The Cloud Is Not Another Data Centre
Many cloud disappointments begin with a reasonable instruction: move what we have and change as little as possible. That instruction can reduce migration risk, and it can also carry oversized servers, always-on test environments, manual approvals, fragile dependencies and unowned costs into a metered environment. This seven-part series explains what actually changes when capacity becomes a meter: why lift-and-shift needs a dated exit, how to choose a treatment for every workload, when serverless is and is not cheaper, why FinOps starts in the design review, how the operating model has to move with the workloads, and when to retain, repatriate or retire. One fictional South African distributor carries the examples, Part 3 includes an interactive workload treatment worksheet, and every figure is illustrative rather than a promise of savings.
7 articles
- Part 1The cloud is not another data centre: the commercial model changedAn on-premises server is bought once for peak demand; a cloud server is billed every hour it runs. Part 1 explains why moving the same estate onto a meter can raise the bill, and what has to change alongside the hosting: capacity, change, ownership and finance.9 min read
- Part 2Lift-and-shift is a migration wave, not a destinationRehosting is a sound way to meet a hard exit date. It becomes a trap when the temporary state has no owner, budget or trigger. Part 2 shows when rehosting is the right call, the seven-field exit ticket every rehosted workload should carry, and how to rightsize before and after the move.9 min read
- Part 3Choose a treatment for every workload: the migration Rs as decisionsAWS lists seven migration strategies, Microsoft eight and Google six, but the decisions underneath are the same. Part 3 turns the Rs into a decision method, explains why an inventory is not discovery, sequences waves by dependency, and includes an interactive worksheet you can download.10 min read
- Part 4Serverless is not automatically cheaperFunctions remove idle cost for bursty, short work, and can cost more than containers for steady, heavy traffic. Part 4 explains how serverless is billed, the five ways its costs rise, when Kubernetes arrives too early, and why the cheapest compute line is rarely the cheapest service.9 min read
- Part 5FinOps starts before deploymentBy the time finance receives a surprising cloud invoice, the decisions that produced it were made months earlier, in architecture reviews and pipeline defaults. Part 5 shows how to connect design, ownership and a unit of business value before anything is deployed.10 min read
- Part 6The operating model must move tooCloud platforms make change fast. An organisation that keeps monthly approval boards, hand-built environments and project handovers will either block that speed or lose control of it. Part 6 shows how to turn governance into reusable controls instead of removing it.9 min read
- Part 7Retain, repatriate or retire: correcting placement without ideologyMoving a workload out of public cloud is not automatically failure, and keeping one on premises is not automatically caution. Part 7 sets out a six-step test for repatriation, shows why the cheapest migration is often a retirement, and closes the series.10 min read
Working through any of this in your own organisation?
These articles come out of delivery work rather than theory. If the problems in this series look familiar, a short scoping conversation is usually the quickest way to work out what is worth doing first.
Talk to CloudNala