probability are focused in this work. The existing proposed work may not be reliable and fault aware.
In general, tasks allocation in the distributed surrounding is an NP-hard problem (Yu & Buyya, 2006). An optimal solution for task scheduling problem using GA is presented by many researchers in the past, but it is not the best method, as illustrates by Lie et al. Zhang et al., (2008). Kumar and Verma (Kumar & Verma, 2012) have proposed a mixture of Max-min and Min-Min policies in GA. The proposed algorithm is faster than GA but time-consuming. Guo et al. Guo et al., (2012) have suggested a PSO method to reduce execution and transfer time. This proposed strategy is more agile than the L-PSO and M-PSO algorithm on a massive scale but stuck in the local optimal solution. Pandey et al. Pandey et al., (2010) have proposed a heuristics approach based on PSO algorithm. The algorithm reduces the execution time and provides three times better cost compared to BRS but stuck in the local optimal solution.
Based on the research found so far, task scheduling in cloud computing is an open issue for all. So, to overcome the existing solutions given by many researchers using GA (Kansal & Chana, 2012)[6]Yiqiu et al., (2019), a new hybrid GA-ANN (Genetic Algorithm-Artificial neural network) scheduling algorithm is proposed. The proposed algorithm combines the advantages of both the methods to minimize the execution time and increase the utilization percentage of computing resources. To validate the results obtained from GA-ANN model, the algorithm is evaluated against GA-Fault (Rekha & Dakshayini, 2019).