Abstract

Nowadays, the surge of Internet contents and the need for high Quality of Experience (QoE) put the backbone network under unprecedented pressure. The emerging edge caching solutions help ease the pressure by caching contents closer to users. However, these solutions suffer from two challenges: 1) a low hit ratio due to edges’ high density and small coverages. 2) unbalanced edges’ workloads caused by dynamic requests and heterogeneous edge capacities. In this paper, we formulate a typical cooperative edge caching problem and propose the MagNet, a decentralized and cooperative edge caching system to address these two challenges. The proposed MagNet system consists of two innovative mechanisms: 1) the Automatic Content Congregating (ACC), which utilizes a neural embedding algorithm to capture underlying patterns of historical traces to cluster contents into some types. The ACC then can guide requests to their optimal edges according to their types so that contents congregate automatically in different edges by type. This process forms a virtuous cycle between edges and requests, driving a high hit ratio. 2) the Mutual Assistance Group (MAG), which lets idle edges share overloaded edges’ workloads by forming temporary groups promptly. To evaluate the performance of MagNet, we conduct experiments to compare it with classical, Machine Learning (ML)-based and cooperative caching solutions using the real-world trace. The results show that the MagNet can improve the hit ratio from 40% and 60% to 75% for non-cooperative and cooperative solutions, respectively, and significantly improve the balance of edges’ workloads.

Problem to solve

Cooperative edge caching problem.

Challenge

1. a low hit ratio due to edges’ high density and small coverages.
2. unbalanced edges’ workloads caused by dynamic requests and heterogeneous edge capacities

Method

To address the two challenges of edge caching solutions discussed above, we formulate a typical cooperative edge caching problem which jointly optimizes the latency, traffic, and workload balance. The problem is proved to be an NP-complete problem. To solve it, we propose the MagNet, a decentralized and cooperative edge caching system. The MagNet has two innovative mechanisms to address the two challenges: 1) the Automatic Content Congregating (ACC) mechanism. First, the ACC utilizes a neural embedding algorithm to generate vectors for all contents, which tries to capture underlying patterns of historical traces. Second, a novel clustering algorithm is designed to cluster the vectors into some types. Third, the ACC guides requests to their optimal edges by type so that requests of the same type tend to congregate in the same edges. As a result, each edge accumulates contents of one dominant type more than other types. Then, more requests of this type are attracted to the edge. This process forms a virtuous circle between edges and requests, which eventually leads to a high hit ratio. 2) the Mutual Assistance Group (MAG) mechanism. When an overloaded edge emerges, the MAG finds some idle edges for it to form a temporary group. The group runs in a master-worker mode where the master, i.e., the overloaded edge, shift some of its workloads to the workers, i.e., idle edges, according to their capacities to balance their workloads. The group dismisses when the overload problem is eliminated.

Result

To evaluate the MagNet performance, we conduct experiments from three different perspectives using a real-world trace dataset. We compare the MagNet with some benchmark caching solutions, including classical, ML-based and cooperative solutions. The result shows that the MagNet can improve the hit ratio from 40% and 60% to 75% for non-cooperative and cooperative solutions, respectively, and significantly balance the edges’ workloads.

Bibtex

@inproceedings{peng2022magnet,
  title={MagNet: cooperative edge caching by automatic content congregating},
  author={Peng, Junkun and Li, Qing and Ma, Xiaoteng and Jiang, Yong and Dong, Yutao and Hu, Chuang and Chen, Meng},
  booktitle={Proceedings of the ACM Web Conference 2022},
  pages={3280--3288},
  year={2022}
}