3 September 2026LANTEK Computers
Every South African business has a view on load shedding by now, and most have bought something to deal with it. What tends to be less considered is the difference between keeping the lights on and keeping the data intact, because those are two different purchases and the second one is cheaper.
What actually goes wrong, beyond the downtime
Losing power for two hours costs you two hours. That is the visible cost and it is the one everybody plans for.
The less visible cost is what happens at the moment the power goes. A computer that is writing to a file when it loses power does not finish the write. Most of the time the operating system copes and nothing is noticed. Sometimes it does not.
Databases and accounting data are the most exposed. A multi user data file being written to by several people is precisely the case that does not survive an abrupt stop cleanly. This is the mechanism behind a particular kind of support call: everything worked before the outage, and afterwards the accounting package will not open the file or reports a locked record that will not release. Related causes are covered in Sage on a network that was never designed for it.
Servers accumulate damage rather than failing outright. A server that is hard-stopped repeatedly, several times a week, for months, is being treated in a way no manufacturer designs for. Disks and power supplies fail earlier. The failure, when it comes, looks unrelated to the cause.
Backups that were running get cut short. If a job runs overnight and the power goes mid job, the result is an incomplete backup and, depending on configuration, possibly the loss of the previous good one. Worth checking specifically: what does your backup do when it is interrupted, and would you know?
Network equipment comes back in a random order. The router, the switch, the firewall and the server all restart at once and negotiate with each other in whatever order they happen to finish. Usually fine. Occasionally something comes up in a state that needs a person.
What to protect, in order of value for money
The instinct is to size a solution to keep everybody working. That is expensive, and it is not the first thing to buy.
One. Enough battery on the server to shut down properly. Not to keep running. To shut down cleanly. A UPS that gives a server a few minutes and, crucially, is connected to that server so it can tell it to shut itself down gracefully, converts every outage from an abrupt stop into a controlled one. This single item removes most of the data corruption risk and it is the cheapest thing on this list.
The connection is the part people miss. A UPS with no signal cable to the server just delays the same abrupt stop by a few minutes. It has to be able to say the power is going, shut down now.
Two. The network kit on battery. A router and a switch draw very little. Keeping them up means anybody with a laptop, or with power at home, keeps working, and it means the phones survive if you are on Teams Phone. This is a small spend with a disproportionate effect.
Three. Whatever the business genuinely cannot pause. Reception, one machine that can take payments, the terminal in dispatch. Identify the specific functions rather than trying to cover everybody.
Four. Laptops instead of desktops, on the next refresh cycle. Not a power purchase at all, but it is the most effective one. A laptop has its own battery. This does not need to be a project: it needs to be the default choice next time a machine is replaced.
Five. Everything else, if the business case supports it. Inverters and solar are a genuine answer to the downtime, and they are a different order of spend. Nothing wrong with getting there. Just do the four items above first, because they address the data and they cost a fraction as much.
The case for moving things off your own premises
The most durable answer to power at your building is not to depend on your building.
Mail and files in a cloud service are not affected by your power at all. Your ability to reach them is, but that is a different and smaller problem, and anybody with a charged laptop and a phone can work from wherever there is power. A business whose mail is on a server in the office loses access to mail whenever the office loses power. A business whose mail is hosted does not.
This is one of the more compelling arguments for moving to cloud services in a South African context specifically, and it has little to do with the usual reasons people give for it.
Two honest caveats. It moves the dependency to your internet connection, which needs its own battery, which is item two above. And it does not help with anything that has to run on a machine in your building, which for most businesses is still something.
Things worth checking this week
All of these are quick and all of them are things we routinely find.
Are the UPS batteries still good? They degrade, typically over a few years, and a UPS with dead batteries reports itself as fine until the moment it is needed. Most can self-test. Ask when yours were last replaced.
Is the UPS actually connected to the server? As above. Very common to find one supplying power with no signal cable.
Does the server come back on by itself? If it needs somebody to press a button, and the power returns at two in the morning, you lose the night. This is usually a setting in the machine’s firmware.
What is plugged into the UPS that should not be? Monitors and printers on a UPS meant for the server consume the runtime you were relying on.
Do backups reschedule after an interruption? Or do they simply wait until tomorrow.
The point
You cannot make the power reliable. You can make an outage boring, and the difference between boring and expensive is mostly a few minutes of battery in the right places and a cable nobody plugged in.
What we cover is on IT Infrastructure and Backup and Disaster Recovery. If your systems have been through a lot of hard stops, it is worth having somebody look at what that has done, because the damage tends to surface as something apparently unrelated.