Adaptive Data Collector

Use the Adaptive Data Collector to leverage adaptive, context-aware data collection and gain relevant diagnostic information without having to rely on continuous, high-volume data collection.

With Adaptive Data Collector, BMC Helix Operations Management adopts a precision-on-demand telemetry that has the following key features:

Resting state: This state triggers low-frequency polling to establish baseline health and trends, minimizing data volume and cost.
Active state: This state automatically enables high-frequency telemetry when a trigger occurs. Examples: deployment events, latency deviations, or alerts
Self-optimizing pipeline: This feature dynamically scales data collection up or down without manual intervention, ensuring that observability aligns with operational reality.

Adaptive Data Collector capabilities

Prevent cost leaks by eliminating waste during the steady state
Capture what matters during change and risk
Reduce the mean time to resolve (MTTR)
Treat observability as a business-aware capability rather than a passive data list

Scenario

Monitoring specific metrics with context-aware data collection

Apex Global's online service is responding slowly. The performance degradation is recent and is not related to an ongoing maintenance window. To quickly mitigate the situation, their administrators need to understand which infrastructure component is affected.

Apex Global can benefit from collecting data only from the specific metrics that are affecting the online service.

Agent types, skills, and prompts

The Adaptive Data Collector supports the following agent types, skills, prompts, and supported models.

Agent types

Health Breach Analyzer: Identify and analyze a health breach in the infrastructure system
Process Correlation: Identify the process that is affected because of the health breach
Deep Monitor Trigger: Identify the subprocesses that are affected because of the health breach
Policy Mapping: Converts selected actionable metrics into deployment-ready policy mappings to enable monitoring

Out-of-the-box skills

Parsing a four-row hierarchy as shown below:
Monitor type > instance > attribute > data
Detecting breaches by using percentile thresholds
Establishing a correlation between health, process, and deep monitoring as shown below:
Health > Process > Deep-monitor
Detecting the lag relation. The following types are valid:
- process_leads
- process_lags
- simultaneous
Estimating causal strength by using causation-scoring heuristics
Recommending an adaptive poll interval. The following values are valid:
- 10s
- 1m
- 10m
Grouping and deduplication of actionable attributes
Performance reporting and scale extrapolation

Out-of-the-box prompts

Detect breached health attributes and windows
Correlate process metrics with breached health signals
Trigger deep monitor selection when SQL/MSSQL context appears
Generate policy mappings for monitor enablement
Recommend polling intervals from volatility/anomaly patterns

Supported models

Azure OpenAI deployment (configured in config.json, current gpt-5-mini)
For more information, see https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/reasoning?tabs=csharp%2Cgpt-5.

User roles and permissions

Make sure that you have the following roles and permissions to configure and use the Adaptive Data Collector:

Role	Description
BMC Helix Operations Management administrator	Runs the Adaptive Data Collector

Process overview

The following diagram explains how the Adaptive Data Collector works:

Adaptive Data Collector use case

The following table lists the task that you can perform by using the Adaptive Data Collector:

Task	Description	Reference
Monitor specific metrics	Leverage adaptive, context-aware data collection and gain relevant diagnostic information without having to rely on continuous, high-volume data collection	Investigating a sudden performance degradation by using the Adaptive Data Collector