Cloud Computing Architecture Guide

Migrating a Portfolio While Changing the Operating Model

Goal

Apply the chapter's model to a 200-application portfolio while increasing migration throughput only as the destination organization can absorb it.

The Short Version

Large migration programs coordinate three workstreams: destination foundations, workload waves, and operating capability. If workload movement outruns governance, security, support, recovery, and cost practices, the program converts data-center risk into cloud risk. If the foundation is perfected before any workload moves, learning arrives too late.

Build a thin production-ready baseline, test it with representative pilots, and expand both platform capability and wave throughput using evidence.

Synchronized Change

Foundation and controls: baseline -> expand -> optimize
Portfolio waves:          pilot -> wave 1 -> wave 2 -> wave n
Operating capability:     train -> co-own -> workload ownership

Segment and Sequence the Portfolio

For 200 applications, segment by business value, deadline, criticality, dependency group, lifecycle, technical condition, migration method, evidence confidence, and owner readiness. Identify obvious retire and replace candidates early. Keep tightly coupled applications and data together unless the program deliberately builds a temporary bridge.

Use reusable patterns for common moves: standard landing-zone placement, connectivity, identity, telemetry, backup, deployment, and acceptance evidence. A migration factory improves consistency, but it needs an exception path for workloads whose data, regulation, latency, or vendor constraints genuinely differ. Exceptions should produce learning, not hidden one-off work.

Grow the Operating Model

Develop platform product ownership, workload ownership, service management, security response, financial operations, support access, and recovery capability alongside the technology. Central teams should provide paved paths and guardrails; workload teams should own service objectives, application behavior, dependency relationships, and business validation.

Use co-ownership during early waves. Platform specialists and workload teams operate pilots together, capture gaps, and transfer responsibility through exercises rather than documents alone. Increase throughput only when alert response, deployment success, restore tests, policy compliance, cost allocation, and stabilization times show sufficient capacity.

Example: A 200-Application Exit

A company has an eighteen-month data-center deadline, incomplete dependency evidence, and a mandate to avoid unnecessary rewrites. It first retires unused systems and selects pilots representing a standard web application, a batch workload, and a database-heavy service. The baseline landing zone supports controlled access, hybrid connectivity, logging, cost allocation, deployment, and backup.

Early waves use rehost and targeted replatform patterns. High-change refactors proceed separately where product value justifies them. A weekly governance forum resolves evidence gaps and exceptions; monthly measures cover stable operation, decommissioned assets, risk reduction, migration lead time, and unit cost. Legacy network paths, licenses, contracts, and support queues have named retirement dates.

Program Exit

Completion requires more than the last cutover. Define ownership transfer, operational acceptance, archive and retention, asset disposal, contract closure, access removal, network shutdown, and benefit review. Temporary coexistence should have an expiry so two operating models do not become permanent.

Common Mistakes

  • Scaling wave volume before support, security, and recovery capacity are ready.

  • Building a rigid factory that cannot learn or handle justified exceptions.

  • Leaving legacy contracts, paths, and ownership active indefinitely.

  • Measuring movement while ignoring stable operations and realized outcomes.

If Someone Asks You

I would segment the portfolio, establish a thin production-ready foundation, and use representative pilots to build evidence. I would standardize common migration patterns, isolate justified modernization, and increase wave throughput only as governance, support, recovery, and workload ownership prove they can absorb it.

What You Should Remember

Increase migration speed only as the destination platform and operating organization earn the capacity to absorb it.

Questions to Check Your Understanding

  1. Which dimensions should segment a large portfolio?

  2. What evidence should gate an increase in throughput?

  3. Why must legacy retirement be part of program exit?

Further Study