Asynchronous Distributed Particle Swarm Optimization for Edge Cloud Architectures

Edge computing poses a range of optimization problems such as load balancing, resource provisioning, and workload placement. Particle swarm optimization (PSO) is a bio-inspired stochastic optimization algorithm, with the objective to iteratively improve the solution of a problem for a given objective. The distribution of PSO workloads to the edge would transfer resource-intensive computational tasks from central large cloud data centers to the edge, resulting in more efficient use of existing resources there. However, this edge architecture introduces performance and fault tolerance challenges, due to the resource-constrained edge environment with a high probability of faults.We present here an asynchronous variant of an edge-distributed PSO algorithm built on top of the Apache Spark distributed computing framework [2]. This PSO variant aims at solving performance problems introduced by the execution in an edge setting. Our asynchronous PSO algorithm that distributes the load across multiple executor nodes can effectively realize both coarse- and fine-grained parallelism, allowing us to obtain a substantial performance increase.