Graph-Based Data Deduplication in Mobile Edge Computing Environment

Ruikun Luo, Hai Jin, Qiang He, Song Wu, Zilai Zeng, Xiaoyu Xia

Abstract: Mobile edge computing (MEC) extends cloud computing by deploying edge servers with computing and storage resources at base stations within users’ geographic proximity. The networked edge servers in an area constitute an edge storage system (ESS), where edge servers cooperate to provide services for the users in the area. However, the potential of ESSs is challenged by edge servers’ constrained storage resources due to their limited physical sizes. A straightforward method to tackle this challenge is to reduce data redundancy in the ESS. The unique characteristics and constraints in the MEC environment, e.g., edge servers’ geographic coverage and distribution, render conventional data deduplication techniques designed for cloud storage systems obsolete. In this paper, we make the first attempt to study this novel Edge Data Deduplication (EDDE) problem. First, we model it as a constrained optimization problem with the aim to maximize data deduplication ratio under latency constraint by taking advantage of the collaboration between edge servers. Then, we prove that the EDDE problem is $$\mathcal {NP}$$ -hard and propose an approach named EDDE-O for solving the EDDE problem optimally based on integer programming. To accommodate large-scale EDDE scenarios, we propose a $$ln\alpha +1$$ -approximation algorithm, namely EDDE-A, to find sub-optimal EDDE solutions efficiently. The results of extensive experiments conducted on a widely-used dataset demonstrate that EDDE-O and EDDE-A can solve the EDDE problem effectively and efficiently, outperforming four representative approaches significantly.