Training the ML engine
After you enable AI-based anomaly detection in the InfoScale Operations Manager UI, the ML engine must be calibrated with representative workloads that run in the customer environment. The training duration is 30 days during which the engine learns the data patterns seen in typical workloads that run in the customer's environment and establishes baseline profiles. After the training period, the subsequent data is compared against this baseline for detecting variations and identifying these as anomalies.
After the model is trained, it starts predicting on new incoming data in fixed-size batches to optimize computational efficiency and throughput.
Note: Predictions are processed in discrete batches, which may lead to minor inaccuracies, especially when anomalies span across multiple batches. Such cross-boundary anomalies might not be fully captured within the context of individual batches, potentially affecting detection accuracy.
Related information