Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Reinforcement Learning for Fault-Adaptive Routing in Eisenstein-Jacobi Interconnection Topologies

Mohammad Walid Charrwi, Zaid Hussain

TL;DR

This work evaluates fault-adaptive routing in Eisenstein-Jacobi interconnection networks by comparing a deterministic greedy baseline, Dijkstra's globally optimal routing, and a PPO-based reinforcement learning (RL) policy. The RL agent learns to navigate around fault clusters using local information, achieving near-optimal performance while avoiding the computational and state-distribution burdens of full topology knowledge. Empirically, RL delivers about 94% effective reachability and 91% packet delivery at nine faulty nodes, closely tracking Dijkstra and substantially outperforming greedy; it also sustains over 0.9 normalized throughput under congestion, often exceeding Dijkstra at low to moderate loads due to implicit load-balancing. These results demonstrate that RL can provide robust, self-healing routing in fault-prone NoCs without requiring global topology knowledge, representing a practical middle ground between greedy efficiency and global optimality.

Abstract

The increasing density of many-core architectures necessitates interconnection networks that are both high-performance and fault-resilient. Eisenstein-Jacobi (EJ) networks, with their symmetric 6-regular topology, offer superior topological properties but challenge traditional routing heuristics under fault conditions. This paper evaluates three routing paradigms in faulty EJ environments: deterministic Greedy Adaptive Routing, theoretically optimal Dijkstra's algorithm, and a reinforcement learning (RL)-based approach. Using a multi-objective reward function to penalize fault proximity and reward path efficiency, the RL agent learns to navigate around clustered failures that typically induce dead-ends in greedy geometric routing. Dijkstra's algorithm establishes the theoretical performance ceiling by computing globally optimal paths with complete topology knowledge, revealing the true connectivity limits of faulty networks. Quantitative analysis at nine faulty nodes shows greedy routing catastrophically degrades to 10% effective reachability and packet delivery, while Dijkstra proves 52-54% represents the topological optimum. The RL agent achieves 94% effective reachability and 91% packet delivery, making it suitable for distributed deployment. Furthermore, throughput evaluations demonstrate that RL sustains over 90% normalized throughput across all loads, actually outperforming Dijkstra under congestion through implicit load balancing strategies. These results establish RL-based adaptive policies as a practical solution that bridges the gap between greedy's efficiency and Dijkstra's optimality, providing robust, self-healing communication in fault-prone interconnection networks without requiring the global topology knowledge or computational overhead of optimal algorithms.

Deep Reinforcement Learning for Fault-Adaptive Routing in Eisenstein-Jacobi Interconnection Topologies

TL;DR

This work evaluates fault-adaptive routing in Eisenstein-Jacobi interconnection networks by comparing a deterministic greedy baseline, Dijkstra's globally optimal routing, and a PPO-based reinforcement learning (RL) policy. The RL agent learns to navigate around fault clusters using local information, achieving near-optimal performance while avoiding the computational and state-distribution burdens of full topology knowledge. Empirically, RL delivers about 94% effective reachability and 91% packet delivery at nine faulty nodes, closely tracking Dijkstra and substantially outperforming greedy; it also sustains over 0.9 normalized throughput under congestion, often exceeding Dijkstra at low to moderate loads due to implicit load-balancing. These results demonstrate that RL can provide robust, self-healing routing in fault-prone NoCs without requiring global topology knowledge, representing a practical middle ground between greedy efficiency and global optimality.

Abstract

The increasing density of many-core architectures necessitates interconnection networks that are both high-performance and fault-resilient. Eisenstein-Jacobi (EJ) networks, with their symmetric 6-regular topology, offer superior topological properties but challenge traditional routing heuristics under fault conditions. This paper evaluates three routing paradigms in faulty EJ environments: deterministic Greedy Adaptive Routing, theoretically optimal Dijkstra's algorithm, and a reinforcement learning (RL)-based approach. Using a multi-objective reward function to penalize fault proximity and reward path efficiency, the RL agent learns to navigate around clustered failures that typically induce dead-ends in greedy geometric routing. Dijkstra's algorithm establishes the theoretical performance ceiling by computing globally optimal paths with complete topology knowledge, revealing the true connectivity limits of faulty networks. Quantitative analysis at nine faulty nodes shows greedy routing catastrophically degrades to 10% effective reachability and packet delivery, while Dijkstra proves 52-54% represents the topological optimum. The RL agent achieves 94% effective reachability and 91% packet delivery, making it suitable for distributed deployment. Furthermore, throughput evaluations demonstrate that RL sustains over 90% normalized throughput across all loads, actually outperforming Dijkstra under congestion through implicit load balancing strategies. These results establish RL-based adaptive policies as a practical solution that bridges the gap between greedy's efficiency and Dijkstra's optimality, providing robust, self-healing communication in fault-prone interconnection networks without requiring the global topology knowledge or computational overhead of optimal algorithms.
Paper Structure (23 sections, 12 equations, 9 figures)

This paper contains 23 sections, 12 equations, 9 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Fault-Adaptive Routing
  4. Reinforcement Learning in Interconnection Networks
  5. Background
  6. Methodology
  7. Fault Model
  8. Greedy Adaptive Routing
  9. Reinforcement Learning Routing
  10. Dijkstra's Shortest Path Routing
  11. Experimental Evaluation
  12. Faulty Nodes Routing
  13. Greedy EJ Routing
  14. RL Routing
  15. Fault Adaptive Routing
  16. ...and 8 more sections

Figures (9)

  • Figure 1: EJ network generated with $\alpha = 3 + 4\rho$
  • Figure 2: EJ Sectors for network $\alpha = 3 + 4\rho$
  • Figure 3: Greedy Routing From Root Node $0$ To Node $5-5\rho$ On Topology $\alpha = 5 + 6\rho$
  • Figure 4: RL Routing From Root Node $0$ To Node $5-5\rho$ On Topology $\alpha = 5 + 6\rho$
  • Figure 5: Adaptive Routing using Dijkstra Algorithm From Root Node $0$ To Node $5-5\rho$ On Topology $\alpha = 5 + 6\rho$
  • ...and 4 more figures