Exit
Or don't
Cases where the cloud is the wrong answer — latency, data sovereignty, cost structure, operational control — argued plainly.
a legitimate ending — we mean it
The worksheet
Every guide at this fork answers one row. Work down the left column until something you believe stops being true.
| What you think | What is actually true |
|---|---|
| The cloud is always the modern answer. | Sustained utilisation, strict data residency and a predictable cost ceiling still favour your own hardware.When On-Premises Wins → |
| Nobody actually moves back. | Repatriation is more common than the industry admits — with its own structural costs going the other way.The Repatriation Argument → |
| A nearby region is close enough. | Roughly 200 km per millisecond through fibre. The floor does not negotiate, and no tuning recovers it.Latency Is Physics → |
The 3 guides, in reading order
Every guide →
When On-Premises WinsAn honest accounting of the workload characteristics — sustained high utilisation, strict data-residency requirements, predictable cost ceilings — where running your own hardware remains the rational choice. Written without nostalgia or anti-cloud axe-grinding.By the ExtraSys desk · Or don't · 3 min read
The Repatriation ArgumentAn explainer of cloud repatriation — moving workloads back on-premises or to colocation — covering the conditions under which it makes sense, what the transition costs look like in structure, and why it is more common than the industry acknowledges.By the ExtraSys desk · Or don't · 4 min read
Latency Is PhysicsA short, blunt piece about the speed-of-light constraint: what distance means for round-trip time, why a regional cloud endpoint is not equivalent to local infrastructure for latency-sensitive workloads, and which classes of application this actually affects.By the ExtraSys desk · Or don't · 2 min read