Selective congestion control for mixed-critical networks-on-chip

Abstract

Modern Networks-on-Chip (NoCs) must accommodate a diversity of temporal requirements, e.g., providing guarantees for real-time senders while reducing adverse performance impact on best-effort (BE) traffic. In this work, we propose a protocol-based adaptive congestion control. By selectively detouring real-time or BE traffic (i.e load balancing) and dynamic throttling of BE, we allow improving the NoC performance without costly hardware extensions. The introduced method offers safe and efficient integration of mixed-critical workloads through the coupling of flow control mechanisms with the path selection based on the current NoC state. The requested real-time reliability of the network is achieved through a predictable synchronization with control messages supported by a formal analysis and an experimental evaluation.