$ cat choices/gcp.md
the call
Cost, control, performance. I've migrated to it more than once for the same three reasons, and they held both times. I don't move clouds for the logo on the slide, only when those three actually move.
The reasons I reach for GCP are boring on purpose: cost, control, performance. Predictable pricing without surprise egress games, primitives that let me own the network and the runtime instead of fighting the platform, and a network and compute layer that just moves. When those three line up, the migration pays for itself. When I’ve made the call, they held.
A cloud migration is one of the most expensive things a team can do, and most of them are vanity. If the bill isn’t bleeding, control isn’t the bottleneck, and performance isn’t the problem, you don’t move. Picking a cloud because it’s trendy, or because a competitor uses it, is résumé-driven infrastructure. The switching cost is real: retraining, re-tooling, re-wiring everything that touches the old provider. It has to earn its way in on the three reasons, or it doesn’t move.
I’ve migrated to Google Cloud more than once. At Brandfolder, and again at CommercialTribe alongside the move to Kubernetes, both times the case was the same three reasons: cost, control, performance. Not a trend, not a logo. The same scorecard, and it held both times. Doing it twice is the proof: when a decision survives a second independent test on the same criteria, it wasn’t luck.— see: works / Brandfolder
Decide on the things that matter and check that they actually moved. Cost, control, performance is a scorecard, not a slogan, and the discipline is refusing to move until the scorecard says move. The best infrastructure calls are the ones you can re-run on different teams, different stacks, different years, and watch come out the same. That repeatability is what separates judgment from preference.
the gaps — what it costs even when it’s right
Lock-in just moves, it doesn’t vanish. Lean on managed services and you’ve traded one provider’s gravity for another’s. The control GCP buys you is real, but the more of its native services you adopt, the harder the next move gets.
The migration itself is the tax. Even a clean cut costs you retraining, re-tooling, and a stretch where the team is fluent in neither the old world nor the new one. The three reasons have to clear that cost, not just beat steady-state.
Smaller ecosystem, sharper edges. Fewer third-party integrations, fewer Stack Overflow answers, fewer engineers who’ve already done the thing you’re doing. You pay for the cleaner primitives with a thinner safety net around them.
used at
works / Brandfolder