Latency Management in Task Offloading from IoT to MEC

Abstract

Mobile Edge Computing is a state-of-art technology which is being used to providereal-time environment by computing and responding in shorter timelines to the IoT generated requests. The task computation requests being sent to these servers is called task offloading, which is a highly complex process. The decision to offload a task for remote computation is done with the aim of receiving responses within few instant but these server in turn gets heavily loaded with thousands of computation requests as each server is connected to a number of IoT devices. This may result in situations like imbalanced workloads and resource starvation. The occurrence of this situation is caused due to adoption of full task offloading policy targeted IoT environment. Many works have already been observed in order to improve this offloading approach but it remains a complex issue. In this research study it is being tried to propose a latency minimizing procedure with optimal task splitting method. This will not only prevent resource starvation but also reduce total incurring latency and lead to quick responses. The proposed work will facilitate parallel remote and local computationof task and thus reducing the total computation time with optimal set of resources. The proposed model has been validated using hypothesis testing including Shapiro-wilk, One-wayANOVA Test, F-Test two-sample Z-test, Multiple Linear Regression Test and was successfully found to be efficient in minimizing latency with the use of partial offloading policy and have resulted in optimal resource allocation when compared to other traditionally existing offloading policies.