Contents

For an MVP or a small project, one or two VPSs running containers are usually enough, while the database and file storage are best handed to managed services from the start. Full cloud infrastructure is needed when load grows unevenly, multi-zone fault tolerance is required or the team wants to spend less time on servers. Compare options by total cost of ownership, including engineering time.

Here is how hosting options differ, how to estimate cost and what to choose at each stage of a product.

How do VPS, cloud and managed services differ?

A VPS is a virtual server you manage yourself: installing the OS, database, updates and backups. The cloud offers a set of services with flexible scaling and pay-as-you-go pricing. Managed services — databases, queues, storage — take the administration of individual components off your hands.

Option Who administers Scaling Resource cost When it fits
VPS You Manual Low MVPs, small projects, predictable load
Dedicated server You Manual, slow Low per unit of power Stable high load
Cloud (IaaS) You and the provider Automatic Medium to high Uneven load, growth
Managed services The provider Automatic Higher than a VPS Databases, queues, storage — almost always

How do you calculate total cost?

Total cost is not just the provider's bill but also engineering time for setup, updates, monitoring, backups and incident response. A cheap VPS with a self-managed database can cost more than a managed one once you count admin hours and the risk of data loss.

What to include:

  1. The cost of servers, storage, traffic and backups.
  2. Time for initial setup and automation.
  3. Monthly hours for updates, monitoring and minor incidents.
  4. The cost of downtime: how much you lose per hour of unavailability.
  5. The risk of data loss and recovery time — see backups.

We covered estimating maintenance costs in website and app support cost.

A hand holding house keys at the entrance Photo: Unsplash

What should you choose for an MVP?

A simple setup suits an MVP: one or two VPSs with the application in containers, a managed database or at least automated backups, S3-compatible storage for files and a CDN. It is cheap, understandable and does not get in the way of moving when the product grows.

A starter setup:

  • the application in Docker on a VPS — see Docker for development and production;
  • a managed database with backups;
  • object storage for user uploads;
  • a CDN for static assets and images;
  • deployment via CI/CD and basic monitoring.

On the budget for a first product version, see MVP development cost.

When should you move to the cloud?

Move to the cloud when load starts fluctuating noticeably, downtime becomes too expensive or manual server management eats a lot of the team's time. The cloud provides autoscaling, multiple availability zones and dozens of ready services, but you have to watch spending.

Signs it is time:

  • peak load is several times the average — sales, seasonality, ad campaigns;
  • you need to run across several availability zones for fault tolerance;
  • engineers spend a lot of time on server routine;
  • you need specialised services: managed Kubernetes, queues, ML infrastructure.

On orchestration and serverless options, see do you need Kubernetes and serverless: when it fits.

What should you consider when choosing a provider?

When choosing a provider, prices matter, but so do data centre locations, data residency compliance, support quality and the ability to leave without huge costs. Projects handling personal data of Russian citizens need a provider that stores it in Russia.

  • data centre locations and legal data residency requirements;
  • managed databases, storage and CDN from a single provider;
  • the service level agreement and the provider's real outage history;
  • egress traffic costs — often an unexpected expense;
  • compatibility with standard tools so you are not locked into one cloud.

On personal data requirements, see personal data on a website.

In short

  • A VPS is cheap and simple, but administration is on you.
  • The cloud offers flexible scaling and ready services at a higher price.
  • Databases and file storage are almost always better taken as managed services.
  • Compare total cost, including engineering time and the price of downtime.
  • An MVP fits on VPSs with containers; move to the cloud as you grow.

FAQ

  • Can you move from a VPS to the cloud later?

    Yes, if the project is containerised, settings live in environment variables and files are kept in S3-compatible storage. Then moving is mostly migrating the database and updating DNS.

  • Where must personal data of Russian citizens be stored?

    Initial recording and storage of Russian citizens' personal data must happen in databases located in Russia. Projects handling such data therefore choose Russian providers or data centres in Russia.

  • How much does hosting a small project cost?

    An MVP often fits on one or two mid-sized VPSs plus object storage, costing tens of dollars a month. A managed database and redundancy add cost but reduce risk.