Dependency-Aware Task Offloading in Vehicular Edge Computing

Abstract

With the advancements in Internet Of Things (IOT), Internet Of Vehicles (IOV) has imposed a new challenge for computation intensive applications of smart vehicles. These applications require task computations in real time e.g obstacle detection. route navigation and congestion avoidance etc. However, the variable execution requirements cannot be met due to limited computa tional resources available locally on vehicles. Volunteer devices to facilitate the end devices applications, Road Side Units (RSUs) are located at various locations to execute the critical computational tasks that are offloaded from nearest vehicles. These critical computation intensive task comprises of exe cution deadlines and dependency constraints in the form of Directed Acyclic Graph (DAG) with data flows transmitted from one computation level to the next level in multi-stage jobs, where tasks are executed in parallel on different processors. However, offloading of these critical DAG tasks with coflows is almost difficult. Proposed approach presents a task offloading framework for the purpose of minimal execution delays of computation intensive tasks in vehicular edge networks. In this regard, our work focuses on real time DAG tasks with coflows execution offloaded from nearby vehicles on road. The main contri bution of our research is to minimize the average service time , queuing delays and enhance the efficiency of our system as compared with benchmark task scheduling algorithm. Our work focuses on offloading of DAG task with 2 coflows, 4 coflows, 8 coflows and 16 coflows that make it more desirable than the traditional techniques. We shall show the proposed system performance by various simulations.