Questions
Which cloud, and what would make you leave?
Choose the provider that runs your real workload at the lowest total cost your team can operate safely, keeps data where your customers and regulators need it, and could be left within a known cost and time. For most small teams, the provider the engineers already know well wins, unless a particular service, region, or price gap is large enough to pay for learning another.
Price the workload you actually run
Start from what you run rather than from a price page. Count the servers and their hours, the storage, the managed databases, and above all the data you send out to users and other systems, because outbound transfer fees surprise more teams than any other line. Then price that workload with the discounts you would really take, such as one- or three-year commitments.
Add the people. A team fluent in one provider builds and fixes faster on it, while retraining, hiring, and running two sets of tools all cost money. A provider that is 15 percent cheaper on paper can prove dearer in practice once the learning is counted.
A worked example
A company with twelve engineers runs 40 servers and 50 terabytes of storage, and sends 20 terabytes a month to users. On Provider A, which the team knows, the workload costs $29,000 a month with a three-year commitment. Provider B quotes $24,000 a month on the same terms, a saving of $60,000 a year.
Moving would take three engineers about four months at a loaded cost of $15,000 each a month, $180,000, plus two months of running both clouds side by side at about $30,000 a month. The one-off cost of $240,000 takes four years to recover, longer than the three-year commitment that produces the saving, so unless Provider B offers something the business needs, staying is the stronger case.
The same numbers show what would change the answer. If the workload doubled, the yearly gap would nearly double and payback would fall to about two years. If Provider B offered migration help that halved the cost of the move, moving would become a serious option.
Leaving is part of the choice
Every provider makes some services easy to start and hard to leave, such as databases and queues built only for its platform. Using them is often sensible, but count each one as part of the exit cost. Moving 50 terabytes out at an assumed $0.08 a gigabyte costs about $4,000, which is small beside the work of rebuilding whatever depended on one provider's own services.
Running on two providers at once sounds like insurance, yet it doubles the tools, the skills, and the places to make mistakes. For most teams it earns its keep only for a specific workload with a clear reason, such as a customer contract or a region one provider lacks.
How the room weighs providers
Staying, moving, and splitting off a specific workload each get their own Champion, and the monthly, migration, and payback sums are computed exactly from your figures. Scout checks public price lists and region availability, and the record keeps the date, because cloud prices change.
Craftsman asks whether your team can operate the new platform with the hours it has, and Gray Wolf hunts support plan fees, transfer charges, and minimum commitments. Grandfather points to the three-year commitment as the part hardest to undo. The Verdict names the workload size or price gap that would change it, and the Human Decision stays with your engineering leaders.
Questions
- Should we take a long commitment discount?
- Only for the part of your workload you are confident will still be running at the end of the term. Commit the steady base, and pay normal rates for growth and experiments.
- Does data location matter?
- It can. Some customers, industries, and countries require data to stay in a particular region, so confirm what your contracts and the rules that apply to you require before comparing prices.
- When is it worth running on two clouds?
- When a specific workload has a clear reason to sit elsewhere. Spreading everything across two providers by default usually costs more than the outage it is meant to prevent.
Further reading
Bring Argumont a Decision worth challenging.
Start with a paid 60 day pilot on your own real Decisions.