Adaptive techniques for scalable optimistic parallel discrete event simulation

Mikida, Eric P.

Permalink

https://hdl.handle.net/2142/104831 Copy

Description

Title

Adaptive techniques for scalable optimistic parallel discrete event simulation

Author(s)

Mikida, Eric P.

Issue Date

2019-04-15

Director of Research (if dissertation) or Advisor (if thesis)

Kale, Laxmikant

Doctoral Committee Chair(s)

Kale, Laxmikant

Committee Member(s)

• Nicol, David
• Snir, Marc
• Jefferson, David

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Parallel Computing
• PDES
• Simulation
• Adaptive
• GVT
• Load Balancing

Abstract

Discrete Event Simulation (DES) can be an important tool across various domains such as Engineering, Military, Biology, High Performance Computing, and many others. Interacting systems in these domains can be simulated with a high degree of fidelity and accuracy. Furthermore, DES simulations do not rely on a global time step and simulated entities are only updated at discrete points in virtual time at which events occur. The particular DES simulation engine handles simulation logic and event scheduling, while the particular models written by domain experts need only focus on model-specific logic. As models grow in size and complexity, running simulations in parallel becomes an attractive option. However, a number of issues need to be addressed in order to effectively run DES simulations in parallel in a distributed environment. The issue of how to synchronize PDES simulations has been addressed in a number of ways, using various types of either conservative or optimistic protocols. Optimistic simulation synchronization has shown several benefits over conservative synchronization, but it is also more complex and brings with it some unique challenges. Two of these challenges are synchronizing event execution across distributed processes, and maintaining a high accuracy in the speculative execution of events. This thesis aims to address these challenges in order to make optimistic simulations even more effective and reliable. Specifically, this thesis explores a variety of GVT algorithms in an attempt to lower synchronization costs, while utilizing other techniques such as dynamic load balancing to maintain a high event execution efficiency and keep work balanced across execution units. Most importantly, these techniques aim to make the simulator robust and adaptive, allowing it to work effectively for a variety of models with different characteristics and irregularities.

Graduation Semester

2019-05

Type of Resource

text

Permalink

http://hdl.handle.net/2142/104831

Copyright and License Information

Copyright 2019 Eric Mikida

Owning Collections