Adaptive thresholds for improved load balancing in mobile edge computing using K-means clustering

Abstract

Mobile edge computing (MEC) has emerged as a promising technology that can revolutionize the future of mobile networks. MEC brings compute and storage capabilities to the edge of the network closer to end-users. This enables faster data processing and improved user experience by reducing latency. MEC has the potential to decrease the burden on the core network by transferring computational and storage responsibilities to the edge, thereby reducing overall network congestion. Load balancing is critical for effectively utilizing the resources of the MEC. This ensures that the workload is distributed uniformly across all of the available resources. Load balancing is a complex task and there are various algorithms that can be used to achieve it, such as round-robin, least connection, and IP hash. To differentiate between heavily loaded and lightly loaded servers, current load balancing methods use an average response time to gauge the load on the edge server. Nevertheless, this approach has lower precision and may result in an unequal distribution of the workload. Our study introduces a dynamic threshold calculation technique that relies on a response-time threshold of the edge servers using K-means clustering. K-means based proposed algorithm classifies the servers in two sets (here K = 2), i.e., overloaded and lightly loaded edge servers. Consequently, workload is migrated from overloaded to lightly loaded servers to evenly distribute the workload. Experimental results show that the proposed technique reduces latency and improves resource utilization.