Sizing and specification
We work out how many cores, how much memory and how much disk your load really needs, with room for three years. This is also where the warranty and the response time for a hardware failure are decided.
The price on the invoice is usually the smaller part of the bill. The rest is made up of licences for the system and the database, the maker warranty, the electricity and the time it takes to look after it. So before we recommend a model we ask what is going to run on it and how many years it has to last. This page is about choosing and starting a server; looking after one that already runs is server administration.
One question always comes first and it sometimes ends the conversation about buying: does this server have to stand at your place at all. If everything meant to run on it is available as a service, your own machine is often dearer and more trouble. More often, though, something keeps the company on site: an industry system needing a local database, large design files, or simply a connection that will not carry cloud work. In wholesalers and shops the answer is usually yes, because the trading database grows every year and it is what sets the hardware replacement date.
Servers are rarely replaced because they stopped working. Far more often it is because the system running on them stopped receiving fixes.
Windows Server 2016 loses support on 12 January 2027, a few months from now. The 2019 release has until January 2029 and 2022 until October 2031. If you do not know which version you are running, that is the first thing worth checking this quarter.
The server will not stop working overnight, but it will stop receiving security fixes. From that point every new hole stays with you for good, and at audits and in customer questionnaires an unsupported system is one of the first things people write down.
Upgrading the system on the current hardware, if the machine still has capacity and warranty. A new server, when its life is ending anyway. Or moving the service to the cloud and giving up the physical box. The choice depends on what runs on the server, not on what is fashionable.
One distinction saves needless panic, because offers sometimes blur it on purpose. The 2022 release leaves mainstream support in October 2026 but keeps receiving security updates until 2031, so nothing has to be done about it. The maker also sells extended fixes for older versions, which can be a sensible bridge when an industry system will not run on a newer Windows and you are waiting for a release from its supplier.
Almost nobody now puts one service on one physical box. What has changed is the cost of the software that makes it possible.
A single physical server runs several or a dozen virtual machines: the industry database on one, files on another, the domain controller on a third. A failure of one does not touch the others, and moving a whole machine to new hardware takes hours instead of days.
Perpetual licences are gone, leaving subscriptions counted by physical core with a minimum of sixteen cores for every processor and sales only in bundles. For a company with two or three servers that means a bill rising every year with no new capability in return.
For smaller environments we usually propose Proxmox, free to use with optional support from its maker. Sometimes Hyper-V is better, because it already sits in the Windows Server licence you are buying anyway. The choice follows what you can work with.
We do not push migrations off VMware for their own sake. If you have a working environment and an acceptable renewal price, staying put is the cheapest decision, because migrating costs too. What we do advise is asking for the renewal quote early and checking two things in it: the number of cores you are paying for, and whether the seller has added an order minimum of their own on top of the maker rules. A late renewal can carry a surcharge as well.
Six stages, each with its own page. You can join at any point, for instance when the hardware is already there and only the care is missing.
We work out how many cores, how much memory and how much disk your load really needs, with room for three years. This is also where the warranty and the response time for a hardware failure are decided.
Mounting, the system, virtualisation, accounts and permissions, and moving the data off the old machine. The switchover is planned outside working hours with a way back to the previous state.
We split one machine into several environments, separate the industry system from files and the domain, and set up snapshots so an update can be undone.
Updates, accounts, watching performance and disk space, with a monthly report. This stage lasts longest and it decides whether the server reaches the end of its warranty without surprises.
A copy of the virtual machine and a test of restoring it, because with a server what counts is not whether a backup exists but how many hours until the company works again.
Some services no longer need a machine in your server room. We compare the three year cost of both routes rather than assuming up front that your own server is cheaper.
What we do not do: sell you a server meant to last ten years, because no such hardware exists and an oversized machine only means a larger invoice and larger licences. Nor do we argue for a particular maker. The hardware you can buy through us or on your own against a specification you get in writing.
The longest part is not the mounting, it is waiting for the hardware. So the order goes in only once the configuration has been worked out.
We write down which systems have to work, how many people use them and how the data grows. We also check the software makers requirements, because those set the floor for the configuration.
You get a specification with a reason for every line and the licences counted, because with servers they can flip which hardware pays off. We also give the cost of running it for three years, not just the purchase price.
Lead times on servers run longer than on office hardware, so we settle them before ordering. Installation, data migration and the switchover happen in a window when the company is not working.
We leave the documentation, the credentials and a description of what runs where, so you are not tied to one company. Further care we can take on a retainer or leave with your own IT team.
A server in a law firm and a server in a production hall differ not in power but in what is demanded of continuity and access to data.
The pages that come back when a server is discussed.
The hardware is often a smaller item than the licences that have to go with it, so comparing invoice prices alone leads you astray. We therefore quote the whole thing: the machine, the system, the access licences, virtualisation and the warranty, with the cost of running it for three years alongside. The sizing itself is free.
First check whether the software running on it works on a newer version, because that is usually what blocks the upgrade. If it does, the decision comes down to whether the hardware still has capacity and warranty. If not, there is paid extended fixes as a bridge or moving the service elsewhere. Worth starting now, because server lead times can run to several weeks.
It depends on what has to run on it. Mail and documents have long worked out better as a service, but an industry system with a local database, large design files or a weak connection can tip the balance towards your own hardware. We cost both routes over three years, licences and labour included, and show the difference rather than recommending up front.
With a small environment often yes, because after the licensing changes a subscription counted by physical core with a minimum of sixteen per processor can be out of proportion to two or three servers. Migration is also a cost and an outage, so we decide only after setting the renewal price against the cost of moving. Sometimes staying turns out cheaper.
We usually plan for five years, which is the span of a sensible maker warranty with next business day attendance. After that the hardware still works, but spare parts grow dearer and harder to find, and a failure starts meaning days of downtime instead of hours. So the replacement is worth having in the budget before it becomes urgent.
The disks in a server work in an arrangement that survives one of them failing, and backup power buys time for a safe shutdown. That does not replace a backup, though, because no disk arrangement protects against files being deleted or encrypted by ransomware. A copy and a test of restoring it are part of every rollout.
Yes, and we do it regularly. We start with a review that ends in a description of the actual state: what runs on it, how it is patched, how the backups look and where the credentials are. You get that document whether or not you decide to work with us further.
The first meeting takes 30 minutes and commits you to nothing. We will tell you straight whether we can help and what we would propose.
2019 - 2024 Interactive Workspace Ltd. - All rights reserved.