Scalable and Efficient Intra- and Inter-node Interconnection Networks for Post-Exascale Supercomputers and Data centers

TL;DR

This paper discusses scalable and efficient interconnection networks designed for the complex, heterogeneous architectures of post-exascale supercomputers and data centers, aiming to address communication bottlenecks.

Contribution

It proposes novel network architectures optimized for high scalability and efficiency in heterogeneous, large-scale computing environments.

Findings

Reduced communication bottlenecks in large-scale systems

Improved data transfer efficiency across nodes

Enhanced scalability of interconnection networks

Abstract

The rapid growth of data-intensive applications such as generative AI, scientific simulations, and large-scale analytics is driving modern supercomputers and data centers toward increasingly heterogeneous and tightly integrated architectures. These systems combine powerful CPUs and accelerators with emerging high-bandwidth memory and storage technologies to reduce data movement and improve computational efficiency. However, as the number of accelerators per node increases, communication bottlenecks emerge both within and between nodes, particularly when network resources are shared among heterogeneous components.