Data centre migration plan: what to settle before moving day

November 13 2026, by Macquarie Technology Group | Category: Data Centres
Data Centre Parallax image

Moving the equipment is only one part of a data centre migration. The harder question is whether everything will work together when it is switched back on. Before you choose a move weekend, map the dependencies, prove the destination is ready and agree how you’ll recover if a cutover does not go to plan.


This guide takes you through seven phases, from discovery to the support period after the move. It covers a move from on-premises infrastructure, another facility or a consolidation into colocation. You’ll find the decisions to make at each stage, a risk register and a responsibility matrix to help turn the plan into work people can carry out.


What is a data centre migration plan?


A data centre migration plan is a structured strategy for moving IT infrastructure, applications and workloads from one facility to another while minimising downtime, operational disruption and data loss. It covers key stages including asset discovery, dependency mapping, target infrastructure design, testing, phased migration, cutover and post-migration support.


Find the dependencies before they find you


A move plan can account for every cabinet and still miss a service that the business relies on. Physical handling needs careful planning, but so do the connections between systems and the sequence in which they recover. Both belong in the project from the start.


Look for the details a rack inventory cannot tell you: an undocumented server supporting a critical application, a hard-coded IP address, a licence tied to hardware or a firewall rule needed by an old integration. Use traffic analysis and conversations with application owners to find them, including services that only run at particular times of the month.


Leave contingency within each wave. A delayed network change or a failed hardware check can consume the next team’s window if the schedule assumes every step finishes exactly on time. Agree what can slip, what must stop and how much time is needed to recover safely.


The seven phases of a data centre migration plan


Give each phase an exit condition and an owner. The date in the schedule is a target; the evidence that the work is complete is what allows the next phase to proceed. Our data centre relocation guide provides additional context for the physical move and its logistics.


Phase 1: Discovery and dependency mapping


Build the physical asset inventory and map logical dependencies. Use approved discovery tools and traffic observation alongside documentation and application-owner interviews. Allow enough observation time to capture scheduled jobs, third-party integrations and infrequent processes. Include hardware-bound licences and physical dongles in the record.


Have application owners review the dependency map and agree how remaining gaps will be resolved. If nobody can confirm what a critical application needs, keep that question open in the plan rather than treating a missing answer as evidence that there is no dependency.


Phase 2: Target design


Settle the target rack elevations, usable power and redundancy, cooling, cabling and IP addressing. Confirm network paths to other sites and cloud services, including the temporary connections needed during migration. Decide what will move as-is, what will be refreshed or consolidated and what can be retired. Check the support and data-retention implications before decommissioning anything.


Phase 3: Build and validation


Work with the provider to validate the target before production arrives. Confirm power, cooling, cabling and access are ready for the agreed load; establish cross connects and test the required paths end to end. Verify that staff and contractors can obtain the access they need, including out of hours. Record which tests are complete and which depend on equipment being installed.


Phase 4: Pilot move group


Choose a small group that is low enough risk to be a sensible pilot but representative of the later moves. Record the actual time and issues at each step. Use the results to adjust the runbook and schedule, and repeat the pilot where unresolved problems would put production waves at risk.


Phase 5: Waved production moves


Group workloads by dependency, then assess whether they should move together or through a tested interim arrangement. A split cluster may introduce new latency, bandwidth or availability constraints between sites. Design and test that temporary state explicitly. Where a phased move is impractical, document why and how the larger cutover will be controlled.


Phase 6: Cutover and verification


Give each cutover a runbook with timed steps, named owners and go/no-go decisions. Define success in business terms, such as an end-to-end transaction completed and accepted by the application owner. Rehearse the recovery route as far as practicable. For a physical move, returning hardware may take longer than the available window, so recovery could require a replica, standby system or other agreed fallback. Set a decision deadline and account for data changes that could make reversal unsafe.


Phase 7: Hypercare


Keep additional support in place through the agreed verification period after each wave. Include the relevant business peaks, scheduled jobs and batch cycles in the acceptance plan. Some cycles may need simulation or a later monitored check rather than keeping the full project team indefinitely. Agree the handover conditions, escalation contacts and ownership of any remaining issues.


The risk register


Use the table as a starting risk register, not a set of measured probabilities. The ratings are illustrative and need to be reassessed for your environment. Assign owners and review the risks at each wave gate, including the feasibility and time needed for recovery.

 

Risk Likelihood Impact Mitigation Owner
Undocumented dependency discovered at cutover High High Automated traffic-based discovery plus signed application owner review Application owners
Cross connect lead time exceeds the move window Medium High Order early with written lead times, and test paths before the wave Network lead
Hardware fails on power-up at the destination Medium Medium Verified restore capability, spares and a feasible recovery route for the wave Infrastructure lead
Licence bound to hardware or MAC address fails Medium High Licence audit during discovery, with vendor rehosting arranged in advance Application owners
Access not provisioned for staff or contractors Medium Medium Submit access requests early and physically test entry before the wave Programme manager
Dependency cluster split across waves Medium High Wave grouping by dependency cluster, with interim states designed and tested Architecture lead
Equipment damaged or lost in transit Low High Specialist movers, chain of custody, insurance, and verified backups Programme manager
Rollback needed but never tested Low Critical Rehearse recovery as practicable; confirm data consistency, route and decision deadline Infrastructure lead
Issue emerges at first month-end or peak load Medium Medium Verify relevant business cycles through observation or simulation, with an agreed handover plan Service owner

 

Roles and responsibilities in a data centre migration


Agree who can make each decision before the pilot. The matrix below is a starting allocation; tailor it to your governance and name the actual people. Give the cutover lead clear authority and escalation rules, with backups for critical roles and confirmed availability during the move window.

 

Role Primary responsibility Critical decision they own
Programme manager Schedule, wave sequencing, logistics and overall coordination Coordinate the go/no-go gate with technical and business approvers
Architecture lead Target design, dependency clustering and interim state design Whether a dependency cluster may be split across waves
Infrastructure lead Physical build, racking, power, hardware and rollback readiness Assess recovery readiness and recommend or invoke rollback under the agreed authority
Network lead Cross connects, IP addressing, firewall rules and connectivity testing Whether network paths are verified and ready for cutover
Application owners Dependency accuracy, licence audit and functional verification Whether their application is verified as working after cutover
Security and compliance Access authorisation, data handling in transit and secure decommissioning Whether data handling and destruction meet obligations
Provider engagement lead Dock bookings, on-site support scheduling and provider escalation When to escalate to the provider during a move window
Business service owner Business verification, user communication and hypercare acceptance Whether the business accepts the service as restored

 

What to expect from your provider


Agree the provider’s scope early: what it will do, what you will arrange and what each activity costs. Confirm cross-connect provisioning, delivery appointments, lift capacity and dimensions, storage, out-of-hours access and contractor authorisation. Make sure those arrangements fit the chosen facility and deployment space, rather than relying on a generic service description.


Ask what on-site engineering support is available during each move window, which tasks are covered and what response commitments apply. Confirm whether the team can receive deliveries, rack equipment and patch connections, and identify any specialist work requiring another supplier. Have the escalation contacts and charges agreed before the first shipment.


Data centre migration planning checklist


Confirm each item before committing to a move date.


• Discovery completed using appropriate tools, traffic evidence, documentation and owner interviews
• Dependency map reviewed and signed by every application owner
• Hard-coded addresses, hardware-bound licences and physical dongles identified
• Assets categorised as move, refresh, consolidate or retire
• Rack elevations, power per cabinet and redundancy configuration designed and agreed
• Cabling schedule and labelling standard defined
• Cross connects ordered, with lead times confirmed in writing
• Target environment validated for the agreed power, cooling and network requirements, with remaining tests recorded
• Access provisioned and tested for all staff and contractors, including out of hours
• Pilot group selected as representative rather than merely convenient
• Pilot timings recorded and used to correct subsequent wave estimates
• Waves grouped by dependency cluster, with any split clusters planned and tested
• Runbook per wave with timed steps, named owners and go or no-go points
• Recovery or rollback procedure rehearsed as practicable, with feasibility, data integrity and decision deadlines agreed
• Verification defined in business terms and confirmed by actual users
• Hypercare, business-cycle verification and handover criteria agreed, with escalation contacts on both sides
• Dock booking, lift dimensions and delivery storage arrangements confirmed
• Insurance and chain of custody arranged for equipment in transit
• Decommissioning and secure data destruction planned for the origin site


Before the move, walk the runbook with the people who will use it. They should know what success looks like, who can stop the cutover and how the service will be recovered. Resolve the unclear hand-offs then, while there is still time to change the plan.


Looking for a data centre partner that can support your infrastructure today and as your requirements evolve? Macquarie Data Centres provides secure, resilient and scalable data centre solutions for hyperscale, enterprise, government and cloud environments.


Contact us to discuss your requirements, learn more about our team, or request a tour to see our facilities firsthand.


Frequently asked questions


What should be in a data centre migration plan?


Include discovery, target design, build and validation, a pilot, production move waves, cutover verification and hypercare. Give each phase an owner and exit criteria. Each wave needs a runbook, recovery route, decision deadlines and business-level verification, with physical logistics and security addressed alongside application dependencies.


How long does a data centre migration take?


The schedule depends on dependencies, connectivity lead times, build work, data transfer and approval windows. Build the plan from those tasks rather than choosing a duration first. Allow contingency within each wave and include the support and verification needed after cutover.


What causes data centre migrations to fail?


Unidentified dependencies, an unready target, hardware problems, access issues and an unworkable recovery plan can all disrupt a move. Discovery, validation and a representative pilot help expose these risks before a critical cutover. Leave time to resolve the findings.


Should I move everything at once or in waves?


Waves can limit the impact of an individual cutover and let the team apply lessons from the pilot. They need to follow dependency groups, with any interim connections tested. Some tightly coupled environments may require a larger move; assess the trade-off and recovery plan rather than applying a blanket rule.


What should I ask a colocation provider about migration support?


Ask who provisions cross connects and the lead time, dock booking and lift dimensions, whether deliveries can be accepted and stored ahead of the move, how out-of-hours access is authorised for staff and contractors, whether on-site engineers are directly employed, and whether they assist with racking and cabling or expect you to supply that labour. Get charges confirmed in writing.


When should rollback be invoked?


Use pre-agreed failure conditions and the latest safe decision time, with authority assigned to a named lead. Include the time needed to recover and any data-consistency constraints. Some situations require an earlier stop; others may need a different recovery route because physical reversal is no longer feasible.