Articles by tag "InfiniBand"

1. Optimizing AI Inference Clusters: MCX7 AAS-NEAT Adapters
   Selene Gong

   25
   As the demand for Large Language Models (LLMs) and generative AI surges across Singapore and APAC, GPU-as-a-Service (GPUaaS) providers and AI engineers face a critical challenge: your inference cluster is only as fast as its network. Microsecond delays in transferring KV caches or synchronizing ...
2. How to Choose Between NVIDIA InfiniBand and Ethernet for AI Infrastructure Deployment
   Selene Gong

   33
   When AI infrastructure projects move from concept to deployment, one of the most important design decisions is whether the network should follow an InfiniBand path or an Ethernet path. This is not just a hardware preference. It affects how the environment scales, how the architecture is planned, ...
3. How to Simplify Global AI Infrastructure Procurement with a One-Stop High-Performance Networking Supply Chain
   Selene Gong

   18
   ...
4. Troubleshooting InfiniBand Subnet Manager Failover High-Availability OpenSM Architecture for MQM8790 Unmanaged Clusters
   Selene Gong

   18
   Quick Take In deployments based on unmanaged MQM8790 switches, the entire InfiniBand fabric control plane relies on an external Subnet Manager (SM). Implementing a high-availability active/standby OpenSM architecture with distinct sm_priority values prevents full cluster outages during ...
5. PCIe Network Card for Servers: PCIe Lanes, Speed, Form Factor, and Compatibility
   Selene Gong

   18
   Quick Take Selecting a PCIe network card is not simply a matter of matching the advertised network speed with an available PCIe slot. A reliable choice must also match PCIe generation and lane width, server slot wiring, port type, Ethernet or InfiniBand requirements, ...
6. How to Choose Networking Hardware for an AI/HPC Cluster: A Practical Sizing and Compatibility Guide
   Selene Gong

   18
   Quick Take Choose the network fabric first, calculate ports based on GPU nodes and single- or dual-rail design, then validate the switch, NIC, cable, PCIe, firmware, and airflow as one complete system. Selecting networking hardware for an AI or HPC cluster ...
7. How to Choose a Data Center Switch: Port Speed, Leaf-Spine Design, AI Networking, and Compatibility
   Selene Gong

   18
   Quick Take Choose a data center switch by network role and fabric first, then compare port speed, port density, uplink capacity, optics, software, congestion management, power, airflow, and lifecycle support. The right data center switch is not simply the model with the highest port ...
8. Top 20 NVIDIA Mellanox Networking Products for AI and HPC
   Selene Gong

   18
   Quick Take There is no single best NVIDIA Mellanox product for every AI or HPC deployment. The right choice depends on the fabric protocol, link speed, server adapter, port density, breakout plan, optical distance, and software environment. This guide compares 20 NVIDIA Mellanox switches, ...