Choose a VPS, VDS or dedicated server by the resources and operating model your application needs. A VPS is a virtual machine; VDS is a provider's product label whose resource guarantees must be checked; a dedicated server gives you a physical machine allocated to you. Buying a product called VDS does not automatically buy exclusive CPU cores.
At RS Computers, VPS and VDS plans are KVM virtual machines with shared vCPUs. The VDS range adds larger RAM and storage allocations and a 10 Gb/s port. That distinction matters when a workload needs more memory or network capacity but does not need to control a whole physical server.
If you are deciding today, answer three questions first: does the application fit in the available memory, can it use more than one CPU thread, and does it require control of the physical host? A small website and an internal database server can need different plans even when they have the same number of users.
VPS vs VDS vs dedicated server at a glance
| Question | VPS | VDS at RS Computers | Dedicated server |
|---|---|---|---|
| What do you administer? | A virtual machine | A larger virtual-machine plan | A physical server and its operating system |
| Are vCPUs exclusive? | Shared on our plans | Shared on our plans | Physical CPU resources belong to that machine; its network may still be shared |
| Typical reason to choose it | Small sites, bots, VPNs and development | More RAM, NVMe capacity or port capacity | Physical-host control or sustained workloads that justify a whole machine |
| What should you check? | Memory, disk space, workload limits and backups | The same checks, plus whether a faster port helps | Hardware replacement, remote access, disks and recovery arrangements |
This comparison uses our own VPS/VDS specifications as of 11 October 2026. Other providers use these names differently. Ask for the resource policy in writing when CPU exclusivity is a requirement.
What KVM isolation gives you
A KVM virtual machine runs its own guest operating system and kernel. That gives you a place to manage your services, users and firewall independently of another customer's guest. It does not mean the physical CPU, storage system or network path is exclusive to you. Isolation and resource exclusivity are separate properties.
QEMU's system-emulation documentation explains how a virtual machine presents CPU, memory and devices to a guest and can use KVM for acceleration. For a hosting purchase, the useful follow-up is the provider's resource policy: CPU scheduling, memory allocation, disk capacity and network limits.
Our Debian VPS images already include SSH, so you can begin administering the server after receiving access. Root access lets you configure the guest; it does not give you access to the physical host. Our Debian hardening guide covers the first account and firewall changes.
Start with the bottleneck, not the biggest plan
More CPU does little for a single process waiting on a remote API. Extra RAM helps a database keep frequently used data in memory, but it will not repair an inefficient query on its own. A larger disk gives you room for uploads and logs; it does not guarantee a particular number of storage operations per second.
Write down the workload before choosing: how many applications, expected concurrent users, database size, upload growth, backup space, and the busiest hour. Then identify whether it is mostly single-threaded, parallel, memory hungry or network dependent. If you already have a server, collect evidence with our slow VPS troubleshooting guide.
A useful sizing decision has two parts: enough resources for the normal workload and enough spare capacity for predictable peaks. There is no honest universal visitors-per-vCPU number. A cached static page and an uncached checkout can consume very different amounts of work for one visit.
What the RS Computers virtual plans actually change
| Plan | vCPU | RAM | NVMe storage | Port |
|---|---|---|---|---|
| VPS Nano | 1 shared | 1 GB | 20 GB NVMe | 1 Gb/s |
| VPS Micro | 2 shared | 2 GB | 40 GB NVMe | 1 Gb/s |
| VPS Mini | 4 shared | 4 GB | 80 GB NVMe | 1 Gb/s |
| VDS Small | 4 shared | 8 GB | 240 GB NVMe | 10 Gb/s |
| VDS Medium | 8 shared | 16 GB | 480 GB NVMe | 10 Gb/s |
| VDS Large | 16 shared | 32 GB | 960 GB NVMe | 10 Gb/s |
Read the whole row rather than choosing by vCPU count alone. For example, VPS Mini and VDS Small both have four shared vCPUs in this lineup, but their RAM, storage and port allocations differ. A move between those two plans addresses a different requirement from simply adding more CPU threads.
The port speed describes the plan's network interface capacity. Transfer speed still depends on the destination, route, protocol and contention. A 10 Gb/s port does not make a database query finish ten times faster or guarantee a 10 Gb/s transfer to every remote server.
Windows Server is available on VPS Mini and the VDS plans. Choose the operating system during order configuration. A Windows desktop with several applications needs memory for the operating system as well as the applications; do not size it using a minimal Linux service's requirements.
A worked sizing example: a small business application
Suppose your own measurements suggest the following peak memory needs. These are illustrative planning numbers, not measurements of a named application or a promise about user capacity:
| Component | Planning allowance |
|---|---|
| Operating system and supporting services | 1 GB |
| Database working memory | 2 GB |
| Four application workers at about 0.25 GB each | 1 GB |
| Room for a short peak or administrative task | 1 GB |
| Total planning target | 5 GB |
A 4 GB allocation would already be below that target. An 8 GB plan would be a more sensible trial candidate, followed by measurement under representative load. This does not mean every small business app needs 8 GB; replace every row with evidence from your application. Shared libraries and caches can complicate process-memory accounting, so also measure the server's total peak working set.
Repeat the exercise for disk space. For example, 48 GB of current data plus six months of growth at 5 GB per month plus 30 GB for a temporary export comes to 108 GB before extra free space. An 80 GB disk will not fit that plan. Database imports, software upgrades and backup creation can need temporary space beyond the steady data size.
CPU sizing needs a different experiment: run a realistic transaction or job at the expected concurrency. If one request is slow on one fully busy thread, doubling the number of available threads may help concurrent requests without speeding up that individual request. Ask about the CPU generation and measure the actual job.
Four workloads and how we would approach them
A small website, private VPN or bot
Start by checking the software's documented memory requirement and leaving space for updates and logs. A small VPS may be enough. Keep an eye on memory and disk growth before paying for a much larger server. If you plan to put a database and several containers on the same machine, add their requirements rather than considering each one in isolation.
A business application with a database
A VDS can make sense when the database and application need more RAM. Measure query latency, active working set and peak connections. Put the application near the database and test a representative transaction. Our database hosting guide covers the operational choices in more depth.
A Windows remote desktop
Count the programs and sessions you intend to run, then check application and Microsoft licensing requirements separately. Extra RAM can help with several open applications. A server in a suitable location can improve responsiveness more than a CPU upgrade when network delay is the problem. Start with the Windows VPS guide.
A sustained compute workload or a virtualization host
This is where a physical server may deserve investigation. Continuous parallel processing can make exclusive hardware useful; running your own virtualization layer also introduces host-level requirements. Confirm the exact CPU, memory layout, disk redundancy, recovery access and replacement arrangements. Bare metal shifts more operational responsibility to the person running it.
Questions to settle before ordering or upgrading
- What is shared? Ask about CPU allocation and the network policy. A dedicated IP address does not imply an exclusive CPU or uplink.
- What changes during an upgrade? Confirm the expected interruption, how the new resources become visible, whether the disk can grow, and whether you need to extend the partition or filesystem inside the guest.
- What is included in management? Clarify who updates the operating system and application, monitors failed jobs and restores data. Root access by itself is not an application-management agreement.
- How does recovery work? Confirm backup coverage, retention, restore access and how to reach support when normal login fails.
- What is the next constraint? Check your expected data growth and peak workload so the new plan fixes more than today's last few megabytes.
For a measured application-specific example, the n8n requirements guide shows why a workload test is more useful than a generic plan recommendation. Keep networking dependencies in the same assessment: a faster guest cannot remove latency in an external API it must wait for.
Backups and recovery belong in the comparison
A bigger server is not a backup strategy. RS Computers VPS/VDS plans include weekly full backups, but an application that cannot lose a week's worth of changes needs additional backups at an appropriate interval. Decide how much recent data you can afford to lose and how quickly the service must return.
Test a restore onto a separate instance. A backup file that exists but has never been restored is incomplete evidence. The off-site backup guide explains practical approaches. Check backup scope, retention and restore arrangements before relying on any provider's headline description.
Use a representative trial
Install the application, load synthetic or properly protected test data, and run the actual user journey. Record response times at a realistic concurrency level. Repeat at different times because shared resources and external services can vary.
Compare the result to a requirement such as a report finishing within a working minute or a checkout responding within your chosen target. Avoid buying around a benchmark number disconnected from your application. See the VPS and VDS specifications and ask our team if you need help matching the workload.
Frequently asked questions
Does VDS mean dedicated CPU?
Not necessarily. Product names vary by provider. RS Computers VPS and VDS plans use shared vCPUs; choose based on the published specifications and resource policy.
Is a dedicated server always faster?
No. CPU generation, single-thread performance, storage and the application all matter. A whole older machine can be slower for a particular task than a suitably sized modern virtual machine.
Should I upgrade RAM or CPU first?
Measure first. Sustained memory pressure and active swapping point toward a memory problem. Busy application threads may point toward CPU, while slow storage or external requests need a different response.
Is a VPS suitable for production?
Yes, when its resources, recovery arrangements and operating practices meet the application's requirements. A production service still needs monitoring, updates and tested backups regardless of whether it runs in a virtual machine or on physical hardware.
Does dedicated IPv4 mean a dedicated server?
No. An IP address assigned to your VPS is a networking resource. It does not change the machine into bare metal or make its shared vCPUs exclusive.
Can I upgrade instead of migrating to another server?
That depends on the provider and the type of resource change. Ask about the supported upgrade path, downtime and guest-side disk steps before planning the work. Keep a current backup even when the upgrade is performed in place.