Replatform the code and the data together

Methodology

Six structured phases

Each phase produces reviewable artifacts, so domain experts stay in control without writing every line of code.

1. 01

   Discovery

   • Map the existing application surface, data flows and access patterns
   • Identify every consumer of the database, including the forgotten ones
   • Produce a reviewable picture of what is actually there
2. 02

   DAL isolation and tests

   • Carve the data access layer out of the application
   • Lock current behavior in with tests before anything changes
   • Give yourself a regression net for every later phase
3. 03

   Migration assessment

   • Identify gaps, risks and the right cloud target for the workload
   • Size the effort against real query patterns rather than table counts
   • Decide what moves, what waits and what gets retired
4. 04

   Schema design

   • Empirically validated schemas, designed against real query patterns
   • Shaped by how the application reads, not by the old table layout
   • Reviewed as an artifact before any code depends on it
5. 05

   New parallel DAL

   • Implement the new data access layer alongside the legacy one
   • Run both against the same traffic until the new one is trusted
   • Switch consumers over deliberately rather than all at once
6. 06

   Data migration

   • Move the data with Dsync, in step with the code
   • Verify both sides agree before the cutover commits
   • Minimum downtime, with a back-out that reverses direction