A Data Migration Journey (Part 1)
I’m certain that even Will Shakespeare, the Bard
Would deem that composing this poem is hard.
But I’m feeling brave, I’ll give it a go
And fashion these verses; like data, they’ll flow.
You’re setting off on a data migration,
Flying data to an exotic location.
Millions of records are being migrated;
Let’s hope hardly any will be complicated.
Big Bang or Phased, what approach will you choose?
Or Dual-running maybe? it’ll drive you to booze.
Pre-load historic? But be on your guard;
Extracting just deltas is frightfully hard.
Imagine the scene, if there were a disaster?
Fix going forward (a strong sticking plaster)?
If feasible, please put a rollback in place
To let the migration be backed out with grace.
A problem demanding your early attention:
The current solution data’s retention.
What’s to be done with the data that was?
Delete it? Or keep it and archive to BLOBs?
Transactions can give some migrators a fright:
Those started but not yet complete, still in-flight.
Identify early, as soon as you’re able.
Design their solution - transition made stable!
And data discovery can’t be ignored;
To find all locations where data is stored.
Forage for places where data could be.
You’re sure to spot data you didn’t foresee.
Returning to basics, it’s just bits and bytes
That make up the data you’re moving ‘tween sites.
But migrations are not just for techies, you see,
You must understand it, hence BAs, like me!
When moving your data between diverse disks,
What bad things can happen? Identify risks.
Take action on those with a decent risk rating;
Their ‘impact’ and ‘likelihood’ need mitigating.
So how to transfer from site A to site B?
Plonk servers in vans? Or SFTP?
The option selected, it must be secure,
Numerous stakeholders you must reassure.
Mapping of data will help you discover
How one system’s data aligns with the other.
Fields in the target the source can’t supply?
Or data from sources the ‘load’ can’t apply?
Finding those mis-matches has to be done;
It causes more work, but it’s part of the fun!
Some data you need, you’ll have to transform;
The new world rules, you must make it conform.
Unstructured data may have some pitfalls:
Documents, images, files and phone calls.
You’ll find old object types your team’s not seen before,
(Types from computing’s old Days of Yore!)
If your data quality’s not as expected,
Choose the best time when it can be corrected.
Before extraction? Or during migration?
Or after receipt at the end destination?
(continued in Part 2)
©️ Ray Cohen, May 2023
