Technical structure: past conventional monitoring
Weave’s technical basis depends on a hybrid information graph structure. It processes totally different information sorts by way of specialised analytical engines. It doesn’t try to power all community information by way of giant language fashions (LLM). This design selection addresses accuracy considerations inherent in making use of generative AI to specific networking information.
“There’s truly a large threat of hallucination in the event you’re processing time collection information by way of LLMs,” Subramaniyan mentioned. “So we truly are very particular and cautious to not course of any time collection information by way of LLMs.”
The system makes use of graph analytics for relationship modeling between community entities. It maintains vector databases for similarity searches. All parts feed right into a unified information graph. This captures each logical relationships (bodily connections) and semantic relationships (practical dependencies) inside the community infrastructure.
Distinguishing state adjustments from anomalies
The core differentiator in Weave’s strategy lies in its skill to tell apart between professional state adjustments and real anomalies in real-time. Conventional monitoring instruments deal with each situations as deviations from baseline. Each require guide investigation to find out acceptable responses.
Weave addresses this by way of temporal evaluation. It considers change patterns over time. This functionality turns into important in large-scale networks. Tons of or 1000’s of configuration adjustments might happen every day. The system learns from community engineer suggestions. It builds institutional information about what constitutes regular operational adjustments versus points requiring intervention.
Integration and deployment mannequin
Weave doesn’t change present community monitoring infrastructure. It positions itself as a topology intelligence layer that enhances present instruments. The agent identifies particular community segments or nodes requiring consideration. This permits conventional monitoring instruments to focus their evaluation efforts extra successfully.