What IT infrastructure that does not fail you looks like: the parts that matter

IT infrastructure is like the wiring in a building: you only notice it when it stops. And when it stops, the damage is not the broken hardware, it is the hours nobody can work. A 20-person office down for a day means 160 lost working hours — usually more than the redundancy you did not want to pay for.

1. The network — the part everyone underestimates

This is where most problems wrongly blamed on “the internet” begin. A few things that make the difference:

• Proper cabling. A badly crimped cable produces intermittent errors, the hardest kind to diagnose.
• Segmentation. Surveillance cameras, office computers and the guest network have no business sharing one space. If a camera is compromised, it must not reach accounting.
• Wi-Fi that was designed, not improvised. Three access points placed where the sockets happened to be cover less than two placed properly.

2. Servers — sized for reality

The most common mistake is not the server that is too small, it is the one that is too big, bought “just in case”. The second is the single server: everything runs on one machine, with no plan for the day it dies.

Virtualisation solves part of the problem, because it separates services and lets you move a virtual machine to another host. But virtualisation on a single physical server is still a single point of failure.

3. Backups — the 3-2-1 rule

Three copies of the data, on two different types of media, one of them in another location. It sounds excessive until ransomware encrypts the backup drive that was left connected to the network.

The part almost nobody does: testing the restore. An untested backup is an assumption, not a safety copy. At least quarterly, somebody should actually restore a file and confirm it opens.

4. Security — in layers, not in one product

• Updates applied on time. Most successful attacks use old vulnerabilities with a fix available for months.
• Least privilege. Not everyone needs administrator rights, even if it is more convenient.
• Two-factor authentication for everything reachable from outside.
• Separate admin accounts from the ones used daily.

5. Monitoring — finding out before the user does

The difference between managed infrastructure and infrastructure left alone is who notices the problem first. If you learn the server is full from an employee who can no longer save, you have already lost time. Monitoring temperature, free space, disk health and backup status costs little and completely changes how you respond.

6. People — the part that does not show in the quote

You can have excellent hardware and still be down for a day if you have nobody to call. In practice, a maintenance contract with an agreed response time is worth more than the difference between two server models.

Where to start, if you already have something

1. Take inventory: what hardware you have, how old it is, what runs on each box.
2. Check the backup — restore a file today.
3. Identify single points of failure: what stops if that one box dies?
4. Only then talk about purchases.

The order matters. Most companies buy hardware before knowing what they are missing, and end up with the same problems, only more expensive.

← Back to the blog