Latency minimization model towards high efficiency edge-IoT service provisioning in horizontal edge federation

Edge computing plays a critical role in IoT as it potentially minimized the computation tasks response latency demanded by time-critical IoT applications. The growth of IoT users with high demanded computation power as well as ultra-low latency tasks may cause the performance degradation. One way to minimize the end-to-end (E2E) latency is to form horizontal edge federation (HEF) so that the computation resources can be shared with each participating edge node. Achieving ultra-low latency in HEF-IoT ecosystem involves setting two factor: resource allocation and task dispatching. This two factor interact with each other yet feasible solutions must provide satisfactory service level to meet latency constraints demanded by target applications. In this paper, we formulate it as E2E latency minimization problem and proposed a two-phase iterative (TPI) approach. The TPI method alternately determines optimal task dispatching and computation resource allocation. We exploit bin packing problem and, genetic algorithm (GA) to determine the edge nodes, and the required computation resources. The simulation results show that by using TPI approach, we can achieve more throughput, minimum E2E latency and optimum number of required edge nodes.