Operations Notes
An unmanaged VPS runs at whatever speed you last tuned it to
You can afford the server. The question you have not answered is who sizes the PHP worker pool, watches the kernel updates and proves the backups restore.
The short answer
Unmanaged is a job list, not a specification: the box arrives with distribution defaults, and distribution defaults are sized for a machine nobody has described, so the difference between a quick VPS and a slow one is almost always tuning nobody did.
Take unmanaged if you will genuinely do that work and enjoy it. Talk to us about management if an hour of your time is worth more than the gap between the two. Below: exactly what needs tuning, in what order, and how to check whether it worked.
By the HostingFast team · Reviewed 15 August 2026
99.9%
Uptime target
24/7
Engineers reachable
Free
TLS on every account
NVMe
Storage class
An unmanaged price does not remove the administration. It leaves it off the invoice, and it returns later as your evenings, as monitoring you never configured, or as a contractor's emergency rate on the worst possible Saturday.
This page is the honest version of that job list. Not a warning, a checklist: the specific things that need sizing on a fresh KVM box, the commands that tell you whether they are sized correctly, and the point at which paying somebody else stops being an indulgence.
What a fresh box actually needs before it is quick
A new server comes up with distribution defaults, and those defaults assume nothing about your workload. The first four things to size are the PHP-FPM pool, the database buffer, OPcache and swap. Everything else on the usual hardening list matters for safety rather than speed.
Size the PHP worker pool against RAM rather than against optimism. Measure the resident memory of a real request, divide the memory you can spare by that figure, and set the maximum children to the result. Set it too high and the box swaps under load, which is far slower than queueing would have been.
Then the database. On MySQL or MariaDB the InnoDB buffer pool wants to hold your working set; if it does not, the disk becomes the bottleneck at exactly the moment traffic arrives. Check it against your actual data size rather than against a blog post's rule of thumb.
The maintenance that is not optional
Patching on a schedule, firewall and access rules that are actually restrictive, monitoring with alerts that reach you wherever you are, and periodic proof that a backup restores rather than merely completes. That is the standing workload, and none of it is difficult — it is simply relentless.
The failure mode we see most often is not a compromise, it is silence: an unmanaged box running on optimism, unpatched and unmonitored, discovered months after the fact. Nothing announces itself, which is precisely why the monitoring has to be configured on day one rather than after the first incident.
Snapshot backups are available on demand across the VPS range, and extra SSD disks attach in seconds. Both are useful, and neither is a substitute for having restored something at least once and timed how long it took.
What our KVM boxes give you to work with
Full root, KVM virtualisation rather than containers, and a choice of 24 Linux and BSD images. Pick London or Dallas as the region, and sit behind more than 1 Tbps of DDoS filtering on hypervisors with a 20 Gbps uplink. Reboot, rebuild and console access all live in your panel.
cPanel can be bolted on whenever you want it, which is worth considering if the reason you are going unmanaged is cost rather than appetite. A panel does not make the box faster, but it removes a large slice of the routine work that was going to eat your evenings.
Deployment is immediate and there is no setup fee. The trade-off to know before ordering: VPS and dedicated servers are built to order the moment payment clears, so they sit outside the refund window that covers shared, business, WordPress and WooCommerce plans.
Deciding honestly, with a number attached
Estimate the hours. Patching and monitoring upkeep, plus the incidents you cannot schedule. Multiply by what your hour is worth. Compare that with the gap between an unmanaged box and a managed arrangement, and the answer usually stops being a matter of pride.
Then check the second cost, which is opportunity. Time spent sizing a worker pool is time not spent on the application, and application work almost always returns more than infrastructure work at small scale.
Management is a service layered on top rather than a property of the server, so starting unmanaged and adding help later is a perfectly ordinary path. Plenty of people administer their own box happily until one genuinely bad week recalculates the arithmetic for them.
The part of this that is not marketing
If you enjoy running servers, run the server. This page exists because the people who regret an unmanaged box are almost never the ones who wanted the work — they are the ones who bought a price and inherited a rota.
Either way the hardware story is the same: KVM with full root, snapshots on demand, and DDoS filtering ahead of the uplink rather than bolted on afterwards.
- Size the worker pool against real RAM
- Buffer pool against real data size
- Restore a backup, and time it
- No refund window on built-to-order servers
Why HostingFast
Standard on every plan
KVM, so the kernel is yours
Full virtualisation rather than containers, with root access and no locked doors anywhere in the stack.
24 images, two regions
Linux and BSD images ready to deploy, in London or Dallas, live the moment payment clears.
Filtering ahead of the uplink
More than 1 Tbps of DDoS filtering in front of hypervisors running a 20 Gbps uplink.
Snapshots and extra disk on demand
Take a snapshot before you change something, and attach more SSD in seconds when you need it.
cPanel when you want it
Bolt a panel on later if the routine administration turns out to be the part you did not want.
Console, reboot and rebuild
All three from your own panel, so a bad configuration change is a recovery rather than a support queue.
Quick Start
From order to online
- 1
Measure one request, then size the pool
Find the resident memory of a real PHP request under load, divide your spare RAM by it, and set the worker maximum to that. Guessing high causes swapping, which costs far more than a short queue ever would.
- 2
Match the buffer pool to the working set
Check your InnoDB data size and set the buffer pool so the hot part fits in memory. If it does not, every busy period turns into disk reads and no amount of CPU rescues it.
- 3
Restore something before you need to
Take a snapshot, restore it somewhere harmless, and time the whole operation. A backup you have never restored is a hypothesis, and the worst moment to test a hypothesis is during an incident.
Built In
Loaded onto every plan
- Full root access on KVM virtualisation, with no locked doors
- 24 Linux and BSD images, deployed the moment payment clears
- London or Dallas as your region, chosen at order time
- 1 Tbps+ DDoS filtering ahead of a 20 Gbps hypervisor uplink
- Snapshot backups on demand, with extra SSD disks added in seconds
- Reboot, rebuild and console access from your own panel
- cPanel available as a bolt-on whenever you decide you want it
- No setup fee, with the server provisioned automatically on payment
- Human support on duty every hour of every day
- Shared, business, WordPress and WooCommerce plans carry 30 days; reseller carries 7
Frequently Asked
What people ask us most often
What actually needs tuning on a fresh unmanaged VPS?
Four things move response times: the PHP-FPM worker pool sized against real RAM, the database buffer pool sized against your working set, OPcache enabled with enough memory, and swap configured so a spike degrades instead of dying. Everything else on the standard hardening list protects you rather than speeds you up.
How do I size PHP-FPM workers for the RAM I have?
Measure the resident memory of one real request under load, subtract what the database and the operating system need from total RAM, and divide the remainder by that per-request figure. Set the maximum children to the result. Setting it higher does not add capacity — it adds swapping, which is dramatically slower than waiting in a queue.
Is there a refund window on a VPS if I get this wrong?
No. VPS and dedicated servers are provisioned to order the moment payment lands, so they fall outside the guarantee. Shared, business, WordPress and WooCommerce hosting carry 30 days and reseller carries 7, which is worth knowing before you commit to running your own box.
Can I add cPanel to an unmanaged box later?
Yes, it bolts on whenever you want it. That is often the right middle path: you keep root and the flexibility that came with it, while handing the repetitive account, mail and backup administration to something that already does it well.
Keep reading
Laravel Hosting Plans
Getting a Laravel deployment onto a server without adopting platform engineering as a second job.
Cheap Dedicated Servers
Where bare metal starts making sense on price per unit of sustained throughput.
Managed Hosting (Glossary)
What management actually covers, and where the boundary sits between you and the provider.
VPS Hosting
KVM virtual servers with full root, snapshots on demand and DDoS filtering in front.
Agency Hosting
Client accounts, staging and care-plan infrastructure built for agencies running many sites.
Changing hosts? Run through our checklist first.
A straightforward sequence for a switch your visitors never feel: which files move first, how to shift email across without losing a single message, the right moment to repoint DNS, and the two mistakes behind almost all the downtime we get asked to rescue.
Root access, and a clear job list.
KVM servers in London or Dallas with snapshots, console access and no setup fee — plus a panel you can add whenever you want it.
View VPS Hosting plans