Skip to content

Verified cloud accounts across 14 providers

Choosing a cloud region: latency, data residency and cost

Region choice is one of the few infrastructure decisions that is awkward to reverse. Three factors decide it, and they often disagree.

Region selection looks like a dropdown and behaves like an architectural commitment. Moving a running workload between regions means moving data, re-creating resources and usually accepting downtime, so it is worth spending real time on before the first deployment rather than after.

Latency: distance to the people and systems that call you

Network latency is bounded by physics. Light in fibre travels at roughly two thirds of its speed in vacuum, and real routes are never straight, so round-trip time grows with geographic distance in a way no amount of optimisation removes. The practical rule is to put compute close to whoever or whatever talks to it most.

That is not always the end user. If an application makes many sequential calls to a database, the distance between the application and the database matters more than the distance to the browser, because each round trip is paid again. Co-locating the chatty components usually beats moving the whole stack closer to users.

When a single region is enough

Multi-region architecture is expensive in engineering time, not just in resources. For many applications a single well chosen region plus a content delivery network in front of static assets gives most of the perceived benefit for a fraction of the complexity. Reach for multiple regions when you have a concrete requirement, such as a legal obligation or a recovery objective that a single region cannot meet.

Data residency: where the law says the data must sit

For many organisations this is the constraint that decides everything else. If personal data must remain within a particular jurisdiction, the region list narrows before latency or price enters the conversation. Two details are easy to miss.

  • Backups and replicas count. A database in an in-scope region with backups replicated elsewhere may not satisfy the requirement.
  • Managed services vary. Not every service is available in every region, and some regional services still depend on a control plane elsewhere. Check the specific services you plan to use, not just the region.

Cost: the same resource is not the same price everywhere

Providers price by region, and the differences are not trivial. Compute, storage and especially data transfer can vary noticeably between regions within the same provider. If a workload is price sensitive and has no residency constraint, comparing a few regions before committing is one of the easier savings available.

Data transfer deserves particular attention. Traffic between regions, and traffic leaving the provider’s network entirely, is usually billed differently from traffic inside a single region. An architecture that moves large volumes between regions can cost more in transfer than in the compute doing the work.

Availability zones are a separate decision

Within a region, most providers offer availability zones: separate facilities with independent power and cooling, connected by low latency links. Spreading across zones protects against a single facility failing and normally costs little or nothing extra in latency. Spreading across regions protects against a whole region failing and costs a great deal more in complexity. Teams often conflate the two and buy the expensive protection when the cheap one was what they needed.

A workable order of operations

  1. Eliminate regions that residency rules forbid.
  2. Of what remains, shortlist those close to the dominant traffic source.
  3. Confirm every managed service you need exists in the shortlist.
  4. Compare price among the survivors, including data transfer.
  5. Use multiple availability zones inside the chosen region by default.
Chat on Telegram
Scroll to Top