VPS vs VDS vs dedicated server: how to choose

A VPS, a VDS and a dedicated server all give you a server of your own, with root or administrator access, your own operating system and your own IP address. What differs is how much of the physical machine is reserved for you, and which features the server exposes. This guide explains each option, shows you how to measure what you need and gives you a checklist for choosing.

The short answer

• Choose a VPS for websites, small applications, development servers, VPNs and bots. You get an isolated virtual machine on a shared physical host.
• Choose a VDS when you need more resources or want to run your own virtual machines. At HyperDC every VDS has nested virtualization available, larger CPU and memory allocations and NVMe storage.
• Choose a dedicated server when you need the whole machine: consistent performance under sustained load, hardware access, strict single-tenant isolation or software that must run on its own physical hardware.

How a VPS works

A virtual private server is a virtual machine on a physical host that runs several virtual machines side by side. A hypervisor divides the host's processors, memory and storage between them and keeps each one isolated: you get your own kernel and operating system, root or administrator access, your own IP address and the resources your plan allocates. Our Linux and Windows VPS plans are KVM virtual machines.

Because the host is shared, providers can allocate more virtual CPUs in total than the host has physical threads, on the assumption that not every server is busy at the same moment. This is normal practice and keeps prices low. It suits workloads that are quiet most of the time and busy in short bursts, such as most websites, development environments and small APIs.

When several virtual machines want the CPU at the same time, each waits a little. On Linux you can see this as steal time: the st value in top shows the share of time the hypervisor gave your virtual CPU to another machine. Occasional low values are harmless; steal time that stays high during your busy hours means your workload needs reserved resources.

What a VDS is at HyperDC

VDS stands for virtual dedicated server. Across the industry the word is used loosely, so always read the plan details rather than the label. At HyperDC the rule is precise: we call a server a VDS only when nested virtualization is available on it, so you can run your own virtual machines, emulators or a hypervisor inside it. Our Linux, Windows and macOS VDS plans offer it, together with larger CPU and memory allocations and NVMe storage. They suit production websites and shops, databases, build servers, game servers and virtualization labs. Read more in VDS and nested virtualization.

Why Apple Silicon macOS is a VPS

On Apple silicon Macs, macOS virtual machines run on Apple's Virtualization framework, which does not offer nested virtualization for macOS guests. Because you cannot run virtual machines inside them, our Apple Silicon macOS servers are sold as Apple Silicon macOS VPS. They still give you a complete macOS system with administrator access for Xcode builds and app testing.

When a dedicated server is worth it

A dedicated server gives you the whole physical machine. Nothing else runs on it, so performance is as consistent as the hardware allows, and you can use what a virtual machine cannot offer: your own disk layout and RAID, a remote management interface where the plan includes one, or a hypervisor of your own. Typical reasons to move to bare metal:

• Sustained load: databases, analytics or game servers that keep the CPU or disks busy for hours.
• Software requirements: some software or vendors require physical, single-tenant hardware.
• Isolation: a single-tenant server makes it simple to show auditors that no one else shares your hardware.
• Hardware access: your own virtualization platform, specific disks or controllers, full control of the network stack.

The trade-off is flexibility. Adding memory or disks to a physical server is a hardware job, and a bigger machine usually means a migration, so plan capacity with some headroom.

Side by side

Measure before you upgrade

Decide with numbers from your busiest hours, not with a feeling. On a Linux server:

nproc
free -h
vmstat 1 10

In vmstat, watch the last columns: wa is time spent waiting for disks and st is steal time. A run queue (r) that stays above the number of CPUs from nproc means processes are waiting for CPU.

Pressure stall information shows how much time tasks spend waiting for CPU, memory or disk:

cat /proc/pressure/cpu
cat /proc/pressure/memory
cat /proc/pressure/io

If the files are missing, PSI is not enabled in your kernel (some distributions need the psi=1 boot parameter); use vmstat instead. For disk latency, install the sysstat package and run:

sudo apt install sysstat
iostat -x 1 5

High %util together with rising r_await and w_await values means the disk is the bottleneck. These signals tell you it is time to move up:

• Steal time stays high while your traffic peaks.
• Memory runs out: the server swaps, or the kernel stops processes because memory is exhausted.
• Disks keep you waiting: high wa and slow queries during peaks.
• Pressure keeps rising in /proc/pressure.
• Audits or software requirements call for single-tenant hardware.

Nested virtualization: a quick check

To see whether a Linux server exposes the processor's virtualization extensions to you, count the vmx (Intel) or svm (AMD) flags:

grep -Ec '(vmx|svm)' /proc/cpuinfo

A result of 0 means you cannot run accelerated virtual machines inside this server, which is normal on a VPS. On a VDS the result is greater than zero.

How to choose in five steps

1. Describe the workload. Is it busy all the time or in bursts? Heavy on CPU, memory or disk?
2. Measure what you run today. Record steal time, memory use, disk wait and response times during peak hours.
3. Check rules and hardware. Compliance rules, software requirements, your own hypervisor, RAID or GPU.
4. Size with headroom. Pick the smallest plan that covers your measured peak with room to spare, in a location close to your users.
5. Review as you grow. Moving from VPS to VDS or dedicated is a normal step when your numbers change.

Common mistakes

• Upgrading blind. A slow site is often a slow database query or a missing cache, not a small server. Measure first.
• Buying a VPS for virtual machines. Without nested virtualization, hypervisors and Android emulators will not run. Choose a VDS.
• Planning to shrink later. Disks cannot be made smaller, so moving to a smaller plan means a new server and a restore.
• Ignoring location. A bigger server far from your users can feel slower than a small one nearby. See choose a data center location.

Next steps

• See what each line includes: Linux VPS, VDS and nested virtualization, dedicated servers.
• Compare plans and prices: cloud servers.
• Is the server struggling right now? High CPU or memory usage.