Skip to main content

Data Center Operations

Data Center Operations is the FactVerse module for facility and equipment operations in data centers. It brings asset identity, data-hall context, BMS point mapping, equipment health, predictive maintenance, alert diagnosis, work-order dispatch, SLA review, and closed-loop evidence into one operating workflow.

Use this guide when a data center operations team needs to connect source-system records with a governed digital twin, review current operating risk, and prepare maintenance actions with traceable evidence.

For AI-assisted analysis surfaces, see Data Center Operations AI Tools.

Current service boundary

Data Center Operations is implemented across several runtime surfaces:

SurfaceCurrent role
FactVerse frontendProvides data center overview, data hall filtering, asset list and detail, predictive queue, diagnosis console, closed-loop view, operations dashboard, BMS mapping view, and model-ops status.
Core backend and Data Center Operations servicesProvide /api/v1/dcops dashboards, KPIs, asset reads, health trends, prediction history, diagnosis history, planning recommendations, dispatch candidates, closed-loop records, BMS mappings, operations analytics, snapshot export, and live event stream.
Inspector and work ordersProvide alert, work-order, feedback, attachment, and field execution context for operating actions.
Predictive MaintenanceProvides the broader equipment health and remaining-life workflow when data center assets need detailed maintenance analysis.
DFSPrepares source data, BMS mappings, meter readings, work records, equipment identity, and governed datasets before operations depend on them.
AI Engine and AdvisorProvide health scoring, remaining-life estimation, alert diagnosis text, and optional standard or NVIDIA-backed execution metadata where enabled by the project.

The module supports operational review and maintenance workflow preparation. Site teams should confirm data mappings, engineering assumptions, and action approval before using outputs for field execution.

Operations workflow flow

What users can do today

WorkflowWhat users can review or perform
Overview dashboardReview asset counts, alert state, work-order state, data hall scope, health score, PUE and WUE indicators when data is mapped, and operating risk signals.
Asset operationsReview data center assets, equipment detail, recent alerts, latest diagnosis, open work orders, health trend, prediction history, and remaining-life intervals.
Health reviewCalculate or read equipment health scores, risk levels, anomaly scores, and aggregate health distribution.
Predictive queueReview 7, 30, and 90 day risk buckets, high-risk equipment, remaining useful life intervals, and maintenance attention candidates.
Diagnosis consoleCreate or read alert diagnosis records, review confidence, trace ID, evidence source, and diagnosis history.
Dispatch and closed loopReview pending dispatch candidates, approve work-order creation, link feedback, and inspect whether diagnosis, action, and feedback evidence are complete.
PlanningReview maintenance planning recommendations, generate draft plans, inspect conflicts, and resolve time-window or resource conflicts.
Operations dashboardReview queue aging, SLA rates, SLA breach list, diagnosis reuse, dispatch conversion funnel, KPI deltas, trends, and exportable operations snapshots.
Live operations streamSubscribe to operations events for alerts, work orders, and risk snapshot updates.
BMS mappingReview, validate, and publish source-point to target-field mapping rules for BMS data.
Model operationsReview the active engine mode, fallback metadata, and status indicators for standard or NVIDIA-backed execution paths.

Before you start

Prepare the tenant, site, and source data needed for the workflow:

RequirementNotes
Tenant and site contextDCOps reads are tenant-scoped. Site and data center filters should match the active operating scope.
PermissionsMain read surfaces use dcops.view. Prediction surfaces use dcops.predict.view. Diagnosis and work-order actions require the matching diagnosis, work-order, or feedback permissions.
Asset identityEquipment records should have stable equipment IDs, names, type, location, criticality, and data hall or room assignment.
Source mappingsBMS points, alerts, work orders, meter readings, and equipment records should map to the same asset identity.
Work ownershipDefine who reviews alerts, accepts dispatch candidates, assigns work orders, resolves SLA breaches, and records feedback.
Data qualityConfirm timestamps, units, source freshness, missing values, and stale mappings before interpreting health or prediction output.
Review recordKeep diagnosis text, trace IDs, work-order links, feedback, and snapshot exports with the operating review.

For source data setup, start with Getting Started with DFS, Connect BMS to a Facility Twin, and Prepare Predictive Maintenance Signal History.

Open Data Center Operations

Open the FactVerse application and use the Data Center Operations workspace. Current frontend surfaces include:

ViewRoute
Overview/datacenterops/dashboard
Asset list/datacenterops/assets
Asset detail/datacenterops/assets/:assetId
Predictive queue/datacenterops/predictive-queue
Diagnosis console/datacenterops/diagnosis
Closed-loop view/datacenterops/closed-loop
Operations dashboard/datacenterops/operations
Integrations/datacenterops/integrations

Module availability depends on tenant configuration. The backend guard uses the Data Center Operations module enablement setting before serving /api/v1/dcops paths.

Prepare operational data

Data Center Operations depends on a consistent operating data package:

Data areaTypical preparation
Asset hierarchyMap sites, data centers, data halls, rooms, racks, equipment, and criticality to stable IDs.
BMS pointsMap source point names to target fields and validate rule coverage before publishing.
AlertsMap alert severity, status, title, source equipment, and timestamps to the asset layer.
Work ordersConnect open, assigned, in-progress, completed, and feedback records to alerts and equipment.
Meter and energy readingsPrepare power, water, or other utility readings with units and time windows for operating review.
Predictive signalsPrepare health snapshots, failure prediction history, and remaining-life inputs for maintained assets.

Use DFS when source systems need connector configuration, source-to-target mapping, sync monitoring, data-quality review, and governed dataset preparation.

Review dashboard and assets

Start from the overview dashboard, then drill into assets:

  1. Select the relevant site or data center scope.
  2. Review asset, alert, work-order, health, and risk summary cards.
  3. Open the asset list and filter by risk or equipment type.
  4. Open an asset detail page to review recent alerts, diagnosis records, open work orders, health trend, prediction history, and feedback status.
  5. Export or save the operating snapshot when the review needs to be shared.

PUE and WUE indicators should be read as operating review signals. Confirm meter coverage, load definitions, and calculation assumptions before using them in management reporting.

Review health and predictive risk

The module can calculate or read equipment health and predictive risk:

AreaWhat to check
Health scoreComposite score, risk level, anomaly score, engine mode, timestamp, and factor breakdown when available.
Health trendRecent health-score history for the selected asset and review window.
Prediction history7, 30, and 90 day failure probabilities, remaining-life range, model version, and engine mode.
High-risk equipmentEquipment whose predicted risk passes the selected threshold and time window.
Predictive queueRisk buckets for near-term maintenance planning and operations meetings.

Treat health and prediction output as evidence for review. Maintenance owners should compare the output with inspections, work orders, source signal freshness, and site constraints before approving action.

Diagnose alerts and close the loop

The diagnosis workflow connects alerts to work orders and feedback:

  1. Open the diagnosis console or an alert detail.
  2. Review the alert severity, affected equipment, recent alert pattern, and source timestamps.
  3. Create or read a diagnosis record.
  4. Check confidence, trace ID, evidence source, and diagnosis history.
  5. Create or link a work order after an owner accepts the recommended action.
  6. Record feedback when the work order closes.
  7. Confirm the closed-loop view includes diagnosis, action, and feedback evidence.

Diagnosis may use Advisor-backed text generation or a rule-based fallback. Both paths should remain visible in the review evidence through source, confidence, trace ID, and audit record.

Plan and dispatch maintenance

Planning and dispatch views help teams turn risk into scheduled work:

SurfaceUse
Planning recommendationsReview candidate maintenance windows, estimated duration, priority, and risk score.
Planning draftGenerate a draft plan for a selected window and inspect resource or time-window conflicts.
Pending dispatchReview open alerts that have no open work order yet.
Batch approveApprove selected dispatch candidates after diagnosis reuse or diagnosis creation has been reviewed.
Bulk operator actionsAssign owners, normalize priority, append escalation notes, retry diagnosis, or reconcile closed-loop evidence when permissions allow it.

Keep action ownership clear. Bulk operations should be used for triage and queue management after the shift lead or operations owner approves the selected records.

Review BMS mapping and model operations

BMS mapping defines how source data becomes useful operational context:

  1. Open Integrations.
  2. Review the latest published BMS mapping version and source.
  3. Validate mapping rules for required sourcePoint and targetField coverage.
  4. Publish mapping changes after source owners confirm the mapping.
  5. Keep the audit record with the operating handover.

Model operations exposes engine mode and status metadata. Use it to confirm whether the workflow is running in the standard path or a project-enabled NVIDIA path, and record fallback or degraded status during acceptance.

API surface

Data Center Operations APIs are grouped under /api/v1/dcops:

GroupExample endpoints
Overview and KPIs/, /dashboard/overview, /dashboard/overview/trends, /dashboard/kpis
Assets and health/assets, /assets/rul-intervals, /assets/{assetId}/detail, /assets/{assetId}/health, /assets/{assetId}/health/trend, /health/summary
Predictions/assets/{assetId}/predictions, /assets/{assetId}/predictions/history, /predictions/high-risk
Diagnosis/diagnosis/from-alert/{alertId}, /alerts/{alertId}/diagnosis, /alerts/{alertId}/closed-loop
Planning/planning/recommendations, /planning/generate, /planning/{planId}/conflicts, /planning/{planId}/resolve
Dispatch and feedback/dispatch/pending, /dispatch/alerts/{alertId}/approve, /dispatch/batch-approve, /recommendations/{diagnosisId}/create-work-order, /work-orders/{workOrderId}/feedback, /work-orders/{workOrderId}/closed-loop
Operations dashboard/dashboard/operations, /dashboard/operations/kpi-delta, /dashboard/operations/queue-aging, /dashboard/operations/predictive-queue, /dashboard/operations/trends, /dashboard/operations/sla-rates, /dashboard/operations/sla-breaches, /dashboard/operations/diagnosis-reuse, /dashboard/operations/dispatch-funnel, /dashboard/operations/snapshot, /dashboard/operations/snapshot.csv, /dashboard/operations/events
Integrations and model ops/integrations/bms/mappings, /integrations/bms/mappings/validate, /integrations/bms/mappings/publish, /model-ops/status, /model-ops/engine-mode
Automation helpers/automation/diagnosis/bulk-retry, /automation/close-loop/reconcile, /dashboard/operations/risk-chips/bulk-assign-assignee, /dashboard/operations/risk-chips/bulk-normalize-priority, /dashboard/operations/risk-chips/bulk-sla-escalation-note

Validation checklist

Before using Data Center Operations output in an operations meeting or maintenance decision, confirm:

  • tenant, site, and data center filters match the review scope;
  • asset, equipment, alert, BMS point, meter, and work-order IDs resolve to the same operating objects;
  • source timestamps, units, and freshness are acceptable for the decision;
  • BMS mapping changes have owner approval and audit evidence;
  • health and prediction outputs are reviewed with inspection and work-order context;
  • diagnosis records include trace ID, evidence source, confidence, and responsible reviewer;
  • work-order feedback is recorded after action so the closed-loop view can reflect the outcome;
  • exported snapshots and CSV files are stored with the operating review package when used in handover.