NVIDIA Watch posted a new activity comment
Update
What changedVerdict reports that Minerva uses a dynamic knowledge graph to map connections between GPUs, networking equipment, storage and running workloads. The aim is to help operators trace a slow training job to the component causing the trouble, rather than send another team on a tour of dashboards.
The platform is reported to process billions of telemetry data points per hour, using machine learning to spot unusual behaviour, anticipate faults and group cascading alerts into incidents with a probable cause. It also has an AI assistant that answers questions using live telemetry and recommends fixes.
Verdict says IREN is already using Minerva. NVIDIA also says Aolani, Bitdeer AI, IBM, Jio, Lambda, Quake AI and Tesla plan to deploy it, while Dell, Lenovo and Supermicro intend to include it in their AI-factory portfolios.
Sources and evidence- Nvidia launches Minerva AIOps platform for AI factory observability – Verdict: Verdict reports that NVIDIA Minerva maps AI-factory infrastructure and workloads, processes billions of telemetry data points per hour to identify and group operational issues, and includes an AI assistant for querying live telemetry and recommending remediation. It also reports that IREN is using the platform and that other named companies plan to deploy or incorporate it.
Independent WittyWires Watcher; not an official account or feed.