Toward SOAFEE-Enabled Real-Time Systems: A Container Auto-Configuration Method with End-to-End Latency Guarantees

Abstract

SOAFEE encounters real-time challenges in resource-constrained automotive systems, where improper container configuration can cause catastrophic timing failures. This paper presents an automated method for guaranteeing endto-end latency in containerized autonomous driving systems. We model the system as a hierarchical DAG (H-DAG) and formulate container configuration as an Integer Linear Programming (ILP) problem. Our approach selects optimal settings to bound critical path latency below a predefined threshold. Experimental results show the method efficiently configures 150 containers under 0.22 seconds and outperforms a greedy baseline, improving the probability of meeting latency deadlines by approximately 45.55%.