LOAM: Low-Latency Communication, Caching and Computation in Data-Intensive Computing Networks

Jinkun Zhang, Edmund Yeh

Abstract: Deploying data- and computation-intensive applications such as large-scale AI into heterogeneous dispersed computing networks can significantly enhance application performance by mitigating network resource bottlenecks, including bandwidth, storage, and computing power. However, current resource allocation methods do not provide a comprehensive solution that jointly considers arbitrary topology, elastic resource amount, reuse of computation results, and congestion-dependent optimization objectives. These aspects are vital when modeling state-of-the-art heterogeneous dispersed computing networks with high demand volume. In this paper, we propose LOAM, a low-latency joint communication, caching, and computation placement framework. LOAM incorporates the above aspects with a rigorous analytical foundation. It tackles the formulated NP-hard cost minimization problem with two methods: an offline method with a constant factor approximation of 1/2, and an online adaptive method with a bounded gap from the optimum. Through extensive packetlevel simulation, LOAM outperforms multiple baselines in both synthesis and real-world network scenarios.