Engineering Architecture Brief

NVIDIA UFM® (Unified Fabric Manager) is a comprehensive fabric management and monitoring platform designed for InfiniBand and high-performance Ethernet AI networks. It provides real-time visibility, telemetry, analytics, and automation for large-scale AI factories, HPC clusters, and cloud data center fabrics. UFM enables administrators to discover, monitor, and optimize network performance, detect congestion hotspots, automate provisioning, and troubleshoot issues across thousands of nodes and interconnects. It integrates deeply with NVIDIA Quantum InfiniBand switches and supports advanced features such as predictive analytics, fabric-wide orchestration, and AI-driven optimization for mission-critical workloads.

NV-UFM-FABRIC-01

NVIDIA UFM® (Unified Fabric Manager)

Advanced InfiniBand and AI fabric management platform providing real-time monitoring, automation, and performance optimization for large-scale HPC and AI networks.

System Metrics Table

AI Optimization AI-Assisted Fabric Optimization and Workload Awareness

Automation Automated Provisioning and Configuration Management

Congestion Detection AI-Driven Hotspot and Bottleneck Identification

Deployment Model On-Premises, Virtual Appliance and Cloud-Enabled

Deployment Scale Hyperscale AI and Enterprise Data Center Fabrics

Fault Management Predictive Fault Detection and Root Cause Analysis

Integration NVIDIA Quantum InfiniBand and Spectrum Ethernet Ecosystem

Monitoring Scope Switches, NICs, Links, and Compute Nodes

Network Visibility End-to-End Fabric Telemetry

Performance Analytics Real-Time and Historical Fabric Performance Analysis

Platform Type Fabric Management and Orchestration Platform

Primary Function InfiniBand and Ethernet Fabric Monitoring and Management

Protocol Support InfiniBand and Ethernet Fabrics

Scalability Designed for Thousands of Nodes and Multi-Fabric Environments

Security Role-Based Access Control and Secure Fabric Monitoring

Telemetry Streaming Telemetry and Deep Packet Insights

Topology Management Dynamic Fabric Topology Discovery

Use Cases AI Factories, HPC Clusters, Cloud Data Centers, Scientific Computing

User Interface Web-Based Dashboard with Real-Time Visualization