Bonree ONE is an AI observability platform that aligns heterogeneous data into trusted entities, so Sage AI operations agents and AIOps can find root causes faster.

Most observability projects don’t fail because there is too little data. They fail because the data doesn’t line up. A single host monitored by a cloud provider, an in-house platform and a CMDB shows up as three unrelated objects, each with its own ID, field names and units. Add more sources, and the confusion grows with them.

This article is based on the Bonree College session Data Integration: Governance & Fusion. It explains how custom integration in Bonree ONE, our AI observability platform, treats integration as more than a data pipe. It is an entity-centric process that makes data trustworthy first, so that AI-driven analysis, root cause analysis and AIOps can be built on a solid base.

Why heterogeneous data breaks observability

Enterprise operations data is heterogeneous by nature: monitoring APIs from public, private and hybrid clouds, in-house platforms or customized open-source tools, log pipelines and CMDB asset records. Each system evolved on its own, with its own entity identifiers, field names and data formats.

In practice, the bottleneck is rarely coverage. It is three structural conflicts:

• Protocol mismatch. Sources package data differently, and batched payloads must be split into individual records before anything else can happen.

• Ambiguous entity identity. The same host may be collected by several platforms, yet each assigns its own instance ID, so records can’t be matched directly.

• Inconsistent semantics. Even after entities are matched, metric units and statistical dimensions can still differ.

Without a governance mechanism, every new data source adds more misaligned entities and ambiguous semantics to the platform.

Two integration paths: standard and custom

Bonree ONE offers two integration paths, matched to how predictable the data is.

Standard integration is for sources with a known structure, such as databases and middleware. Field structures and entity relationships can be modeled in advance, and collection runs through SmartAgent, ready to use out of the box.

Custom integration is for sources whose structure isn’t known up front: in-house monitoring systems, data reported over generic protocols, or requirements that can’t be defined in advance. These can’t follow a fixed template, so users configure entity mappings and association rules based on the raw data. Technically, custom integration collects data through SmartGate, controls scheduling with collection tasks, and manages processing with data streams.

The logic is simple. Where requirements are clear, prebuilt capabilities save effort. Where they aren’t, configurable data streams handle the transformation. Both paths feed the same metric model and entity relationship graph.

How custom integration works: three layers of entity governance

What sets custom integration apart from simple data forwarding is that it resolves entities as well as moving data. The process has three layers.

1. Protocol parsing and batch expansion. Cloud monitoring APIs usually return batched payloads. In a typical case, two raw batch records expand into 232 individual data records. This step is the foundation: if it goes wrong, nothing downstream works.

2. Entity uniqueness. Recognizing the same operational entity depends on key distinguishing fields. In the AWS EC2 demo, Bonree ONE uses a composite identifier made of the host identifier and the data center. This mirrors dynamic CMDB registration: entity instances are created and registered from field values, which avoids the risk of a single ID being duplicated or going stale.

3. Cross-source entity merging. Suppose the platform already holds historical monitoring data for a host, and metrics reported from the cloud side for that same host are then connected. Rules recognize that both belong to one entity, so the host doesn’t end up with several isolated data views.

Together, the three layers form a complete governance chain, from raw heterogeneous data to analyzable entities.

Keeping it running: verification, reuse and layered troubleshooting

Identifying entities is only half the job. Once a configuration takes effect, the platform reports operational metrics such as collection traffic, parsed data volume and task status, so configuration flaws surface early instead of in the middle of an incident.

Built-in metric models, dashboards and alert rules can also be reused for newly governed sources, which lowers the cost of building an observability system. These presets are applied on demand; they don’t take effect automatically.

Because non-standard integrations are highly customized and rarely configured perfectly on the first pass, Bonree ONE provides a layered troubleshooting path that follows the processing chain:

• Connection layer. Confirm that raw data actually reaches the processing chain. For collection tasks that stop updating, check the synchronization status between collector and probe, and the network path.

• Parsing layer. Validate data processing. For dirty data or abnormal record counts, compare the raw field structure against the custom mapping rules.

• Semantic layer. Check statistical accuracy. When data lands successfully but metrics look off, review units, statistical dimensions and other definitions, then refine the entity mapping.

This keeps the whole path, from ingestion to parsing to alignment, operable and traceable.

From trusted data to intelligent operations

Data integration isn’t a preparatory step for observability. It is the foundation. Custom integration governs the data first, and only then supports intelligent analysis. The value builds across three layers:

• Data layer: a trusted foundation. Entity uniqueness and cross-source merging turn scattered, heterogeneous raw data into entity-unique, consistently defined data, giving AI anomaly analysis and root cause analysis a reliable base.

• Platform layer: reusable capabilities. Newly governed sources reuse built-in metric models, dashboards and alert rules. Standard integration works out of the box and custom integration extends on demand, balancing efficiency with breadth of coverage.

• Business layer: intelligent operations. A unified entity view connects observability data across domains and breaks down data silos, so anomalies are found sooner and root causes pinpointed more accurately, keeping business systems stable and efficient.

This trusted data is also what lets Sage AI, the AI agent operations platform in Bonree ONE, reason over one consistent picture of your environment.

If your data sits in several systems that describe the same host differently, Bonree ONE custom integration aligns it into one trusted entity view, so AI root cause analysis and Sage AI work from a single consistent picture of your environment.

Next in the series: how Sage AI turns database inspection from hours into 4 minutes. To learn more about Bonree ONE, visit en.bonree.com.

FAQ

What is custom integration in Bonree ONE?

Custom integration is the path for data sources whose structure can’t be defined in advance, such as in-house monitoring systems or data sent over generic protocols. Using SmartGate, users configure entity mappings and association rules, and data streams handle parsing and normalization before the data enters the shared metric model and entity graph.

When should I use standard integration instead?

Use standard integration for sources with a known structure, such as databases and middleware. Their fields and relationships can be modeled in advance and collected through SmartAgent, ready to use out of the box.

How does Bonree ONE keep one host from appearing as multiple entities?

It identifies entities by key fields (in the AWS EC2 demo, a composite of host identifier and data center). When data from another source arrives, rules recognize that it belongs to an existing entity and merge it, rather than creating a duplicate.

What should I check if data is ingested but the metrics look wrong?

This is usually a semantic-layer issue. Review metric units, statistical dimensions and other definitions, then refine the entity mapping. Bonree ONE’s layered troubleshooting (connection, parsing, semantic) helps isolate which stage the problem sits in.

Do I need to rebuild dashboards and alerts for every new data source?

No. Governed data sources can reuse the platform’s built-in metric models, dashboards and alert rules. These presets are applied on demand rather than automatically.

Why does trusted data matter for AI observability and AIOps?

AI anomaly detection and root cause analysis are only as reliable as the data beneath them. Entity-aligned, consistently defined data prevents duplicate entities and conflicting metrics from skewing the results.