3D Torus network topology

Processor power is centerpiece to determine the performance of an HPC system, but it is not the only factor. One of the key aspects of parallel computers is the communication network that interconnects the computing nodes. The network is the one guaranteeing fast interactions between CPUs and allowing the processors to cooperate: this is essential to solve complex computational problems in a fast and efficient way.

HPC systems are also increasingly asked to be more scalable. Petascale and then exascale installations require and will require hundreds of thousands of cores to efficiently work together.

As a switchless way to connect nodes in a HPC system, the 3D torus network topology is a good answer to both problems of speed and scalability. It can assure very low latency and linear system scalability.

Aurora 3d Torus

The 3D Torus is an optional component of the Aurora Systems.

Eurotech proposes a solution that leverages the 3D Torus benefits, but, at the same time, leaves to the users the flexibility and the freedom to run all types of applications.

The approach is called Unified Network Architecture, which means that the Aurora systems have 3 different networks working in concomitance on the same machine: 2 fast independent networks (Infiniband, 3D Torus) and a multi-level synchronization network.

Rather than with a costly application-specific integrated circuit (ASIC) technology, Aurora 3D Torus is driven by FPGAs circuits, injecting more flexibility in the hardware. This flexible approach is mirrored in the use of both a GPL and a commercial distribution of 3D torus software.

Partitioning functionality

Each link in the torus architecture is physically implemented by two lines (main and redundant) that can be selected (in software) to configure the machine partitioning (full 3D Torus or one of the many 3D sub-tori available). In this way, redundant channels allow system repartitioning on-the-fly. The possibility of partitioning the system into sub-domains permits to create system partitions that communicate on independent tori, effectively creating different execution domains. In addition, each subdomain can benefit from a dedicated synchronization network.

This site uses technical or similar cookies and third parties profiling cookies in aggregate form to improve your navigation experience and the use of the site. If you want to know more or opt out of some cookies consult our cookies extended policy. By closing this banner or continuing the navigation you are agreeing to our use of cookies.