Real-Time Systems Optimization with Black-box Constraints and Hybrid Variables

TL;DR

This paper tackles real-time system optimization with black-box schedulability constraints and a hybrid mix of continuous and discrete variables. It introduces NORTH+, a coordinate-descent extension of the NORTH framework, which alternates between continuous optimization (NMBO/VE) and discrete optimization (heuristics like RM) to achieve scalable improvements. The approach yields roughly 20% better solutions than the original NORTH on simulated task sets, highlighting practical gains in schedulability-aware design. The discussion outlines trade-offs between optimality, applicability, and efficiency, and sketches avenues for future work in multi-objective optimization and deeper discrete optimization integration.

Abstract

When optimizing real-time systems, designers often face a challenging problem where the schedulability constraints are non-convex, non-continuous, or lack an analytical form to understand their properties. Although the optimization framework NORTH proposed in previous work is general (it works with arbitrary schedulability analysis) and scalable, it can only handle problems with continuous variables, which limits its application. In this paper, we extend the applications of the framework NORTH to problems with a hybrid of continuous and discrete variables. This is achieved in a coordinate-descent method, where the continuous and discrete variables are optimized separately during iterations. The new framework, NORTH+, improves around 20% solution quality than NORTH in experiments.

Figures (4)

• Figure 1: NORTH+ is an extension of the NORTH optimization framework Wang2023RTAS. The NMBO and VE components are shown in Fig. \ref{['main_framework_fig_north']} and explained more in Section \ref{['section_north_review']}, the discrete optimization is introduced in Section \ref{['section_discrete_opt']}. The continuous and discrete variables in NORTH+ are optimized separately: when optimizing continuous variables, the discrete variables are treated as constants, and vice versa.
• Figure 2: Optimization framework NORTH from Wang2023RTAS that optimizes only continuous variables.
• Figure 3: From Wang2023RTAS, an example that visualizes how NORTH optimizes the continuous variables $\textbf{C}$ in the figure. Expanding the schedulable region at $\textbf{C}^{(k)}$ gives NORTH more chances of optimization.
• Figure 4: Performance comparison between NORTH and NORTH+RM.