UpCloud: MaxIOPS storage, plan shapes and who it suits

Most small cloud providers compete on price. UpCloud competes on storage, which is a narrower pitch and a more defensible one, because disk latency is the constraint that quietly limits more workloads than CPU ever does.

The storage argument

UpCloud’s block storage layer, marketed as MaxIOPS, is its own distributed system rather than a resold or off the shelf design. The point of it is consistently high IOPS available to ordinary instances, rather than performance reserved for an expensive tier.

Whether that matters to you depends entirely on the workload. A stateless web server does not care. A PostgreSQL instance with a working set larger than memory cares enormously, and so does anything doing heavy random reads: search indexes, analytics, message queues under load, CI systems checking out large repositories.

The honest test is to benchmark your own workload rather than trusting anyone’s numbers, including the provider’s. If your current bottleneck is not disk, this advantage is worth nothing to you.

The plans

The catalogue is small and readable:

• Developer plans. Small, cost efficient shapes for development and light services.
• General purpose. A balanced ratio for application servers and typical web workloads.
• High CPU. More cores relative to memory, for compute bound work.
• High Memory. The inverse, for caches and databases.

Instances are flexible enough that you can adjust CPU and memory rather than only stepping through fixed rungs, which reduces the amount you pay for capacity you do not use.

Locations

The footprint covers northern and western Europe well, with a Finnish origin and sites across the continent, plus North American, Asian and Australian locations. For European workloads it is a genuine alternative to the larger operators, and it is an EU company, which matters for some procurement processes.

The rest of the platform

• Managed Databases. PostgreSQL, MySQL, Redis compatible caching and OpenSearch, with the same storage layer underneath.
• Managed Kubernetes. Worker nodes billed as instances.
• Object Storage. S3 compatible, priced with included capacity and transfer.
• Load Balancer and SDN private networks. Software defined networking between instances, with private traffic free.
• Backups. Scheduled or manual, billed per gigabyte.

The API and Terraform provider are complete enough to run infrastructure as code properly, which is not true of every provider at this size.

What to check on the account

As with every provider in this category, outbound SMTP is restricted on new accounts and opened by request. Resource limits also apply per account and are raised once there is a payment history. Neither is unusual; both are worth confirming before you plan a fleet or a mail sending service.

Who it suits

Teams running self managed databases, disk bound applications, or anything where storage latency shows up in the user facing response time. Also European organisations that want an EU operator with a modern API and a small learning curve.

It is not the cheapest option in this bracket. If raw price per unit of compute is the deciding factor, Hetzner is lower, and if you want the largest community and documentation set at this size, DigitalOcean is the safer default. UpCloud earns its place when the workload is genuinely disk sensitive.