Regulations and compliance · Practical guide

Moving a GxP data centre: what needs to be demonstrated before cutover?

An application can remain unchanged when its data centre moves. New network paths, identity services, backups and operational responsibilities may still affect the GMP process. Define the migration scope for an application used in human medicines manufacturing, identify existing evidence you can reuse and close outstanding questions before switchover.

Table of contents

At a glance

Do we need to revalidate the whole system when we change data centres?

Not automatically. Assess which infrastructure, data flows, controls and operating conditions change and how they affect the intended GMP use. Document the baseline and test the affected functions and relevant unchanged dependencies. Previous evidence can remain applicable if its underlying conditions still hold. Completing the technical move does not demonstrate data completeness or suitability of the new operation.

  • Identical application software does not imply identical operating conditions.
  • Data transfer, identities, and reprocessing require specific acceptance criteria.
  • The switch decision must take into account open findings and provide a feasible rollback option.

1. Distinguish the migration from the GMP process

Source date is 7 October 2026. The basis is Annex 11 as amended in January 2011 for computerised systems in the GMP context. Its principle distinguishes between validated applications and qualified IT infrastructure. The 2025 consultation draft is not treated as a current additional requirement.

First, describe what is actually being migrated: virtual machines, databases, storage, network connections, or the entire application environment. A switch to a different cloud provider may have different implications than a change in location with the current provider. Simply naming data centers is insufficient as a change description.

The specific subject of inspection is, for example, the continued correct assignment of a laboratory result to an approved test request. The technical components underlying this process are derived accordingly. Not every enterprise application is solely due to its location considered GMP-relevant.

Sources: [1] [2]

2. Identify changes and dependencies that remain unchanged

A change map links the technical transition with its potential failure consequences. Document what remains unchanged and provide supporting evidence for such stability. For example, an unchanged program version does not justify accepting unverified connectivity to a newly implemented user management system.

The following assignment matrix helps distribute responsibilities among application owners, IT, operators, and quality assurance. It is not a prescribed form structure. In accordance with Annex 11, all changes—including configuration—must be controlled; the review effort shall follow the documented impact.

Change Management Matrix for a Data Center Transition
ChangePotential impactRelevant evidence
New network and interface pathways Orders are missing, processed twice or delayed. Defined transmission with reception, assignment and error handling
Other identity binding Unauthorized rights or locked privileged users Test the role with permitted and unauthorised actions
New time stamp Events may be incorrectly timed. Time basis and presentation of related events
New Security and Recovery Application starts, required data are still missing Restoration with technically readable datasets
New operational responsibility The disturbance is technically resolved; GMP implications remain open. Agreed Notification, Assessment, and Decision

Sources: [1]

3. Purposefully reuse existing evidence

Neither carry over the entire previous validation nor automatically start from zero. Map each relevant requirement to evidence that remains applicable and to the new tests. Evidence for a calculation may still apply unchanged. A restore test for a retired backup procedure usually does not.

Supplier documentation may cover a specific aspect if version, configuration, environment, and testing scope align with migration requirements. A vendor's statement that their data center is certified does not alone confirm that task-specific data restoration functions as intended.

Gaps remain visible: for example, an unverified interface compatibility or a modified database configuration. Decide on additional checks based on the potential impact of errors. Missing information is not evidence of minimal impact.

Sources: [1] [3]

4. Verify the data transfer at the actual switching point

The following example describes an application with eight released activities in the output export. During the pre-copy phase, two additional activities are generated. The relocation plan specifies that these uniquely identified follow-up activities shall be taken over prior to the productive start. This is expected to result in ten distinct activities with their respective dependencies.

A simple quantity comparison is insufficient. Ten target lines may arise from nine distinct processes and one duplicate. Therefore, the reconciliation includes identifiers, technical values, status, and required linkages. During data transfer, Annex 11 further requires verification that the value and meaning remain unchanged.

Define in advance which processing activities are paused during the transition, how intermediate messages are detected, and who evaluates a non-closing reconciliation. A reset must not inadvertently lose business transactions that have occurred during the interim period.

Example: Comparison of eight starting batches and two follow-up batches
Assessment pointDefined expected resultFind and Decision
Identity V01–V08 inclusive, plus V09 and V10 each once Counting only ten lines is not sufficient
Duplicate No second instance of V08 V08 appears twice and V10 is missing. Reconciliation fails.
Status The approval remains linked to the relevant transaction Same number of statuses upon reversed assignment is insufficient
Relationship Every procedure maintains its required result-oriented alignment. Existing main statements without result file are incomplete
Late arrivals V09 and V10 controlled and accepted Resolve the missing later transaction before the intended use

Sources: [1]

5. Plan jointly the switch-over and the rollback

A rollback is only viable when its conditions and boundaries are known. Define up to which step the previous environment can be retained and specify how new data is handled. Two parallel, concurrently operational environments may generate conflicting records; intentionally prevent or control this state.

Verify the restart procedure in accordance with its significance for the process. Annex 11 requires appropriate provisions for system outages and adequate documented testing for critical processes. Therefore, a universal zero-downtime target and a uniform restart procedure for all systems cannot be assumed.

The decision includes technical findings and technical usability. A successful application does not answer whether open quality decisions, relevant history and blocked stocks are still recognisable.

Sources: [1]

6. Clarify operational responsibility prior to site transfer

Check whether the data center also involves changes to contractual partners or sub-contractors. Annex 11 requires clear agreements for relevant third-party services. In the case of outsourced GMP activities, Chapter 7, with its responsibility and approval rules, must also be considered.

Assign responsibilities for monitoring, security, troubleshooting, changes, data availability, and system recovery. Providing a phone number alone does not constitute a complete operational agreement. For the user, it must be clearly identifiable who investigates the technical root cause and who evaluates the impact on GMP records and the process.

A specific country or certification does not replace this assessment. Data protection, information security, and any potential contractual site requirements must be independently evaluated; solely from Annex 11 it does not follow that every data center must be located in Germany.

Sources: [1] [4]

7. Release checklist for the new operation

Summarize the closure in such a way that an authorized person can recognize the remaining risks and the scope of the decision. A schedule or a general notification "migration successfully completed" is insufficient for this purpose. The checklist below serves as a practical working aid.

  • Clear description of source and destination environments as well as affected GMP processes.
  • Evaluates changes and relevant unchanged dependencies.
  • Supplier and customer validation records have been reviewed for compatibility.
  • Data reconciliation including required relationships and follow-ups completed.
  • Permissions, interfaces, and recovery procedures have been assessed according to their impact.
  • Open findings with implications, responsible party, and justified decision documented.
  • Switching point, return path, and handling of newly generated data aligned.
  • Operational documentation and responsible contacts have been transferred to the target site.

Sources: [1] [3]

8. What a SME should prepare

The largest avoidable effort typically arises from an unclear starting point: no one knows all interfaces, feedback loops, or storage locations. Therefore, prior to technical scheduling, compile a system description, role matrix, backup strategy, and the most critical process documentation.

The IT department explains the technical difference, the process owner defines the required functionality, and quality assurance emphasizes the importance of open findings for the intended GMP use. This collaboration represents a sensible organizational approach; it is not a universally mandated body with fixed role designations.

Following the relocation, a targeted monitoring approach remains advisable: Are agreed-upon monitoring activities active, and do bottlenecks occur, providing the first tangible evidence of system performance? The scope is adjusted in line with changes. A successful startup does not relieve ongoing responsibilities for maintaining controlled operations.

Sources: [1] [3]

Project experience and collaboration

Test the data centre migration against the process that relies on it

During a data centre move, IT and the process owners trace a known transaction through the new infrastructure. We check connectivity, timekeeping, permissions, backup and correct processing together. Existing evidence may remain applicable where the application stays the same. Changed dependencies need specific checks. We plan the handover and reconcile the final dataset. The evidence must show the required process working at the new site, beyond confirming that the server responds.

Frequently asked questions about Moving a data centre

Is a change of data center automatically considered a complete revalidation?

No. Assess the impact on intended use, infrastructure, data, and operations. Existing documentation remains valid provided its underlying assumptions continue to hold. Modified or affected unchanged functions require a clearly defined and traceable evaluation or review scope.

Sources: [1] [3]

Is it sufficient for the software version to remain unchanged?

No. Changes to database settings, network paths, identity services or backups can affect the same software differently. Document the target configuration and link the differences to the processes they affect.

Sources: [1]

Do we need to manually verify each data point individually?

Annex 11 does not prescribe a universal manual individual check. Select an appropriate and justified reconciliation approach, for example, automated identity and value verification combined with professional review of key relationships. The scope must address the actual receiving risks.

Sources: [1] [3]

Is a copy of the virtual machine already sufficient as complete migration evidence?

It can support technical adoption. In addition, external dependencies such as user management, interfaces, time supply and backup are to be considered. Furthermore, prove that the application fulfils its intended tasks in the target environment.

Sources: [1]

Which data are most easily generated between relocation steps?

Examples include new orders, processed results, and messages in queues. Define a temporal and subject-matter handover point. Document which procedures are retrospectively accepted or temporarily withheld, and how their completeness is verified.

Sources: [1]

Can the old environment be immediately shut down?

Only after the further access to required data and records has been addressed and the agreed return path is no longer needed, should a system shutdown be decided. The decision to deactivate must be based on data retention, pending investigation findings, and ongoing operational requirements, not solely on the successful completion of the login procedure.

Sources: [1]

Who is authorized to release the new operation?

Responsibility must be defined within your quality system and change control procedure. Technical operability and professional usability must be demonstrably assessed. Annex 11 does not require a single, identical list of signatures for all companies.

Sources: [1]

Is a certified data center sufficient?

An appropriate certificate may support a specific scope of validation. It does not automatically demonstrate correct configuration, data transfer, or application recovery. Review the exact scope of applicability and supplement any missing task-specific validation evidence.

Sources: [1] [4]

Do we need a test of the return path?

Verify the return path in accordance with its role in the transfer and maintenance of critical processes. Consider, in particular, data that has been generated since the process began. A documented return path lacking realistic assumptions cannot ensure the intended level of safeguarding.

Sources: [1]

What documents should we provide to the service provider prior to submitting the proposal?

Useful elements include system description, data volume and data types, interfaces, user roles, critical operating times, and backup and retention requirements. This enables the supplier to describe its actual capabilities and makes outstanding evidence visible prior to scheduling.

Sources: [1] [4]

Sources and review date

Source date: 7 October 2026. Human medicinal products – GMP; transfer of a system used for this purpose. Data protection and national site-specific requirements are not finalised. Tables and case studies are internal work aids and do not represent completed product testing. Draft versions do not establish enforceable additional obligations.

  1. European Commission: EU GMP Annex 11, Computerised SystemsJanuary 2011 revision, applicable from 30 June 2011. Principle and sections 1, 3, 4.2–4.3, 4.8, 7, 10–13 and 16–17.
  2. European Commission: EudraLex Volume 4Current Directory, retrieved 07.10.2026; Human medicines: Annex 11, Revision January 2011
  3. EMA/ICH: Q9(R1), Quality Risk ManagementEMA/CHMP/ICH/24235/2006 Corr.2; Publication date 31.01.2025; Q9(R1) effective from 26.07.2023; Sections 3, 4.3–4.6 and 5.1: risk-based decision-making, uncertainty and formality
  4. EU-GMP Chapter 7: Outsourced ActivitiesRevision 1, applicable from 31 January 2013. Sections 7.4–7.8, 7.11 and 7.14–7.17 address responsibility, subcontracting, contracts and evidence.

Sources checked: .

Daniel Herrmann, Daniel Herrmann Consulting

Daniel Herrmann Consulting

Link the data center transition to the GMP process

We support you in translating technical changes, reusable justifications, and unresolved technical decisions into a actionable change plan. The appropriate scope begins with your system and the actual intended transition.

  • Computer system validation →

    Bring your GxP systems into validated operation, from laboratory software, MES and eQMS to ERP. We define the scope, plan risk-based tests using GAMP 5 and hand the evidence to QA for release.

Send a project enquiry Customer testimonials →