Rambus indicates that AI and high-performance computing (HPC) workloads are transforming the data center network, creating distinct front-end and back-end network domains. These two types of networks differ in terms of performance, scalability, and operational requirements.

Among them, the front-end network is responsible for connecting users, applications, storage, and managing the system; the back-end network provides the high-speed connections required between GPU, accelerators, and computing nodes.

This white paper explores the architecture and security considerations related to these dedicated networks, and explains how data center operators can protect critical data and communications without affecting the performance required by the workloads of the distributed AI and HPC.

Key points of the white paper include:

• The different roles played by front-end and back-end networks in modern AI and HPC infrastructure
• Requirements for each network domain in terms of performance, operations and maintenance, and security
• How to Protect Data and Communications in an Increasingly Distributed Computing Environment

Figure 1: Cybersecurity in AI NIC. Source: Rambus