Cloud infrastructure for an MVP: what do you really need?
An MVP needs a small, understandable infrastructure; this article shows what the minimum is and what you can safely postpone.
An MVP usually needs just an environment with an application server or platform service, a database, storage for files, a backup and basic monitoring. Keep it simple, document the set-up and postpone scaling provisions until usage gives a reason for them.
Who is it for: Founders and small teams who want to put a first version of a product online without building too much infrastructure.
What is the purpose of the infrastructure for an MVP?
An MVP, a minimum viable product, exists to learn whether people want to use it. The infrastructure therefore does not need to scale to millions of users. It must be reliable enough to collect feedback and simple enough to change quickly.
By cloud we mean infrastructure here: servers, databases and storage. The cloud for files and collaboration, such as shared folders for your team, falls outside this.
What is the minimum?
Start with components that make the product run and that you can recover if something goes wrong.
- A place for the application: a single server or a platform service that runs code.
- A database, preferably a managed variant so that backups and updates are taken care of.
- Storage for files such as uploads, preferably object storage with an S3-compatible interface.
- A backup of the database and files, in a different place from the production environment.
- Basic monitoring: an alert if the service is unreachable or disk space is running out.
- Access management with MFA on all admin accounts.
What can you postpone?
Many teams build too early for a scale that does not yet exist. Orchestration with Kubernetes, multiple regions and automated scaling solve problems that an MVP usually does not have. They do add management work.
Postponing is not ignoring. Write down at which level of usage you will look at this again, for example when the server is structurally too full or when an outage really affects customers.
How do you leave room to grow?
Choose components that you can move later. Run the application in a container, keep configuration outside the code and describe the set-up in a script or document. Then a move to a larger set-up takes days instead of weeks.
MVP components and the moment to scale them up
| Component | Now for your MVP | Look again when |
|---|---|---|
| Application | One server or platform service | The load is structurally high or downtime affects customers |
| Database | Managed database with backup | Queries become slow or you need to be able to recover faster |
| Files | Object storage | You get a lot of traffic or strict retention periods |
| Monitoring | Availability and disk space | Several services depend on each other |
| Scaling | Scaling up manually | Peaks in usage return predictably |
Who is it not ideal for?
- Products that process sensitive data from day one: you then need stricter security and legal review sooner.
- Services with known large peaks, such as an event: then plan capacity deliberately.
- Teams who want to manage their own server without technical knowledge: a platform service is more suitable then.
Frequently asked questions
Do I need a test environment for an MVP?
A small separate environment helps you try out changes without affecting users. It can stay small and be switched off outside working hours.
Should I choose Kubernetes straight away?
Usually not. Using containers is sensible, but an orchestration layer such as Kubernetes adds management work that rarely pays off for a small product.
How often do I test a recovery?
Do it at least once before launch and again after major changes. A backup that you have not restored is not yet proven.
