Virtana

Virtana provides hybrid infrastructure observability spanning application to infrastructure mapping, network and data flow visibility, and what it calls AI Factory Observability.

That last area is where the product has moved most decisively and it addresses a genuinely new capacity problem. GPU performance and cost optimisation targets utilisation to reduce infrastructure spend, and training and inference performance work increases throughput while minimising failed training runs. A failed training run is a distinctive kind of waste, consuming days of the most expensive compute available and producing nothing, which makes it worth monitoring in a way ordinary job failures are not.

Token economics and forecasting is the most novel component. Token spend is now a material line in many organisations' budgets, behaves nothing like traditional infrastructure cost because it scales with usage rather than provisioning, and almost no existing observability tool measures it at all.

The conventional capabilities remain substantial: application to infrastructure mapping across the full stack, deep network observability, and business transaction tracing following end-to-end user journeys across services. Power and sustainability intelligence supports sustainable operation at scale, which matters increasingly as AI workloads dominate data centre power draw, and AI security and guardrails cover compliant operation. Pricing is not published.