Edge-Accelerated Parallel Algorithms for Real-Time Large-Scale Simulations

Abstract

Big-scale scientific simulations are core to the modern use of engineering and computational science applications, but are typically characterized by high latency and poor scalability when running on conventional central or cloud-based platforms. These challenges are discussed in this paper, and an edge-accelerated parallel simulation framework is proposed to allow real-time computationally-intensive workloads. The suggested methodology combines the domain decomposition strategy with hybrid parallel computing of the CPU, GU, and distributed edge nodes, enabling effective division of workload, and low-latency execution of tasks. It has the addition of a latency-conscious scheduling system to perform optimally when allocating tasks and reduce communication overhead between computing layers. Much experimental analysis shows that the suggested framework has a huge decrease in execution time with a high level of gains in speedup and computational efficiency over traditional parallel and cloud-based strategies. These findings affirm that edge computing and parallel simulation algorithm integration offers a high-performance, scalable solution, which is well-suited to real-time large-scale simulations in scientific and engineering areas like climate modeling, smart grids, and computational fluid dynamics.