A VPS is a virtual machine with a fixed bundle of vCPU, memory, disk and transfer, plus full root or administrator access. It is the simplest step up from shared hosting when you need your own software stack. Size it from measured CPU, memory pressure and disk wait, not from plan names, and treat backups as a separate decision.
What a VPS gives you
A virtual private server is a slice of a physical host, carved out by a hypervisor (typically KVM on Linux platforms). Inside the slice you have a normal server: your own kernel, your own operating system, your own firewall, and root access to install anything. Outside it, the host's CPU cores, memory bandwidth and storage are shared with other virtual machines.
That sharing is what makes a VPS affordable, and it is also its main risk. On most VPS products vCPUs are scheduled onto physical cores alongside other tenants. When the host is busy, your virtual CPU waits; Linux reports that wait as steal time. If consistency matters more than price, a virtual server with reserved allocation is the next step up.
BroadNet Cloud quotes VPS hosting on established supplier platforms. The quote names the supplier and region, the exact vCPU, memory, disk and transfer, whether CPU is shared or dedicated, and what backups are included. There is no instant checkout today: you request a quote and we provision after you accept.
Who a VPS suits, and who it does not
A VPS suits you if
- You need software shared hosting will not allow: a specific runtime, a queue, a container engine, a custom daemon.
- Your load is steady and one server can carry it.
- Someone on your team is comfortable with SSH, updates and a firewall, or you add management.
- You want a predictable single line on the bill.
Look elsewhere if
- You only run a website and email and do not want to manage a server: shared hosting.
- You need several machines, detachable volumes or scale-out: cloud hosting.
- Latency-sensitive work cannot tolerate CPU steal: virtual server or bare metal.
- Regulated data where you need to evidence controls: start at secure hosting.
What you can specify
| Item | Options we quote | What to check |
|---|---|---|
| vCPU | 1 to 16, shared or dedicated class where offered | Ask the CPU generation and whether vCPU means a thread or a core. |
| Memory | 2 GB to 64 GB | Is memory reserved, or can the host reclaim it (ballooning)? |
| Disk | SSD or NVMe-backed, size per configuration | Local or network storage; any IOPS cap. |
| Operating system | Current Linux LTS images; Windows Server where licensed | Windows licensing is a separate line Confirmed at quote |
| Access | Root or administrator, SSH keys, out-of-band console where the platform provides one | A console matters when a firewall rule locks you out. |
| IPv4 | One included per server is common; extras on request | Each extra IPv4 is usually charged. |
| IPv6 | Where the platform supports it | Ask for the prefix size you will be routed. |
| Reverse DNS | PTR records set on request | Required if the server sends mail. |
| Transfer | Monthly allowance and port speed | Overage rate and how it is measured. |
| Snapshots and backups | Optional, schedule and retention chosen by you | Confirmed at quote |
Reference configurations
These shapes are starting points for sizing, not stocked plans. Disk sizes are typical; any can be changed in the quote.
| Shape | vCPU / memory | Disk | Typical fit | Price |
|---|---|---|---|---|
| Small | 1-2 vCPU / 2-4 GB | 40-80 GB | A single WordPress or Laravel site, a VPN endpoint, a small API | Price shown at quote |
| Medium | 2-4 vCPU / 4-8 GB | 80-160 GB | An app with its database on the same server, a store with moderate traffic | Price shown at quote |
| Large | 4-8 vCPU / 8-16 GB | 160-320 GB | Several apps in containers, a busy database, CI runners | Price shown at quote |
| Extra large | 8-16 vCPU / 16-32 GB | 320-640 GB | Analytics jobs, larger databases, game or media servers | Price shown at quote |
| Memory-heavy | 16 vCPU / 64 GB | Sized to data | In-memory caches, search indexes, databases whose working set must fit in RAM | Price shown at quote |
Sizing from signals, not plan names
The right size is the smallest one where these signals stay quiet at your busiest hour. Measure on your current server for at least a week before choosing; the VPS sizing tool turns the numbers into a shape, and the sizing guide explains each signal in depth.
| Signal | Where to read it | What it means | Usual fix |
|---|---|---|---|
| CPU steal (st) | top, vmstat 1, mpstat | Your vCPU was ready to run but the host gave the time to someone else. | Persistent steal is a host problem, not a size problem: ask for a move or a dedicated class. |
| High user CPU | top, mpstat -P ALL | Your own code is busy. | More vCPU, or profile the hot path first. |
| I/O wait (wa) | vmstat 1, iostat -x | CPU idle while waiting for disk. | Faster storage tier, more RAM for page cache, or fix queries. |
| Memory pressure | /proc/pressure/memory, free -m | Tasks stalled waiting for memory to be reclaimed. | More memory; check for leaks. |
| Swap in/out (si/so) | vmstat 1 | Active memory is being pushed to disk. | More memory. Sustained swapping ruins latency. |
| Load average above vCPU count | cat /proc/loadavg, top | Runnable or disk-blocked tasks queueing. | Split by the signals above to see which resource. |
Worked example: reading vmstat
A 2 vCPU server shows, at peak, us 55 sy 10 id 5 wa 25 st 5 and memory PSI avg60 near zero. CPU is nearly fully used, but a quarter of it is waiting on disk and memory is fine. Doubling vCPU would help little. Moving to faster storage, or adding RAM so the database's hot pages stay in cache, attacks the 25 % wait first. If the same server showed st 20, the fix would be a different host or a dedicated CPU class, not a bigger plan.
Operating system images and root access
We quote current long-term support releases of mainstream Linux distributions, such as Ubuntu LTS, Debian stable and RHEL-compatible distributions like AlmaLinux or Rocky Linux, plus Windows Server where the supplier platform licenses it. Pick a release with years of security support left; an image near its end of life means a forced upgrade soon after launch. Exact images per region are confirmed in the quote.
Root access means the server is yours to configure, and yours to secure. On a self-managed VPS you own:
- Applying security updates, including kernel updates and the reboots they need.
- SSH configuration: keys instead of passwords, no direct root login over SSH.
- A host firewall that only opens the ports you serve.
- Logs, monitoring and alerts that tell you before your users do.
The Linux hardening guide gives a baseline you can apply on day one.
IPv4, IPv6 and mail
Public IPv4 addresses are scarce and usually priced per address, so plan how many you need: most servers need one, and TLS with SNI lets many HTTPS sites share it. IPv6 is plentiful; where the platform supports it, ask what prefix is routed to your server so you can give services their own addresses.
If the server will send email, three things matter: a reverse DNS (PTR) record matching the hostname, SPF, DKIM and DMARC records for your domain, and whether outbound port 25 is open. Many platforms block port 25 by default to stop spam, and opening it can require a request. Ask before you plan a mail server; a transactional email service is often the simpler route.
Backups and snapshots as quoted options
Neither is assumed. Each is a line in the quote with its own schedule and retention, so you can see what you are and are not buying.
- Snapshot: a point-in-time image of the whole disk, fast to take and restore, usually kept on the same platform. Take one before every upgrade.
- Backup: a copy kept separately, on a schedule, with retention you chose, and ideally a copy off the platform. This protects you from deleted servers and compromised accounts.
Databases need application-consistent dumps alongside disk images. Whatever you choose, test a restore before you need one: a backup that has never been restored is a hope, not a plan. See the backup and recovery guide.
Use cases
Moving off shared hosting
A site that hits shared-hosting process limits or needs a newer runtime moves to a Small or Medium VPS. The reason is control: you set PHP workers, cache sizes and cron jobs yourself.
An internal tool or API
A line-of-business app with a modest database fits comfortably on one Medium VPS. One server is easier to reason about than a cloud layout, and a nightly backup off the platform covers most risk.
Containers on one host
A Large VPS running a container engine can host several small services with a reverse proxy in front. Watch memory pressure: containers make it easy to overcommit your own server.
Development and CI runners
Build jobs are bursty and CPU-heavy. Watch steal time during builds; if it is high, a dedicated CPU class pays for itself in shorter queues.
VPS compared with the alternatives
| Question | Shared hosting | VPS | Virtual server | Cloud hosting |
|---|---|---|---|---|
| Root access | No | Yes | Yes | Yes |
| Who patches the OS | Provider | You, unless managed | You, unless managed | You, unless managed |
| CPU consistency | Shared with many accounts | Shared; watch steal | Reserved, not overcommitted | Depends on class |
| Scaling | Bigger plan | Bigger plan | Bigger reservation | Resize or add instances |
| Best for | Sites and email | One steady server | Steady load needing consistency | Multi-machine, variable load |
Responsibility matrix
| Task | BroadNet | Supplier | You |
|---|---|---|---|
| Host hardware, hypervisor, network | - | Owns | - |
| Sizing advice and the quote | Owns | - | Approves |
| Provisioning and first access | Owns | - | - |
| OS updates and reboots | - | - | Owns |
| Firewall, SSH, user accounts | - | - | Owns |
| Snapshots and backups (if ordered) | Sets up | Stores | Tests restores |
| Reverse DNS and extra IPs | Requests | Applies | - |
| Application and data | - | - | Owns |
| Host faults and supplier escalation | Owns | Fixes | Reports |
Want BroadNet to own updates, monitoring and restores? That is managed hosting, and the matrix changes in writing.
Support terms
Your quote states the contact routes, staffed hours, severity definitions and response targets for your contract, plus how planned maintenance is announced and the renewal and cancellation terms. We do not publish a one-size response time. On a self-managed VPS, support covers the platform, access, networking and billing; help inside your operating system is available as quoted work or through managed hosting. See support.
How buying works
- Measure. Collect a week of CPU, steal, I/O wait, memory and transfer from your current server, or estimate with the sizing tool.
- Request a quote. Use the configurator on pricing or contact sales.
- Review the written quote. Supplier, region, shape, CPU class, image, addresses, backup options, support terms and price.
- Provision. After acceptance we build the server and send access details securely.
- Resize if needed. After a few weeks of real traffic, check the same signals and adjust.
Frequently asked questions
What is CPU steal and how much is too much?
Steal is time your virtual CPU was ready to run but the hypervisor ran something else. Occasional low single-digit steal is common on shared hosts; steal that stays high at your peak hour means the host is contended, and the fix is a different host or a dedicated CPU class.
Can I upgrade a VPS later?
Usually yes: moving to a larger shape is common and often needs a reboot. Shrinking a disk is harder than growing one, so start disk sizes modestly. The resize terms are in your quote.
Do I get a control panel?
A VPS comes as a plain server. You can install a hosting control panel yourself, and some panels need a paid license. If you mainly want a panel, shared hosting or managed hosting may be a better fit.
Are backups included?
Only if they are in your quote. Snapshots and backups are separate options with their own schedule and retention, so you can see exactly what is covered.
Can I run Windows Server?
Where the supplier platform licenses it, yes. Windows licensing is a separate cost and is confirmed in the quote.
How quickly can I get a VPS?
BroadNet Cloud is quote-led, with no instant checkout yet. Timing depends on the supplier and region and is stated when we send the quote.
Sources
- proc(5): /proc/stat fields including steal, Linux man-pages
- vmstat(8) manual page, Linux man-pages
- PSI: Pressure Stall Information, Linux kernel documentation
- RFC 6066: TLS Extensions, Server Name Indication, IETF
- Ubuntu release cycle, Canonical
Related
Last reviewed . Facts on this page are checked against the sources listed; anything not yet contracted is marked “confirmed at quote”.