
The CIAM Migration Diary · Part 4 of 6
The CIAM Migration Diary, Part 4. The migration itself is proving more straightforward than any of us expected. The difficulty is somewhere else entirely.
In Part 1 I wrote that I was not ready to tell anyone that replacing a critical CIAM platform is easy. My view has changed, and the diary is supposed to say so when it does.
The migration itself is proving more straightforward than any of us expected, me included. That covers what we have proven so far. We have not gone live yet.
What has been proven is most of it. The functionality works, with very few defects against the replacement platform. Performance has already been proven at several times what the customer requires in the lower environments, with a full performance test still to come. Integration has been proven across the waves we have tested. The bulk movement of identity records, which is the part people tend to fear most, has been the least eventful piece of work on the programme.
I should be fair about why. Part of it is the approach I described in Part 3. We built an environment of our own, close enough to the customer's to behave the same way, and proved almost everything there before attempting it in theirs. Problems that would otherwise have surfaced in front of everybody surfaced in front of us instead. The other part is experience. Over the last eight years complex identity migration has become something we do well, and last year we completed four of them, in four different countries, all at tier one or tier two banks. The data migration being uneventful is not luck. It is practice.
So if the migration is not the hard part, what is?
Two things, and neither of them is technical.
- The first is everything outside our control. Other programmes moving their dates. Teams whose priorities change, for good reasons that have nothing to do with us. Dependencies on suppliers and platforms you do not run. None of it is unusual and none of it is anybody's fault. But any of it can land on your timeline, and none of it can be fixed by working harder.
- The second is the exit date. The incumbent has a licence and support end date that does not move. That sounds like a commercial detail and it is actually the defining constraint on the whole programme. Miss it and the penalty is paying to extend the platform you are trying to leave, which is the one outcome the programme exists to avoid.
Put those together and you can see where the real exposure sits. It is not in the engineering. It is in the gap between a date that cannot move and dependencies that keep moving. The year end change freeze makes that gap smaller still.
The clearest illustration arrived recently, and it is something I do not think anyone could reasonably have planned for. The cloud provider hosting the platform could not supply the capacity needed for the production environments in the region. Not a pricing problem or a contractual one. The capacity simply was not there. I have been building enterprise platforms for a long time and I had not encountered that before.
For a while it looked as though the production environments would have to move to a different provider altogether. The answer then changed more than once before it settled, and in the end capacity was found and the platform stayed where it was. That outcome also spared the bank's infrastructure team a significant rebuild, which matters far more to the programme than how we got there.
It helped that the bank is not dependent on a single cloud provider. A credible alternative is what gives you leverage when you need a supplier to find a solution. That is, after all, the reason this series exists. Being dependent on one vendor for something critical is a weak position, whether the vendor sells identity or infrastructure.
There is a lesson in the response as much as in the problem. When the cause is outside your control, the resolution often is too, and it may not arrive the way you would have chosen.
The good news is in what the uncertainty never touched. Functional proving was already complete, and functional behaviour does not depend on which cloud the platform runs on. So while the infrastructure question was open, we kept moving on everything that did not depend on the answer. We built an environment of our own to mimic the target and ran the non functional work there, including performance, blue green deployment and failover, along with the runbooks and benchmarks that go with them. Pre-production now exists, and the work done while the answer was uncertain means we are ready to use it.
Nothing about the migration changed. Something entirely outside the programme did, and the answer was to keep working on whatever did not depend on it.
That brings me to what I got wrong. The signs that capacity might become a problem were visible before it became one, and we waited for them to resolve rather than escalating. We left it later than we should have. In Part 3 I wrote that we have escalated almost nothing, and I still think that is right for disagreements between teams. It is wrong for early signals of risk from outside the programme. Those are worth raising while they are still small, even when they turn out to be nothing.
Most of the important decisions on the programme have turned out to be the same decision in different forms: removing dependencies on things we do not control.
In Part 2 I described how running both platforms in parallel meant the signing key material had to stay aligned across the incumbent and the replacement, which tied our go live to that of a separate programme building a new version of one of the consuming systems. We removed that dependency by making the change on the incumbent in production ahead of time. We go live against the version of their system that exists today, not the one they are building. Every dependency on another programme's date is a dependency on their risk as well as yours.
The lesson for anyone weighing this decision is not the one I expected to be writing. The question is not really whether you can migrate. It is whether the dependencies you do not control are going to come through, and whether you hold them to account early enough to find out while there is still time to act.
The CIAM Migration Diary is Ajit Gupta's newsletter on LinkedIn. Subscribe on LinkedIn ↗

