ABSTRACT

The amount of information transmitted is continually risen while modern communication methods have been developed. To successfully meet that rising computational demands, the volume of datacentres networks (DCNs), which are materials built up of servers connected by sustainable switches, rapidly grown across the globe. Classical switches are ineffective in meeting the demands of DCNs. Over the latest days, the software-defined network (SDN) was already considered as a novel networking standard for managing DCNs, controlling network switches, and deploying new network services. One major issue with DCNs, therefore, was balancing the workload among computers. Machine learning (ML) technologies might be used to handle data transfer requirements as one possible solution to this issue. Deep learning (DL) is a new effective ML approach that generates predictions, classifications, and choices using vast quantities of data. Despite deep learning (DL) is gaining popularity in such a variety of disciplines, has few advantages in networks. Inside this article, a DL method for load-balancing SDNbased DCNs is presented. The different loading levels across connections were used to educate the DL networks. The overall reaction speed of a DL approach for load balancing is contrasted to that of several ML techniques, including an ANN, SVM, and logic regression method. The findings show perhaps ANN & DL methods have faster reaction times than support vector machines & logistic regression methods. Furthermore, DL efficiency is superior to ANN efficiency. As a consequence, DL seems to be an excellent load balancing solution. Keywords: Machine learning, Load balance, SDN, Deep learning, Switches; Environment
Publication date: 01/11/2021

https://ijbpas.com/pdf/2021/November/MS_IJBPAS_2021_NOV_SPCL1079.pdf

Download PDF
https://doi.org/10.31032/IJBPAS/2021/10.11.1079