Field note
Two 35TB migrations
Same volume, one weekend versus three months and three aborted attempts. The variable was the data model.
Two migrations, same size, opposite outcomes.
The first: a Friday morning note telling me that 35TB of call detail record data I had been expecting the following month had arrived early. Around 20,000 compressed files, well formatted, sensibly named. We had no load process at all. I spent the day writing shell and Perl — this was a while ago — that generated SQL to load the files in parallel, straight into the database without uncompressing them first. I started the job at 4pm on the Friday. I came in on Monday and the whole 35TB was loaded and usable.
The second: a 35TB reporting system, a few hundred tables, migrated with the standard recommended toolset. Three months and three aborted attempts before the database was finally across.
Identical volume. One completed over a weekend with a script written in an afternoon; the other took a quarter of a year with proper tooling and experienced people. The difference was not size, skill, budget or product. It was the data model.
Where this feeds the guide: Complexity, not volume and Nothing survives twelve hours.