Collaborative service caching for delay minimization in vehicular edge computing networks

Abstract

Vehicular edge computing (VEC) is beneficial to reduce task offloading delay and service acquisition delay by pushing cloud functions to the edge of the networks. Edge servers have the computation and storage capacity to execute vehicular tasks and cache the services required for tasks execution. Due to the limited caching resources of a single edge server, the vehicles will obtain services from cloud servers when they can't obtain from their own associated edge server, which results to the increase of service acquisition delay. To this end, we establish a multiple edge servers collaboration caching framework to minimize the heterogeneity tasks execution delay of the all vehicles, including tasks offloading delay, services acquisition delay and tasks processing delay. Specifically, the edge servers collaboratively make slot level caching decisions, i.e., what to be cached in each slot according to vehicular tasks requirements. Based on this framework, we formulate a long-term optimization problem to minimize the heterogeneity tasks execution delay of the all vehicles under the long-term energy constraints. To solve it, we firstly construct a virtual energy deficit queue, and then we transform the target problem into a delay drift-plus-energy consumption minimization problem by utilizing Lyapunov optimization theory. The equal transformation problem is a 0-1 multi-knapsack problem, which is a NP-hardness problem. To solve it, we improved the greedy algorithm that retains the selection process of the greedy algorithm and the comparison and selection of the genetic algorithm. Extensive simulations illustrate that the proposed scheme achieves near optimal delay performance while strictly satisfying long-term energy constraints, and outperforms other baseline schemes in terms of time-averaged delay and time-averaged energy consumption.