Migration to the cloud is in full flight. The most advanced problem concerned in finishing these migrations shouldn’t be shifting the information however migrating the information processing code to work on new infrastructure within the cloud.
The basic problem is to create code that performs the identical enterprise course of or returns the identical consequence on the brand new platform, fairly than merely making the outdated code run on the brand new platform. This historically includes a protracted, guide technique of copying the information, changing the code, testing the code and verifying that the migrated code has the identical habits as the unique code.
Philosophically there are three totally different migration approaches – listed in rising order of danger:
1. Lift and shift
- Move the prevailing code performance to the cloud
2. Lift, modify and shift
- Complete some code redesign in the course of the migration
3. Total redesign
- Rebuild the whole lot from the bottom up
Whatever method you are taking, take into account the Five Hidden Costs in code migration:
Hidden Cost #1 – Underestimating the scope of the problem
Migration tasks are massive, and timelines are unpredictable. The choice emigrate should bear in mind the price, however with out an correct estimate of the problem timelines slip and prices balloon:
Questions to ask embody:
Where is the information processing code and the way a lot code is there?
- The apparent place to begin is in information processing pipelines, however custom-made code might be present in all method of locations, together with BI reporting instruments like Tableau.
Does the whole lot must be migrated? Is some code being run for no good function?
- As a results of historical past and development, continuously a good portion of information processing code has been operating for no good cause.
How do you construction and plan a big migration?
- Converting the code and testing have to be staged. Best at school migrations embody a strong, pushed roadmap primarily based on finish to finish information processing. Ideally a DAG that covers all of the processing code is offered to plan the migration in logical phases.
Hidden Cost #2 – Target Platform Qualification
Can the goal platform truly carry out the whole lot in an analogous approach to the supply platform? If not then main code rewrites are required, challenge timelines slip and prices over-run.
Typically the goal platform has been examined and certified primarily based on a number of information processing pipelines. Inevitably when the whole codebase is explored options not out there on the goal platform are discovered. Working round these issues requires the assist of material specialists (SMEs) on the supply platform – entry to the big variety of SMEs develop into blocking points which delay the code migration.
With correct, automated code conversion, a whole code base might be certified in opposition to the goal platform with out counting on the reminiscence of the SME. Discovering all potential points early within the course of helps scope/price/resolve Amdahl’s regulation.
Hidden Cost #3: The Non-Standardization of SQL
“Standard SQL” is a fantasy. It’s tempting to think about that as a result of a syntax is authorized in any two SQL dialects, that it means the identical factor and returns the identical worth. A guide code translation, notably by an engineer who shouldn’t be an SME in each dialects, will are inclined to deal with profitable execution first, and depend on a protracted tail of testing and validation to substantiate that the proper calculation is being carried out on the brand new platform. This takes time, and is susceptible to error.
The hidden price shouldn’t be solely delays in migrating code, but additionally points/errors which can be found lengthy after the migration is full. A typical guide migration has errors present up for round a yr after switching to the brand new platform, and it’s frequent for an automatic migration to…
