Start with the engine and access pattern.
Provide the database engine, supported release, extensions and application compatibility requirements. Include current size, growth, connection counts and periods of heavy activity. Describe whether the workload is transactional, analytical or a mixture, and share representative slow operations where available.
CPU, memory and storage affect different parts of database behaviour. Working datasets, indexes, query plans and application connection patterns can matter more than a larger nominal server. Agree what evidence is needed to compare a proposed environment meaningfully.
A database within a shared engine, a separate database instance and a dedicated physical server provide different boundaries. Compare access separation, resource contention, tuning freedom and maintenance control. A dedicated allocation is most useful when those requirements justify its cost and operating responsibility.
Define the protection and recovery requirements.
State how much recent data the business can lose and how long recovery can take. Choose a backup and logging approach that supports those objectives for the selected engine. Include consistency, encryption, retention and restore authorisation in the plan.
Replication and failover need their own design decisions about consistency, lag and application behaviour. A replicated change can include an accidental deletion, so historical recovery remains a separate concern. Restoration should end with application validation and checks on important records.
Agree the management boundary.
Limit database access to the applications and administrators that require it. Separate application credentials from privileged maintenance access, and define how credentials are rotated. Include network restrictions, audit needs and the handling of diagnostic data.
Managed database work may include engine updates, monitoring, backup checks or performance review, depending on the agreement. Schema design, query changes and application debugging need explicit owners. Clarify who approves maintenance and how compatibility is tested before a release change.
Move data with a validation plan.
For a migration, identify the source size, write rate, version differences and available cutover window. Decide whether a test restore, replication-assisted move or another supported approach fits the workload. Include a rollback decision before the final switch.
Acceptance should cover record checks, important queries, application transactions, backup execution and restoration. The quotation should distinguish infrastructure, engine licences, administration, migration and any specialist optimisation. This gives both application and operations teams a clear basis for service ownership.
