Sign in
← AI PulseSep 22, 2026

Wire · news · Single-source brief

NVIDIA Topograph Enables Topology-Aware Workload Scheduling

NVIDIA has released Topograph, an open-source toolkit that discovers and normalizes cluster topology from cloud APIs or on-premises fabric systems, publishing this data for workload managers to optimize GPU workload placement.

By Illumora Editorial

Source · Sep 22, 2026, 5:16 PM · On Illumora · Sep 22, 2026, 5:23 PM

Media from the primary source — shown here so you can stay on Illumora.

Rewritten from one allowlisted primary — not independent enterprise reporting. Lanes →

Brief drafted by Illumora’s editorial model from the linked primary source. Ops desk reviews flagged pieces. How we write →

Read the source →NVIDIA Developer Blog — Topology-Aware Workload Scheduling with NVIDIA Topograph | NVIDIA Technical Blog
Save

NVIDIA has introduced Topograph, an open-source toolkit designed to improve GPU workload placement in AI factories. This tool discovers cluster topology from various sources and normalizes it into a common model, which schedulers can then use to place AI workloads in efficient locality domains. Topograph continuously regenerates its topology view to keep schedulers current without manual maintenance.

Key Points

  • NVIDIA Topograph discovers cluster topology from cloud APIs or on-premises fabric systems.
  • It normalizes topology data into a common model for schedulers.
  • Topograph publishes topology data as Kubernetes node labels, Slurm configuration, or Slinky ConfigMaps.
  • Supported cloud providers include Google Cloud, Lambda, Nebius, Nscale, and Oracle Cloud Infrastructure.
  • On-premises support is available for InfiniBand, Spectrum-X, and Multi-Node NVLink domains.
  • Topograph integrates with KAI Scheduler for topology-aware gang scheduling.
  • Operators can deploy Topograph via Helm on Kubernetes or as native packages on Slurm clusters.
  • Simulation utilities are available for testing without production hardware.
  • The toolkit's upstream main reflects support as of September 16, 2026.

Context

According to the NVIDIA Developer Blog, AI factories are power-limited systems where GPU workload placement is a key optimization. Poor placement can fragment topology domains, force traffic across shared links, reduce throughput, and increase job costs. Distributed workloads benefit from communication locality, which NVIDIA NVLink and NVIDIA Spectrum-X Ethernet provide.

Topograph addresses the challenge of maintaining an accurate view of GPU and fabric relationships as clusters change. It provides this current view to schedulers, enabling them to favor nearby resources and avoid bottlenecks that arise from spreading tightly coupled workloads across distant domains.

Why It Matters

Builders and operators can use NVIDIA Topograph to optimize resource utilization and performance in AI factories. By providing schedulers with accurate, up-to-date topology information, the toolkit helps ensure that GPU workloads are placed efficiently, reducing latency and improving throughput for AI training and inference tasks.

What To Do

  • Deploy Topograph from the dsx-ai-factory/topograph GitHub repository.
  • Explore the DSX OS ecosystem for cluster orchestration to understand broader integration.
  • Test Topograph using its simulation utilities to evaluate its impact without requiring production hardware.
  • Note the supported cloud providers and on-premises fabric systems for compatibility with existing infrastructure.