Implementation Playbooks

Phase 0: Pilot Programme Proposal

Before committing billions to full sovereign cloud migration, a structured 24-36 month pilot programme validates technology choices, coordination mechanisms, and migration approaches with non-critical workloads.

This pilot addresses the PAC/NAO concern: "You want billions committed before any proof this approach works? Where's the pilot? Where's the gateway review?"

1. Executive Summary

Phase 0: Prove Before You Scale

Duration: 24-36 months | Budget: €300-500 million | Scope: 2-3 non-critical services per jurisdiction

Objective: Demonstrate that CloudStack-based sovereign infrastructure can successfully host real government workloads, achieve interoperability across jurisdictions, and deliver acceptable performance/security before committing to full programme.

Go/No-Go Decision: At Month 30, Cabinet Office/Treasury (UK), Commission (EU), TBS (Canada), and Finance (Australia) make binding decision based on defined success criteria.


2. Pilot Scope

2.1 Candidate Workloads per Jurisdiction

Workloads selected for pilot must be: (a) non-critical (failure doesn't affect citizens directly), (b) representative of common patterns, (c) currently on US cloud, (d) departmentally willing.

Pilot Workload Candidates
Jurisdiction Candidate Workloads Complexity Current Provider
United Kingdom 1. Internal policy collaboration platform Low-Medium Azure
2. Non-public datasets analytics platform Medium AWS
3. Development/test environments Low GCP
European Union 1. Internal document management system Low-Medium Azure
2. Staff HR portal Medium AWS
3. Non-production regulatory data platform Medium-High AWS
Canada 1. Internal collaboration tools Low Azure
2. Development environments for digital services Low-Medium AWS
Australia 1. Policy analysis platform Medium AWS
2. Internal learning management system Low Azure

2.2 What the Pilot Will NOT Include


3. Budget

Pilot Programme Budget (€ millions)
Category UK EU Canada Australia Shared Total
Infrastructure (pilot scale) €25 €50 €20 €25 €15 €135
Platform development €15 €30 €10 €15 €30 €100
Migration/integration €10 €20 €8 €10 €5 €53
Security certification €8 €15 €6 €8 €10 €47
Training €5 €10 €4 €5 €5 €29
Programme management €5 €10 €4 €5 €10 €34
Contingency (20%) €14 €27 €10 €14 €15 €80
TOTAL €82 €162 €62 €82 €90 €478

Budget notes: This represents approximately 5% of the full programme Likely Case estimate. Pilot infrastructure is sized for workloads listed, not full production. Equipment purchased for pilot can be incorporated into full programme if Go decision made.


4. Timeline and Gates

Pilot Timeline with Gateway Reviews
Phase Months Activities Gateway
P0.1 Setup 0-6 Establish governance, procure pilot infrastructure, deploy CloudStack core, set up ISCC coordination mechanisms Gate 0: Infrastructure ready; governance operational
P0.2 First workload 6-12 Migrate first workload per jurisdiction (dev/test environments); establish interoperability baseline Gate 1: First workload operational in all 4 jurisdictions
P0.3 Production pilot 12-24 Migrate remaining pilot workloads; validate cross-jurisdiction interop; security certification progress Gate 2: All pilot workloads operational; interop validated
P0.4 Evaluation 24-30 Operate under production conditions; performance monitoring; cost analysis; security audit; lessons learned Gate 3: GO/NO-GO DECISION
P0.5 Transition 30-36 If GO: Prepare Phase 1 business case, procure additional infrastructure
If NO-GO: Wind down pilot, migrate workloads back, document lessons
Phase 1 launch or orderly exit

5. Success Criteria for Go/No-Go Decision

The Gate 3 Go/No-Go decision requires achievement of all mandatory criteria and at least 70% of target criteria.

5.1 Mandatory Criteria (Must Achieve)

Criterion Measure Threshold
M1: Workloads operational All pilot workloads running on sovereign infrastructure 100% migrated and operational
M2: Uptime Infrastructure availability over 6-month assessment period ≥99.5% (excludes planned maintenance)
M3: Security certification Platform certified to national security standard At least one national certification per jurisdiction
M4: Data sovereignty Zero data leaving designated jurisdiction boundaries 100% verified by audit
M5: No security incidents No critical/high security incidents during pilot Zero critical; ≤2 high (contained)

5.2 Target Criteria (Aim for 70%+)

Criterion Measure Target
T1: Performance parity Response time vs previous US cloud Within 20% of previous
T2: Cost efficiency TCO comparison to US cloud equivalent Within 130% of US cloud cost
T3: Interoperability Cross-jurisdiction API compatibility ≥90% of defined APIs interoperable
T4: User satisfaction End-user and operator satisfaction survey ≥70% satisfied or better
T5: Skills transfer Staff certified on sovereign cloud operations ≥80% of target headcount certified
T6: Migration velocity Time to migrate typical workload Within 150% of planned duration
T7: Governance function ISCC coordination mechanisms working ≥80% of decisions made within SLA
T8: Supplier delivery Contracted suppliers meeting commitments ≥90% on-time, on-quality delivery
T9: Budget adherence Pilot spend vs approved budget Within 110% of budget (excl contingency)
T10: Full programme confidence Independent assessor confidence in scalability ≥70% confidence rating

6. Pilot Governance

6.1 Oversight Structure

Body Composition Role
Pilot Board Senior officials from each jurisdiction + independent chair Strategic oversight; gateway decisions; resource allocation
Technical Authority Chief architects from each jurisdiction Technical decisions; standards; architecture approval
Security Authority NCSC/ENISA/CSE/ASD representatives Security standards; certification oversight; incident response
Independent Assurance IPA (UK)/equivalent per jurisdiction Gateway reviews; delivery confidence assessment

6.2 Reporting


7. Key Risks and Mitigations

Risk Likelihood Impact Mitigation
CloudStack doesn't meet requirements Medium High Early technical validation; alternative evaluation (OpenStack, OCI Dedicated)
Interoperability fails Medium High Kubernetes-native approach; standard APIs; conformance testing from Month 6
Security certification delayed High Medium Early engagement with NCSC/ENISA/CSE/ASD; parallel certification tracks
Coordination gridlock Medium Medium Independent chair; clear escalation; majority voting for non-critical decisions
Skills shortage High Medium Early hiring; contractor augmentation; cross-jurisdiction training
Budget overrun High Low 20% contingency; strict change control; monthly financial review

8. No-Go Exit Strategy

If Gate 3 results in NO-GO decision, the following exit process applies:

Step Duration Activities Cost
1. Decision confirmed Month 30 Formal No-Go decision; exit plan approval -
2. Data migration back Months 30-34 Migrate pilot workloads back to US cloud providers €15-25M
3. Infrastructure disposal Months 34-36 Secure data destruction; hardware resale/reuse; contract termination €5-10M
4. Lessons documented Month 36 Comprehensive lessons learned report; technology recommendations €2M
5. Staff redeployment Months 30-36 Redeploy/outplace pilot staff; preserve skills Normal costs

Stranded Capital Risk: Maximum stranded capital in No-Go scenario is approximately €478M (full pilot budget) minus recoverable hardware value (~€80M) = €400M maximum loss. This represents ~0.5% of full programme Likely Case, an acceptable risk for de-risking a €85B programme.


9. Required Approvals

Jurisdiction Approving Authority Requirements
United Kingdom Cabinet Office + HM Treasury Business case approval; Spending Review allocation
European Union European Commission (DIGIT) Budget line approval; procurement authorization
Canada Treasury Board Secretariat Investment approval; GC IT standards compliance
Australia Department of Finance + Cabinet New Policy Proposal approval; Digital Investment Framework compliance

Prove It Before You Scale It

A €478M pilot is a small price to de-risk an €85B programme. The pilot either validates the approach (proceed with confidence) or reveals fatal flaws early (exit with €400M loss vs €85B failure).