Reference Architecture Diagram + Narrative (measure → allocate → optimize → attest)
┌─────────────────────────────────────────────────┐
│ SOURCES OF USAGE │
Cloud (IaaS/PaaS/SaaS) │ Colo/DC (power/PUE) │ Network/CDN/SIP │ GPU/AI │
Bills/ CURs / APIs │ Meters (kWh), DCIM │ NetFlow, QoE │ Jobs │
└───────────────┬───────────────┬────────┬────────┘
│ │ │
▼ ▼ ▼
┌──────────────────────────────────────────────────────────┐
│ COLLECTORS & NORMALIZATION (ETL/ELT + Streaming) │
│ • Vendor APIs (CUR, billing) • DCIM/meters • NetFlow/IPFIX│
│ • SKU/region maps • Instance/volume tags • Labels │
└──────────────┬──────────────────────────────┬─────────────┘
│ │
▼ ▼
┌──────────────────────────────────────┐ ┌────────────────────────────────────┐
│ COST & ALLOCATION ENGINE │ │ CARBON & ENERGY ENGINE │
│ • Showback/chargeback rules │ │ • Emission factors (vendor/region)│
│ • Tagging remediation / heuristics │ │ • Scope 1/2/3 mapping (GHG Prot.) │
│ • Shared spend apportionment │ │ • Location vs market-based calc │
│ • Budgets & alerts (per BU/service) │ │ • EAC/REC offsets & provenance │
└───────────────┬──────────┬──────────┘ └───────────────┬──────────┬─────────┘
│ │ │ │
▼ ▼ ▼ ▼
┌───────────────────────────────┐ ┌──────────────────────────────┐
│ OPTIMIZATION RECOMMENDATIONS │ │ POLICY & GUARDRAILS │
│ • Rightsize/stop/scale-to-zero│ │ • Budget caps, alerts, freeze│
│ • Reserved/Savings coverage │ │ • Egress/CDN/cache policies │
│ • Storage tiering/lifecycle │ │ • GPU/AI queues/schedules │
│ • Data egress / placement │ │ • Sustainability SLOs (gCO2e)│
└──────────────┬───────────────┘ └──────────────┬───────────────┘
│ │
▼ ▼
┌──────────────────────────────────────┐ ┌──────────────────────────────────┐
│ WORKFLOWS & APPROVALS (ITSM/CMDB) │ │ REPORTING & ATTESTATION │
│ • Change tickets for actions │ │ • Exec dashboards (cost/carbon) │
│ • CAB gates for risky savings │ │ • SEC/CSRD/ESG packs (WORM) │
│ • Owner routing (service/BU) │ │ • Supplier scorecards, offsets │
└──────────────────────────────────────┘ └──────────────────────────────────┘
Telemetry loop ──► AIOps (anomaly, forecast) • SIEM/SOAR (fraud/leak) • GRC (policy) • Data Catalog/Lineage (source-of-truth)
Narrative (how money and carbon stay visible, governed, and reduced)
1) Purpose & posture
- Objective: Create a cross-domain fabric that ingests real usage, allocates cost and carbon to owners/services/regions, drives optimization safely, and produces audit-ready attestations (SEC/CSRD/GHG Protocol).
- Posture: Data-first, policy-led budgets, owner-routed actions, and immutable proof of both financial and environmental stewardship.
2) Ingest & normalize (syntax)
- Pull cloud bills/CURs, SaaS invoices, DCIM meters (kWh, PUE), NetFlow/CDN/SIP minutes, GPU job hours.
- Normalize SKUs to internal service taxonomy; enrich with tags/labels and business owner from CMDB; backfill missing tags via heuristics.
3) Allocate (semantics)
- Showback/chargeback by service/BU/region/environment; apportion shared assets (egress, NAT, firewalls) with rules that survive audit.
- Budgets drive alerts; forecasts blend seasonality and planned events (launches, sports, holidays).
4) Carbon quantification (meaning + custody)
- Compute Scope 2 (location- vs market-based) with grid factors; Scope 3 (use of cloud services) with vendor-published EF or credible proxies.
- Track EAC/REC purchases and attach them to accounts/regions; store provenance and retirement certificates immutably.
5) Optimize (grammar of safe change)
- Recommendations include rightsizing, stop/hibernate windows, reserved/Savings plans, storage lifecycle, CDN cache improvements, egress minimization (data placement), GPU queueing/scheduling, model distillation.
- Each action is runnable via ITSM (change ticket), checked against SLO/SLA and compliance (PCI/PHI residency). Auto-rollback hooks are required for risky moves.
6) Guardrails & policy-as-code (pragmatics)
- Enforce budget caps and carbon SLOs (e.g., ≤X gCO₂e/request, ≤$ per 1k requests).
- Gate deployments that violate residency/egress/cost guardrails (ties to DevSecOps/GitOps, Arch 19) and Zero-Trust access fabric (Arch 16).
7) Reporting & attestation (truth on demand)
- Executive dashboards: unit economics ($/request, $/GB, $/agent-hour) + carbon intensity (gCO₂e/request, gCO₂e/GB).
- Regulatory packs: SEC climate, CSRD, ESG—with data lineage, emission factors, and offset proofs in WORM storage.
8) Resilience & integrity
- AIOps detects anomalies (sudden cost spikes, rogue egress); opens ITSM incidents.
- SIEM/SOAR flags fraud/leak (unexpected data transfer patterns) and can auto-throttle egress/CDN; every change is auditable.
Reference KPIs
- Unit cost: $/request, $/GB, $/agent-hour (trend ↓ month-on-month)
- Carbon intensity: gCO₂e/request or gCO₂e/GB (trend ↓)
- Coverage: ≥98% spend/tag attribution; ≥95% carbon attribution
- Savings realized: ≥70% of accepted recommendations in 90 days
- Budget variance: ≤5% per BU/service; Alert MTTA: <15 min
Minimal BOM (aligned with prior stack)
Collector/ETL for bills & meters, SKU maps, Tagging remediation, Allocation engine (showback/chargeback), Carbon factors/registry, Optimization engine, Budget/guardrail policy-as-code, ITSM/CMDB integration, AIOps (forecast/anomaly), SIEM/SOAR hooks, Reporting (dashboards + ESG packs), WORM evidence store, Data catalog/lineage.