High-performance computing framework with desynchronized information propagation for large-scale simulations

摘要

The parallel implementation of complex, often micro-scale simulation systems such as social, artificial life, or traffic systems poses a significant challenge for scientists and often requires the use of super-computer devices. At the same time, it is quite difficult to develop a software system capable of being scaled up to hundreds or thousands of nodes due to the inherent algorithmic problems hampering efficiency (such as the need to propagate information about the state of the environment-in other words, the inevitable need to have a means of state synchronization). In this paper, we propose a framework based on a desynchronized method for the distribution of information inspired by the propagation of smell. It does not inhibit the overall scalability and efficiency, making use of available HPC resources. The implementation presented here leverages an actor model for parallelization as well as Akka Cluster distribution mechanisms for cluster management, providing a seamless tool-chain for large-scale simulations following realistic rules found in nature. As an example, three real world-inspired simulations are presented and tested, proving a linear scalability of up to 3456 computing cores and correctness with a growing degree of distribution. (C) 2018 Elsevier B.V. All rights reserved.