01The tool fee
How does AWS bill for AWS Transform?
AWS publishes the current pricing model on the AWS Transform pricing page. Read that page on the day you budget. Ace8 does not quote a rate here, because AWS changes service pricing, and the model can differ between transformation types such as Java upgrades, .NET, mainframe and VMware.
When you read the page, write down four things: the billing unit (for example per line of code, per job or per unit of agent time), whether there is a free tier or a free usage allowance, whether any transformation types are priced differently, and which other AWS services your run uses and bills separately.
For what the tool changes in a Java codebase, see AWS Transform for Java. Ace8 is not affiliated with AWS.
02Beyond the fee
What else does the project cost?
A tool can produce a patch in minutes. Getting that patch to production is where the budget goes.
- Engineer review: a senior engineer reads every diff. Review time grows with the size of the diff and the age of the code.
- Test repair: failing and missing tests have to be fixed or written before you can trust the result.
- Fix-up: the tool leaves compile errors and unsupported libraries, and someone fixes them by hand.
- CI and environment: build agents, new JDK images, container base images and staging time.
- Rework: runs that must be repeated because the first output was wrong or the scope changed.
- Coordination: change windows, approvals and the teams that own each application.
03Worked formula
A cost formula you can fill in
Per application: cost = T + R + F + Q + E, then multiply R + F + Q by (1 + X). Total project cost is the sum across applications plus program management.
Notice what dominates. For most estates T is small beside R, F and Q. A cheap tool on untested code still produces an expensive project.
| Variable | Meaning | Where to get it |
|---|---|---|
| T | Tool fee for the application | AWS pricing page and your usage estimate |
| R | Review hours x engineer rate | Diff size from a trial run, times your review speed |
| F | Fix hours x engineer rate (compile errors, unsupported libraries) | Errors left after a trial run |
| Q | Test repair or writing hours x rate | Test coverage and failing tests today |
| E | CI, environments and cloud costs for the run | Your platform team |
| X | Rework allowance, as a percentage of R + F + Q | Your own judgement; start higher for untested code |
04Method
Measure before you budget
Do not estimate R, F and Q from opinion. Run the tool on two or three representative applications and log the output: files changed, compile errors left, tests failing, and minutes of senior review. Then scale those figures to the rest of the estate by size and test coverage.
We are running this test on an open-source Java 8, Spring Boot 2 and javax application, comparing AWS Transform with OpenRewrite. The cost per run will be published here when it is complete. We are not stating a figure until then.
05Limits
When this formula misleads
- A single small application: the fixed cost of setting up a trial may exceed the project.
- Code with no tests and no owner: X is unbounded, so run an assessment first.
- Work that is mostly a platform change, such as an app server exit: the code edit is the minor part. See Java modernization.
06Next step
How Ace8 helps
Ace8 is the AI-assisted application modernization practice of AceMQ. We run tools such as AWS Transform on client code, measure the trial, and put the result into a priced plan. The assessment produces the inventory and the trial figures that fill in the formula above. Start with AWS Transform at Ace8 to see how we run it.