Unified Operations Data Layer
Unified Operations Data Layer
The Lumi observability warehouse that collects events, the CI/CMDB that defines operated assets and their owners, and the ontology that links IT and business objects work as one layer.
Common data foundation
Observability data, CMDB and ontology in one layer
Observability events are linked to CIs in the ingest pipeline, and CIs become ontology objects connected to IT and business relations. Agents query this layer only within allowed scope and keep evidence rows for every decision.
- Events linked to CIs and workspaces at ingest
- Owners, support groups and change approval per CI
- CI classes double as ontology object types so IT and business share one graph
Suited to these organizations
- Operations teams whose logs and metrics pile up separately in each tool
- Teams that need CI owner and change approval records
- Organizations that must explain the business impact of IT incidents
Common situations before adoption
- Scattered observability dataLogs, metrics and events pile up per tool, so the same incident is seen on different screens.
- Assets without ownersRecords of which server or service belongs to whom, and who may change it, are missing or stale.
- Meaning that does not connectIT services and business plans are not linked, so the business impact of an incident cannot be explained.
Unified Operations Data Layer
Key features
Observability ingest
Collects sources through 9 transports (collect, SOAR, syslog, NetFlow, SFTP, Kafka, Splunk, OTLP, RUM) and 10 collectors.
Loglake pipelines
Refines sources in 17 stages. Parsing takes 7 and golden takes 4. CI resolution and fingerprinting take 1 each. Hygiene and redaction take 4.
CI/CMDB
Defines operated assets and dependencies with 164 CI classes and 48 relation types. Each CI belongs to one system.
Owners and scope
Each CI has a technical owner, business owner and support group, with permissions split by customer group > tenant > workspace > system.
Ontology
CI classes become object types and the 48 relations plus 2 become 50 link types. Business plan objects join the same graph.
Evidence rows and query guard
Agent queries are limited to read-only and workspace scope, and the rows used are kept as evidence.
From observability event to approval
The path from an observability event to a human approval.
Observability event
Logs, metrics and changes collected
CI linking
Hosts and IPs mapped to CIs and workspaces
Ontology object
CI class = object type, relation = link
Agent decision
Root cause and blast radius
Human approval
Approve with evidence before running and auditing
FormatObservability data attaches to CIs, and CIs become ontology objects.
Roles
Agents link and propose; people own and approve.
| Stage | AI agent | People |
|---|---|---|
| Ingest and refine | Runs pipelines, detects ingest anomalies | Decide sources and retention |
| CI linking | Links hosts and IPs to CIs | Review unlinked assets |
| Relations | Proposes observed relations | Approve additions and changes |
| Ontology | Proposes object and property changes | Review, approve, roll back |
| Query and decide | Queries evidence rows with scoped SQL | Approve with evidence in view |
CI ownership
Defines who owns, changes and approves each CI.
| Role | Recorded in | Responsibility |
|---|---|---|
| Technical owner | CI owner (owned_by) | Configuration and operations, change requests |
| Business owner | Business owner, department and cost center | Business impact, priorities |
| Support and managing groups | Support group and managed-by group | Incident response, daily operations |
| Change approval (CAB) | Change record and approval request | Dual approval for normal changes, single fast approval for emergencies |
| AI agent | Approval request (requesting agent) | Only proposes relation changes, applied after human approval |
| Scope administrator | Scope grant | Customer group > tenant > workspace > system boundaries and expiry |
Scale confirmed in code
The number of definitions implemented in the product.
164
CI classes
Basis Same as ontology object type seed
48
CI relation types
Basis Link types add 2 more for 50
17
Loglake pipeline stages
Basis Parse, golden, CI resolution, fingerprint, hygiene
9
Ingest transports
Basis With 10 collectors
17 Loglake pipeline stagesThe number of stages that refine source logs, by kind.
Delivery options
Ways to use this product. Each option lists what is included, who operates what and the steps to start.
Installed SaaS
Installed in your environment and run by you- What is included
- Use Lumi Cloud or install Lumi in your cloud or environment. Source data is gathered through 9 collection paths and 10 collectors.
- Operation
- You operate it, and AIONIS handles installation, updates and technical support. You decide what to collect and how long to keep it.
- Getting started
- Log and metric collection paths for core systems are connected first. After CIs and owners are organized, the ontology expands.
Consulting
Civil services and AX- What is included
- We design CMDB ownership, permission scope and collection targets.
- Operation
- Run by AIONIS consultants and engineers.
- Getting started
- Start with a review of your current asset list and collection status.
Education
AI Innovation Center- What is included
- A course covering the basics of observability data and AI agents.
- Operation
- Run by the AIONIS AI Innovation Center.
- Getting started
- In stage 1 of the AI Native transformation course, you build your first observability dashboard and a data query agent.
RolloutConnect small, then expand.
Connect ingest
Connect logs and metrics from core systems
Output Ingest inventory
Clean up CMDB
CIs and systems with owners and support groups
Output CI ownership table
Extend ontology
Add business objects and links, change approval
Output Object and link list
Connect agents
Start decisions with scoped queries and evidence rows
Output Decision reports with evidence
Use cases
Typical use cases carried out with features built into the product. These are not specific customer cases.
Attaching events to assets
- Situation
- Logs carry only host names, so the affected service is unknown.
- What the AI agents do
- The CI resolution stage turns hosts and IPs into CIs and workspaces.
- Result
- Events appear with their service and support group.
Adding observed relations
- Situation
- Real call relations are missing from the CMDB.
- What the AI agents do
- Agents propose observed calls as relation additions, and owners approve.
- Result
- Only approved relations enter the CMDB, nothing automatic.
Explaining business impact
- Situation
- It is hard to explain which business plan an IT incident affects.
- What the AI agents do
- Business plan objects join the same graph, and impact is traced along CI relations.
- Result
- Impact is reported with the evidence path.
Adoption consulting for your industry and operating environment
Tell us your industry, size and current systems. We will propose the right products and delivery option.
