Cloud Migration: Assessment, Planning & Execution

Cloud migration moves workloads from on-premises or legacy environments to cloud platforms. This skill produces comprehensive migration plans covering 7R assessment, workload analysis, wave planning, landing zone design, execution patterns, and post-migration optimization. [EXPLICIT]

Principio Rector

Migrar sin estrategia 7R es mover problemas de datacenter a la nube. Cada workload merece una clasificación explícita (rehost, replatform, refactor, repurchase, retire, retain, relocate). Wave planning reduce riesgo. Cutover rehearsal es obligatorio — nunca se hace un cutover en producción sin haber ensayado en staging.

Filosofía de Migración Cloud

1. 7R assessment per workload. No existe "migrar todo igual". Cada aplicación tiene contexto, dependencias, y restricciones que determinan su estrategia óptima. [EXPLICIT]
2. Wave planning reduces risk. Migraciones big-bang son apuestas. Waves incrementales permiten aprender, ajustar tooling, y escalar throughput progresivamente. [EXPLICIT]
3. Cutover rehearsal is mandatory. Si el runbook no se ha ejecutado end-to-end en staging, no está listo para producción. Incluye rollback — siempre. [EXPLICIT]
4. Retire ruthlessly. Cada workload que no migra es costo evitado. La clasificación "retain" y "retire" son decisiones legítimas de arquitectura. [EXPLICIT]

Inputs

The user provides a migration program or portfolio name as $ARGUMENTS. Parse $1 as the program/portfolio name used throughout all output artifacts. [EXPLICIT]

Parameters:

• {MODO}: piloto-auto (default) | desatendido | supervisado | paso-a-paso
  • piloto-auto: Auto para discovery y 7R classification, HITL para wave sequencing y cutover decisions. [EXPLICIT]
  • desatendido: Cero interrupciones. Plan de migración completo automáticamente. Supuestos documentados. [EXPLICIT]
  • supervisado: Autónomo con checkpoint en 7R classification y landing zone design. [EXPLICIT]
  • paso-a-paso: Confirma cada workload classification, wave assignment, landing zone component, y cutover step. [EXPLICIT]
• {FORMATO}: markdown (default) | html | dual
• {VARIANTE}: ejecutiva (~40% — S1 7R classification + S3 wave plan + S5 cutover) | técnica (full 6 sections, default)

Before generating migration plan, detect existing infrastructure context:

!find . -name "*.tf" -o -name "*.yaml" -o -name "inventory*" -o -name "*.csv" | head -20

If reference materials exist, load them:

Read ${CLAUDE_SKILL_DIR}/references/migration-patterns.md

---

When to Use

• Planning migration of workloads from on-premises to cloud
• Classifying applications using the 7R framework
• Mapping application dependencies for migration sequencing
• Designing cloud landing zones for migrated workloads
• Planning migration waves and cutover execution
• Validating post-migration performance and decommissioning legacy systems

When NOT to Use

• Designing cloud-native architecture for new applications --> use cloud-native-architecture skill
• Infrastructure platform design (VPC, compute, storage) --> use infrastructure-architecture skill
• Current-state analysis without migration intent --> use asis-analysis skill
• Enterprise portfolio strategy --> use enterprise-architecture skill

---

Delivery Structure: 6 Sections

S1: Migration Assessment & 7R Classification

Classify every workload using the 7R framework to determine optimal migration strategy. [EXPLICIT]

7R Strategies:

• Rehost (Lift-and-Shift): Move as-is to cloud VMs. Fastest, lowest risk, no modernization. Best for: quick wins, legacy apps with no code access.
• Replatform (Lift-and-Reshape): Minor adjustments (managed DB, container runtime). Moderate effort. Best for: apps that benefit from managed services without rewrite.
• Refactor (Re-architect): Redesign for cloud-native. Highest effort, highest cloud benefit. Best for: strategic apps with 5+ year lifecycle.
• Repurchase (Drop-and-Shop): Replace with SaaS equivalent. Best for: commodity functions (email, CRM, HR).
• Retire: Decommission. Best for: unused, redundant, or replaced applications.
• Retain: Keep on-premises. Best for: compliance constraints, recent investment, nearing end-of-life.
• Relocate (Hypervisor-level): Move VMs at hypervisor level (VMware Cloud on AWS). Best for: rapid datacenter exit.

Classification Decision Matrix:

Factor	Rehost	Replatform	Refactor
Business criticality	Low-medium	Medium	High (strategic)
Technical complexity	Any	Low-medium	High (code access required)
Timeline pressure	Urgent (<3 months)	Moderate (3-6 months)	Long-term (6-18 months)
Cloud benefit needed	Minimal	Moderate	Maximum
Budget	Low ($)	Medium ($$)	High ($$$)

Output: Application-by-application classification table with strategy, rationale, effort estimate, risk level, and wave assignment.

S2: Workload Analysis & Dependency Mapping

Automated Discovery Tools:

Tool	Provider	Method	Discovers
AWS Application Discovery Service	AWS	Agent or agentless	Server config, utilization, network connections, process data
AWS Migration Hub	AWS	Aggregator	Unified view across discovery + migration tools
Azure Migrate	Azure	Agent or appliance	Servers, databases, web apps, VDI; includes assessment + migration
Azure App Service Migration Assistant	Azure	Web app scan	.NET/Java web app readiness for Azure App Service
Google Cloud Migration Center	GCP	Agent or API import	Technical fit, TCO, migration complexity scoring
Cloudamize	Multi-cloud	Agent-based	Performance profiling, right-sizing, cost projection, dependency mapping
Flexera One	Multi-cloud	CMDB integration	License optimization, cloud cost modeling, portfolio rationalization

• Start with agentless discovery for broad inventory (network flow analysis, DNS logs). Deploy agents on Tier 1/2 applications for detailed dependency and performance data.
• Run discovery for minimum 30 days to capture full monthly patterns (batch jobs, month-end processing).

Application Inventory Fields: Name, owner, business unit, criticality tier (T1-T4), technology stack, infrastructure requirements (CPU, memory, storage, network), performance baseline (response time, throughput, utilization), compliance requirements. [EXPLICIT]

Dependency Graph:

• App-to-app: API calls, file transfers, shared databases
• App-to-infrastructure: specific hardware, network, licenses
• Data dependencies: shared databases, ETL feeds, replication chains
• External: third-party APIs, SaaS integrations, partner connections

Data Gravity Analysis:

• Data gravity score per workload: (data_volume_TB * access_frequency * compliance_anchors) -- higher score = harder to move
• Transfer time estimate: (data_size_GB / bandwidth_Gbps) * 8 / 3600 = hours + 30-50% verification overhead
• Connectivity: Direct Connect/ExpressRoute for >5TB; VPN for smaller transfers; Snowball/Data Box for >50TB
• Anti-pattern: Underestimating data volumes and WAN throttling. Start continuous replication weeks before cutover.

Application Difficulty Scoring:

• Score 1-5 on: technical complexity, dependency count, data gravity, compliance constraints
• Composite: (technical * 0.3) + (dependencies * 0.25) + (data_gravity * 0.25) + (compliance * 0.2)
• Score >4.0: requires dedicated migration architect and extended parallel-run
• Group tightly coupled applications into the same wave

S3: Wave Planning & Sequencing

Migration Factory Model:

A migration factory standardizes repeatable processes across waves. Structure:

Role	Responsibility	Staffing
Factory Manager	Wave scheduling, escalation, metrics	1 per program
Migration Architect	Technical design per application	1 per 10-15 apps
Migration Engineer	Execute rehost/replatform using tooling	2-4 per wave
Test Lead	Validation scripts, acceptance criteria	1 per wave
App Owner (from business)	Requirements, UAT sign-off, cutover approval	1 per application

• Automation backbone: AWS Cloud Migration Factory, Azure Migrate, or custom Terraform + Ansible pipeline.
• Metrics: apps migrated/week, avg cutover duration, rollback rate, post-migration incidents.
• Target throughput: 15-25 apps/month at steady state after Wave 2.

Wave Design:

• Wave 0 (Foundation): Landing zone, connectivity, shared services, monitoring. Duration: 4-6 weeks.
• Wave 1 (Pilot): 2-3 low-risk, low-dependency applications. Duration: 2-4 weeks. Purpose: validate tooling and process.
• Wave 2-N (Production): Grouped by dependency clusters, business domain, or team. Duration: 2-6 weeks each.
• Final Wave: Complex, high-risk, deeply integrated systems.

Pilot Selection Criteria:

• Low business criticality (failure is recoverable)
• Few dependencies (isolated, self-contained)
• Representative technology (validates migration tooling)
• Willing team (champions who provide feedback)
• Measurable success criteria defined upfront

Wave Calendar:

• 1-2 week buffer between waves for retrospective and tooling fixes
• Go/no-go gate before each wave starts
• Parallel streams: infrastructure, application, data, testing, cutover

S4: Landing Zone Design

Account Structure:

• Management: billing, organizational policies, SCPs
• Shared services: logging, monitoring, security tooling, CI/CD
• Network hub: transit gateway, VPN/Direct Connect, DNS
• Workload accounts: per-environment or per-application

Networking:

• Hybrid connectivity: VPN (quick, lower bandwidth) or Direct Connect/ExpressRoute (dedicated)
• Transit Gateway / Azure Virtual WAN: hub-and-spoke inter-account routing
• DNS: split-horizon for gradual cutover
• IP address planning: non-overlapping CIDR ranges with future growth headroom

Security Baseline:

• IAM: federated identity (SSO via SAML/OIDC), service roles, least privilege
• Encryption: at rest (KMS), in transit (TLS 1.2+), key rotation
• Network: security groups, NACLs, WAF, DDoS protection
• Logging: CloudTrail/Activity Log, VPC Flow Logs, centralized SIEM

Governance Guardrails:

• Preventive: SCPs/Azure Policy blocking dangerous actions (public storage, root usage)
• Detective: Config Rules, Security Hub, compliance dashboards
• Tagging: mandatory (owner, environment, cost-center, migration-wave)
• Budget alerts: per-account, per-wave, anomaly detection

S5: Migration Execution Patterns

Rehost Tools: AWS MGN (Application Migration Service), Azure Migrate/Site Recovery, GCP Migrate for Compute Engine. Process: install agent, replicate, test, cutover, decommission.

Database Migration: Homogeneous (native replication, DMS continuous) or heterogeneous (schema conversion + DMS). Zero-downtime via continuous replication. Validate: row counts, checksums, sample comparison.

TCO Calculator Methodology:

Cost Category	On-Premises	Cloud
Compute	Hardware amortization (3-5yr)	Reserved + on-demand instances
Storage	SAN/NAS, disk replacement	S3/EBS tiered pricing
Facilities	Power, cooling, DC lease ($8-15/kW/mo)	Included
Staff	FTE * fully loaded cost	Reduced (managed services)
Licenses	On-prem perpetual or subscription	BYOL, cloud-native, or included
Network	WAN circuits, load balancers	Egress fees, cross-AZ traffic
Migration (one-time)	N/A	Dual-run, tooling, consulting, training

• Break-even: typically 18-36 months. Include hidden costs: egress fees, premium support, cross-AZ traffic.
• Use AWS Pricing Calculator, Azure TCO Calculator, or Google Cloud Pricing Calculator for estimates. Validate with 3-month post-migration actuals.

Cutover Rehearsal Checklist:

Perform a full dress rehearsal 1-2 weeks before each production cutover:

1. Execute complete cutover runbook end-to-end in non-prod environment
2. Measure actual duration vs. planned window (target: <80% of window)
3. Test data sync: verify final replication lag <1 minute
4. Execute DNS switch and verify propagation
5. Run full smoke test suite against cloud endpoints
6. Execute rollback procedure -- restore source, revert DNS, verify source healthy
7. Validate monitoring/alerting fires correctly in cloud environment
8. Time the rollback procedure (target: <30 minutes)
9. Document gaps and update runbook before production cutover
10. Obtain sign-off from app owner and operations team

Rollback Decision Criteria:

Trigger rollback if ANY of the following occur during cutover:

Condition	Threshold	Action
Error rate spike	>5x baseline for 15+ minutes	Immediate rollback
Latency degradation	P99 >3x baseline for 15+ minutes	Immediate rollback
Data integrity failure	Any checksum mismatch on critical tables	Immediate rollback
Functionality gap	Any Tier 1 feature non-functional	Immediate rollback
Cutover window exceeded	>80% of planned window with steps remaining	Assess and likely rollback
Monitoring blind spot	Key dashboards/alerts not functioning	Pause and remediate; rollback if >30 min

• Keep source running during validation period (1-2 weeks minimum).
• Reverse replication from cloud to on-prem for data-heavy workloads.
• Rollback window: define per application (typically 24-72 hours post-cutover).

Common Anti-Patterns:

Anti-Pattern	Consequence	Mitigation
Skip discovery; migrate by guesswork	Missed dependencies cause outages during cutover	Run automated discovery 30+ days
Migrate without application owner	No UAT, no sign-off, unclear accountability	Require named owner before wave assignment
No cutover rehearsal	Surprises on go-live night, panic rollbacks	Mandatory dress rehearsal for every wave
Big-bang migration of entire portfolio	Maximum blast radius, no learning curve	Wave-based with pilot first
Underestimate data transfer time	Cutover window exceeded, forced rollback	Calculate transfer time + 50% buffer; use continuous replication
Retire nothing	Cloud costs add to existing costs	Enforce retire/retain classification in 7R assessment
No rollback plan	Cannot revert when issues found	Define rollback triggers and test procedure

S6: Validation & Optimization

Parallel-Run Validation:

• Run source and target simultaneously for 1-4 weeks
• Compare: response codes, data checksums, latency percentiles
• Escalation gates: <1% discrepancy = proceed; 1-5% = investigate; >5% = halt

Functional Validation: Automated test suites, manual smoke testing of critical flows, integration testing with dependent systems, data validation.

Performance Baseline: Compare cloud vs. on-premises for latency, throughput, resource utilization. Document regressions.

Cost Optimization (post-migration):

• Right-size after 2-4 weeks of utilization data
• Reserved instances / savings plans for steady-state
• Spot for batch. Storage tiering. Delete orphaned resources.
• Target: 15-30% cost reduction within 90 days of migration completion.

Decommission:

• Validation period complete and signed off
• Archive required data per retention policy
• Reclaim licenses. Secure hardware disposal.
• Update architecture diagrams and CMDB

---

Trade-off Matrix

Decision	Enables	Constrains	When to Use
Rehost	Speed, low risk	No modernization	Datacenter exit deadline
Replatform	Some cloud benefit, moderate effort	Partial optimization	Managed DB, container runtime swaps
Refactor	Full cloud-native benefit	High effort, long timeline	Strategic, 5+ year lifecycle apps
Migration Factory	Repeatable, metrics-driven, scalable	Setup overhead, process rigidity	>20 applications, enterprise programs
Big Bang	Clean cutover, no hybrid period	High risk, long outage	Small portfolios (<10 apps), maintenance windows
Wave-Based	Incremental risk, learning curve	Hybrid period, dual costs	Large portfolios, enterprise migrations
Direct Connect	High bandwidth, low latency	Cost, 4-8 week lead time	Large data (>5TB), long-term hybrid

---

Assumptions

• Target cloud platform selected (AWS, Azure, GCP, or multi-cloud)
• Application inventory exists or can be discovered
• Budget approved for migration (dual-run costs included)
• Team has or will develop cloud operations skills
• Business stakeholders available for cutover scheduling and sign-off

Limits

• Does not design cloud-native architecture for refactored applications (use cloud-native-architecture skill)
• Does not design infrastructure platform from scratch (use infrastructure-architecture skill)
• Does not assess current state without migration context (use asis-analysis skill)
• Organizational change management (training, team restructuring) acknowledged but not deeply covered

---

Edge Cases

Datacenter Exit with Hard Deadline: Favor rehost for speed. Accept technical debt. Plan post-migration optimization waves. Prioritize by lease expiry. [EXPLICIT]

Shared Database Serving Multiple Applications: Cannot migrate database independently. Options: migrate all consumers together, introduce API layer to decouple, or replicate and gradually cut over consumers. [EXPLICIT]

Mainframe Workloads: Specialized tools (Micro Focus, AWS Mainframe Modernization, Azure Mainframe Migration). Replatform to cloud-hosted emulation first, then gradually refactor. [EXPLICIT]

Compliance-Restricted Data: Data residency may limit regions. Encryption requirements affect replication tooling. Audit trail must be maintained through migration. [EXPLICIT]

No Source Code Available: Rehost is the only viable strategy. Containerization may be possible for binary apps. Evaluate repurchase with SaaS alternative. [EXPLICIT]

---

Validation Gate

Before finalizing delivery, verify:

• Every application has a 7R classification with documented rationale
• Automated discovery tool deployed and ran for 30+ days
• Dependency graph covers all inter-application and data dependencies
• TCO comparison completed (on-prem vs. cloud, including migration costs)
• Wave plan sequences applications by dependency and risk
• Migration factory roles and automation defined
• Landing zone meets security, networking, and governance requirements
• Cutover rehearsal checklist completed for pilot wave
• Rollback decision criteria defined with specific thresholds
• Performance baselines captured for pre/post comparison
• Cost optimization plan ready for post-migration execution

---

Output Format Protocol

Format	Default	Description
markdown	Yes	Rich Markdown + Mermaid diagrams. Token-efficient.
html	On demand	Branded HTML (Design System). Visual impact.
dual	On demand	Both formats.

Default output is Markdown with embedded Mermaid diagrams. HTML generation requires explicit {FORMATO}=html parameter. [EXPLICIT]

Output Artifact

Primary: A-01_Cloud_Migration_Plan.html -- Executive summary, 7R classification table, dependency map, wave plan, landing zone design, execution playbook, validation checklist, cost optimization targets.

Secondary: Application inventory spreadsheet, wave calendar, cutover runbook, rollback procedures, TCO comparison, decommission checklist.

---

Author: Javier Montano | Last updated: March 18, 2026