Deadlock-free routing for Full-mesh networks without using Virtual Channels

2025-10-16

AI Summary

Existing deadlock-free adaptive routing for high-radix, low-diameter fully connected networks (e.g., HyperX, Dragonfly) relies heavily on virtual channels (VCs), incurring significant area, power, and scalability overheads. This paper proposes TERA—a VC-free, topology-embedded adaptive routing algorithm. Its core innovation is the first use of embedded physical subnetworks to construct deadlock-free non-minimal paths, circumventing the performance degradation of conventional link-ordering schemes under adversarial traffic. TERA provides formal deadlock-freedom guarantees while drastically reducing hardware complexity. Experimental evaluation demonstrates up to 80% throughput improvement under adversarial traffic (100% for application kernels), 50% reduction in buffer requirements, and a 32% performance gain over state-of-the-art VC-free routing on 2D-HyperX topologies.

Technology Category

Planning, Routing, and Scheduling: Optimization of Spatio-temporal SystemsMachine Learning: Hardware-aware MLSearch and Optimization: Distributed Search

Application Category

Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphsSystems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deploymentsResponsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

High-radix, low-diameter networks like HyperX and Dragonfly use a Full-mesh core, and rely on multiple virtual channels (VCs) to avoid packet deadlocks in adaptive routing. However, VCs introduce significant overhead in the switch in terms of area, power, and design complexity, limiting the switch scalability. This paper starts by revisiting VC-less routing through link ordering schemes in Full-mesh networks, which offer implementation simplicity but suffer from performance degradation under adversarial traffic. Thus, to overcome these challenges, we propose TERA (Topology-Embedded Routing Algorithm), a novel routing algorithm which employs an embedded physical subnetwork to provide deadlock-free non-minimal paths without using VCs. In a Full-mesh network, TERA outperforms link ordering routing algorithms by 80% when dealing with adversarial traffic, and up to 100% in application kernels. Furthermore, compared to other VC-based approaches, it reduces buffer requirements by 50%, while maintaining comparable latency and throughput. Lastly, early results from a 2D-HyperX evaluation show that TERA outperforms state-of-the-art algorithms that use the same number of VCs, achieving performance improvements of up to 32%.

Problem

• Eliminating packet deadlocks in Full-mesh networks without virtual channels
• Reducing switch overhead from virtual channels in high-radix networks
• Improving routing performance under adversarial traffic without VCs

Innovation

• Proposes TERA algorithm for deadlock-free routing
• Uses embedded physical subnetwork without virtual channels
• Reduces buffer needs by 50% while maintaining performance