Kamatera and the Configurable Server Model

Most cloud providers sell machines from a menu. You pick a size, and CPU, memory and storage come bundled in a fixed ratio. Kamatera takes a different approach: you assemble the machine from components, choosing processor count, memory and disk more or less independently.

Why fixed sizes exist at all

Fixed instance families make capacity planning simple for the provider and predictable for the customer. The cost is that real workloads rarely match the ratios on offer. A memory-hungry application with modest CPU needs is forced up the ladder, paying for cores it will never use.

What configurability buys

Being able to pick memory without dragging CPU along is genuinely useful for a few common shapes:

• In-memory caches and search indexes, which want RAM far more than cores.
• Build and CI machines, which want cores in bursts and little else.
• Database servers with a working set that must stay resident, where an extra increment of memory is worth more than any other upgrade.

The trade-off is that you have to know your workload. A menu protects you from your own guesses; a component builder does not. If you are unsure, start from a balanced configuration and adjust once you have measurements, as described in how to size your first cloud server.

Network behaviour and port 25

The other characteristic worth understanding is outbound mail. Providers restrict outbound traffic on port 25 by default to limit spam originating from their networks, and lifting that restriction is an account-level decision rather than a firewall setting you can flip yourself.

An account that already carries this permission removes an approval step for applications that send mail directly rather than through a relay. The general topic is covered in port 25 and outbound mail on cloud servers.

Current catalogue options

Who this suits

Kamatera fits teams that have a clear picture of the resource shape they need and want to pay for that shape rather than the nearest fixed size above it. It suits mail-adjacent and network-specific workloads where the port configuration matters.

It is less obviously the right choice for teams that want a large managed-service catalogue around the compute, which is where the hyperscale providers such as AWS and Google Cloud are stronger.

After taking control

Change the account password, enable any available second factor, rotate API keys, and review existing servers for software you did not install. Then check firewall rules before exposing anything, because inherited rules reflect someone else’s intentions rather than yours.