Performance and energy consumption balanced optimization of ternary optical computers based on partially synchronous vacation queuing.
TL;DR
A novel TOC task scheduling strategy and corresponding three-stage queuing model is established by introducing the partially synchronous vacation mechanism, which provides a new theoretical basis and practical scheduling guidance for performance-energy bi-objective optimization of TOC systems.
Abstract
This paper addresses the core challenge of balancing high real-time performance and low energy consumption in practical ternary optical computer (TOC) systems with hardware-induced state-switching overhead including switching delay and switching energy consumption. By introducing the partially synchronous vacation mechanism, we establish a novel TOC task scheduling strategy and corresponding three-stage queuing model. We thoroughly analyze the influences of key system parameters on queueing performance, total energy consumption, and social utility under Poisson traffic, with the no-vacation scheme as benchmark. Numerical results reveal that the proposed model can achieve optimal system social utility under various task arrival rates by tuning the number of vacation-enabled processors and vacation-related parameters. Supplementary bursty traffic experiments further verify its robustness under fluctuating real-world workloads. This work provides a new theoretical basis and practical scheduling guidance for performance-energy bi-objective optimization of TOC systems.