OPTIMIZING SERVER-SIDE DYNAMIC LOAD BALANCING IN SDN USING NOVEL ALGORITHMS BASED ON CPU, RAM, AND CONNECTION METRICS

Main Article Content

Maghrib Abidalreda
Maky Alrammahi
Mohanad Yahya Al-hamami
Ali Mohammed Taher

Abstract

A software-defined network (SDN) is a network architecture that decouples the data forwarding plane, which forwards data, from the controller plane, which conducts network management and decisions. SDN has exhibited considerable advantages over traditional non-SDNs. However, traffic allocation in SDN impacts efficiency and introduces various challenges. For instance, an unbalanced load distribution in the SDN significantly impacts the performance of the network. This study aims to optimise the utilisation of cluster resources by addressing the load-balancing issue on the server side. It attempts to improve the performance of servers with varying processing capabilities while improving the quality of service in terms of delay, waiting time and service time. Four dynamic load balancing techniques are proposed: least random access memory (LRAM), least control process unit (LCPU), LRAM with LCPU (LCPURAM) and least connection (LC) with LCPURAM (LCLCPURAM). The default dynamic load balancing algorithm based on LC is compared with these strategies. The evaluation employs a Mininet emulator and an Open-flow switch with a POX controller. The evaluation results indicate that the proposed LCLCPURAM algorithm significantly enhanced the delay metric by 15.5%. In addition, it achieved the highest improvement in the waiting time metric, with a 19.2% enhancement. In the meantime, the LC algorithm exhibited the greatest improvement in the service time metric, with a 27% enhancement.

Article Details

Section

Articles

References