A proactive caching and offloading technique using machine learning for mobile edge computing users

Abstract

The mobile edge computing (MEC) paradigm provides cloud and application services at the “edge” of user networks for providing ubiquitous access to resources. The heterogeneous services cause varying network traffic that sometimes increases delay. In edge-based services, concurrency in data distribution requires caching and offloading features. This article introduces a proactive caching technique with offloading (PCTO) ability by considering the need for parallel user services. The proposed method performs demand-aware offloading to meet the concurrent service dissemination requirements. Network-level caching and its forecast in concurrent service distribution are performed to reduce the response time. The offloading and caching processes are streamlined using deep recurrent learning for the failing service distribution intervals. In the learning process, the machine is trained for prior failures and for pursuing offloading instances. Based on the learning output, the caching level and offloading rate are determined for the queuing services. The performance of the proposed method is verified using the metrics service ratio, response failures, latency, offloading rate, and caching ratio.