Virtana announced the grant of U.S. Patent No. 12,340,249 B2, titled "Methods and System for Throttling Analytics Processing." 

The patented design introduces a priority-aware scheduling and backpressure mechanism that dynamically reorders and resubmits analytic tasks based on real-time resource availability, preventing overload, reducing long-tail latencies, and maintaining service levels under heavy demand.

Virtana secures U.S. patent for AI analytics orchestration, ensuring predictable performance and higher throughput.

The patented orchestration system applies the same priority-aware throttling and queue management to these AI analytics streams, so teams can:

• Protect critical model-health signals (e.g., drift, data quality, p95/p99 latency) during traffic spikes
• Avoid GPU memory pressure cascades by pacing downstream analysis and enrichment
• Keep LLM inference and retrieval pipelines observable without starving non-AI analytics.

Why it matters for customers

• Predictable performance underload: Cuts variance and long-tail latency for critical analytics, including AI model-health signals, making SLOs easier to meet.
• Higher effective throughput: Keeps pipelines moving by matching work to available capacity instead of stalling or crashing.
• Operational resilience: Applies controlled backpressure and intelligent retries that stabilize noisy, bursty workloads across AI and non-AI domains.
• Cost control without overprovisioning: Maintains performance headroom through smarter scheduling rather than permanent capacity increases on CPU/GPU resources.

"Enterprises run analytics at massive scale, and AI workloads are only exacerbating already beleaguered infrastructure and the teams that manage them. This patent formalizes a practical way to keep those pipelines stable and performant, especially when demand spikes," said Paul Appleby, CEO and President of Virtana. "The result is more predictable operations, fewer incidents, and better cost discipline across hybrid and AI environments."

The invention applies to high-volume analytics pipelines (e.g., metrics, logs, traces, events, and topology processing) and AI /ML telemetry. Tasks are queued with explicit priority indicators. When capacity is constrained, the system:

• Evaluates task priority and current queue position
• Defers or repositions lower-priority work instead of dropping it
• Resubmits tasks when resources are available
• Sustains flow by continuously selecting the next best task for current conditions.

"This patent gives our platform real-time control over analytics pipelines—so critical signals for AI systems like LLM inference, RAG, vector search, and GPU metrics stay prioritized under load," said Amitkumar Rathi, SVP of Product and Engineering at Virtana. "Customers get steadier SLOs, faster incident triage, and cleaner cost profiles without overprovisioning."

The patented capability underpins Virtana's analytics services across its observability platform and is available today as part of standard product updates.