When you outgrow a VPS: dedicated and GPU servers

You have outgrown a VPS when memory, CPU or disk run short at your busiest time even on the largest plan, or when your workload needs a GPU or isolation a shared machine cannot give. Before that point, a bigger plan, a second server or a separate database usually solves it. If none of those fit, a dedicated server gives you a whole machine, with far more memory and disk than a VPS offers. At Aria Cloud, dedicated and GPU servers are available by request: you tell us what you need, we talk it through, set the price and set it up. This post covers the signs, the cheaper fixes to try first and what to prepare for that conversation.

We sell both, and we recommend growing with Aria: first a bigger plan, then a dedicated or GPU server set up with you.

How do you know you have outgrown a VPS?

Do not guess. Measure at your busiest time, not on average.

Sign	How to check	What to try first
Memory runs out	Free memory falls below about a fifth at peak, or containers restart	Move up one plan
Builds slow deploys	Builds run one at a time and other deploys wait	A larger plan, or a second server for builds
CPU stays busy	CPU stays high for long periods, not brief spikes	A plan with more vCPUs
Disk fills	Database, files or logs fill the disk	A plan with more disk, or move files to object storage
The server is full of apps	Several apps share one server and none has room left	A second server
A database needs its own machine	One database dominates memory and disk	Run it on its own server
A GPU workload	Models, transcription or image or video generation	A GPU server
Isolation	One customer’s data must not share a machine	A separate server, or a dedicated one

Aria samples your usage every minute and suggests a bigger plan, and emails you, when memory, CPU or disk run short (billing docs). Its sizing tool also tells you how much room is left before you add an app. See how to size a server for several apps. What a managed server covers is on managed VPS for developers.

What should you try before buying a bigger machine?

Move up a plan. Aria’s private VPS plans run from Starter (2 vCPU, 4 GB, 40 GB, £9 a month ex VAT) to Scale Plus (16 vCPU, 32 GB, 400 GB, £60). A bigger plan needs one restart. That covers most growing products. All the numbers are on limits.
Split the stack. Put the busiest database or the heaviest worker on a second server. Two mid-sized servers are often cheaper and easier to run than one very large one.
Trim the work. Cache more, fix slow queries and reduce what builds need. These cost nothing but time.
Buy part of the stack elsewhere. A managed database or an object store can take the biggest load off a server. The trade is another bill.

If you reach Scale Plus and still run short, or the workload has one of the special needs below, it is time to talk about something larger.

What is a dedicated server?

A dedicated server is a whole physical machine that only you use. Nobody else’s workload shares it. It can have far more memory, cores and disk than a VPS: we can arrange servers with up to 120 GB of RAM and beyond. Because you take the whole machine, you pay for all of it, whether you use it or not.

It makes sense when:

your stack needs more memory or disk than the largest VPS plan;
a database is large enough that you want it on its own hardware;
you want consistent performance with no shared neighbours;
compliance or contracts ask for hardware that is not shared.

Measure first, so you pick the right size. If a second VPS would do the job, Aria sets that up as well. Aria takes nightly encrypted backups of your databases and volumes, stored off the server, on every plan.

How does a dedicated server work at Aria?

By request, with a consultation. There is no order button and no public price list, on purpose. Your needs decide the server, so the conversation comes first:

You tell us what you need. Memory, cores, disks, and where it should run.
We talk it through.
We set the price.
We set it up.

We publish no prices for dedicated or GPU servers because the right machine and price depend on the workload. Start from the dedicated section on the pricing page.

What should you bring to the conversation?

Come with numbers. The more exact you are, the better the answer.

Bring	Why it matters
Peak memory and CPU use	Sizes the machine from facts, not guesses
Database size and growth per month	Decides disk and backups
What must be fast	Disk, memory or network
Where your users and data should be	Decides the location
What happens if it is down	Decides how much redundancy you need
Your current `aria.yaml`, if you have one	Shows what you run today
A budget range	Helps us suggest the right class of machine

Ask how deploys, backups, access and updates work on the machine. Some of those may differ from a VPS plan, and you should hear the answer before you decide. The same questions are worth asking any provider.

What about GPU servers?

A GPU server is for workloads that need a graphics processor: running AI models yourself, transcription, image or video generation, 3D rendering. They are by request only, with no public price, for the same reason as dedicated servers. In aria.yaml, a service can ask for a GPU with the gpu field, and a stack that asks for one only fits a GPU server (aria.yaml reference). If your code only calls a hosted AI API, you do not need a GPU. Leave the field out.

Tell us the model or workload, the GPU memory it needs and how many requests or jobs you expect. A smaller GPU is often enough for a small model, and you should size to the workload, not to the biggest card you can find.

Why grow with Aria?

You keep the same Aria on top at every size: the same aria.yaml, the same Aria AI and MCP tools, the same approvals and the same backups. The step from a plan to a dedicated or GPU server is a conversation, not a migration to another platform. Our fit is a stack you want on one machine, at a price you can predict, with a server run for you.

FAQ

How big can a dedicated server be?

Up to 120 GB of RAM, and beyond that on request. Tell us your memory, core and disk needs and we will say what is possible.

How much does a dedicated server cost?

We do not publish prices. We set the price with you after we have talked through what you need.

Can I get a GPU server?

Yes, by request, with no public price. Tell us the workload and the GPU memory it needs.

Do I need a dedicated server for a big database?

Not always. First try a bigger VPS plan or a separate server for the database. Measure memory and disk use at your busiest time before you decide.

How do I work with a dedicated server?

The same way as a VPS plan. Aria AI or your coding agent writes aria.yaml, and Aria deploys. Ask in the consultation about access, deploys and backups, and we will tell you how they work on your machine.

Is a dedicated server faster than a VPS?

We make no speed claim. A dedicated machine has no neighbours and can be larger. Whether that helps depends on what limits your app today, which is why we ask for measurements.

Ready for more than a VPS? Look at the dedicated section on the pricing page, or get started on a private VPS and tell us when you outgrow it.