Towards the Optimization of Parallel Discrete Event Simulation on Multi-core Shared-memory machines

The increase of software and hardware parallelism along with data volume has made more efficient ways of managing memory necessary, especially considering shared-memory architectures and the overhead related to common operating systems’ services. Parallel Discrete Event Simulation (PDES) has been the target of many optimizations over the years. However, PDES on multi-core shared-memory machines still lacks a broader consideration of memory-awareness to enhance its performance. The aim of the work carried out during my PhD is to provide a multi-layer approach to fully exploit speculative PDES platforms and improve the overall memory utilization leveraging Linux kernel’s services, tackling event processing in a memory-aware way, improving traditional speculative PDES operations like state saving in terms of intrusiveness reduction when using operating systems’ facilities, up to reducing the delay of critical-path operations such as simulation’s state inspection through output collection