Replace the server, or move to the cloud? The honest trade offs

There is a version of this conversation that is really a sales conversation. Hardware resellers make money selling hardware. Cloud partners make money on consumption. Neither fact makes anyone dishonest, but both mean you should be wary of anyone who answers this question before asking what your server does all day.

So start there, because the answer changes completely with the workload. A file server holding large drawings behaves nothing like a database behind an application, which behaves nothing like a domain controller handling logins. Most small businesses have all three on one box and think of it as the server, singular. It is usually three decisions, not one.

List every job that machine performs and who depends on it. It is half an hour of work and it makes the rest of the decision far easier.

The usual test people apply is a broadband speed test, and it is the wrong test. Download speed tells you how quickly a large file arrives once it starts arriving. What makes cloud systems feel fast or slow, minute by minute, is latency: the time one small message takes to reach the other end and come back.

Applications that chat constantly with a database are the ones that suffer. On a local network the round trip is a fraction of a millisecond and the software feels instant. Move the database to a data centre and each of thousands of small exchanges takes tens of milliseconds. The link is not saturated, the speed test looks fine, and the software feels like wading through treacle. It is the most common reason a cloud migration disappoints, and it is predictable if anyone checks.

Upload matters too, and it is the number people forget. A lot of business broadband is still heavily asymmetric, so a line that downloads quickly may crawl uploading, and once your systems are in the cloud most of what you do all day is uploading.

Timing matters as well. A line that is comfortable at eight in the morning can be at its limit at half past eleven when everyone is on video calls. Get a fortnight of monitoring before deciding, because guessing at this is expensive.

Licensing is dull and it moves the numbers more than anything on the hardware quote. It is also where a lot of cloud business cases fall apart quietly, three months after they were approved.

Two things to check. First, whether the licences you already own can move. Some software bought outright can be transferred to a cloud environment, some cannot, and some can only move under conditions that are easy to breach without noticing. Second, whether the vendor of your main application charges differently for hosted use. It is not unusual for the same software to cost more per user once it runs somewhere other than your own hardware.

The other half is the running cost model. On premises you pay a large amount every five or six years and little in between. In the cloud you pay a moderate amount monthly, forever, and it rises when you use more. Neither is inherently cheaper. What differs is which shape suits your cash flow and how confident you are about the next five years.

Compare like for like over the same period. A fair comparison sets hardware, licences, backup, power, support and the replacement at the end against the monthly cloud cost including storage, backup, egress and support. Comparisons tend to leave several of those out, and usually on the same side.

This is the question owners actually care about and the one that gets the airiest answer. The honest position is not that it never happens. It is that you decide in advance what you do when it does.

With everything on premises, an internet outage is annoying. Email stops, anything web based stops, but people can still open files and use the local application, so the business limps on. With everything in the cloud, an outage stops work completely. Same event, very different day.

The fix is not to avoid the cloud. It is to accept that the connection has become the thing your business runs on and treat it accordingly. That usually means a second line from a genuinely different provider, with a router that fails over automatically rather than needing someone to plug a cable in. Often it means a mobile broadband fallback, which will not carry a full working day but keeps phones, mail and a few critical people going. And it means knowing which staff can work from home on their own connections.

Whatever you choose, test it. Unplug the main line on a quiet afternoon and see what keeps working. Almost everyone who does this finds something they assumed was covered and is not.

The pure versions of this decision are rare. Everything in the cloud makes sense for a business whose people are spread across homes and sites, whose applications are already web based, and whose files are ordinary documents. Everything on premises makes sense when a specialist or older application will only run locally, when people move very large files all day, or when a regulator or a customer contract says so.

Most businesses sit in between, and the sensible design reflects that. Mail, documents and collaboration go to the cloud, where they already work well. The one application that hates latency, or whose vendor will not support it hosted, stays on a properly specified local machine that is supported, monitored and backed up in more than one place. Backups go somewhere neither of those can reach.

You can get most of the way to the right answer without spending anything, and doing so protects you from whichever way the next proposal happens to lean.

We do both, which is the only reason our opinion here is worth anything. Sometimes the right answer is a new machine in the cupboard, properly specified and actually monitored this time. Sometimes it is a staged migration. Often it is some of each.

If you have a quote in front of you and are not sure it is the right shape, send it over with a sentence about what the server does. We will tell you what we would do and why, including when the answer is to stay put for another two years.