The way a computer uses 'load balancing' to handle website traffic

Think of a massive website like a popular seaside cafe on a bank holiday. If every tourist shoved their order at the same poor waiter, the whole place would collapse into a heap of spilled tea and tears.

Load balancing is the digital equivalent of a sharp-eyed manager at the door. Instead of letting one server drown in requests, it catches every incoming click and shuffles it off to whichever computer is currently twiddling its thumbs.

It’s all about avoiding that unnecessary faff. By spreading the weight, the system ensures no single machine catches fire while the others sit around doing nothing.

Wait, how can the manager tell who’s actually twiddling their thumbs?

The manager isn't just guessing; they’re keeping a tally. In a basic setup called 'Round Robin,' they simply go down the line: Waiter A, then B, then C, regardless of how many heavy trays they're already carrying.

A sharper manager watches closer, checking who has the shortest queue or even shouting, 'Are you still alive?' every few seconds. If a waiter doesn't shout back because they've tripped over a seagull, the manager stops sending them orders.

It’s all about constant pestering to ensure the work gets done without a nasty pile-up.

But what if one waiter is just much stronger than the others?

Spot on. In the tech world, we call this 'Weighted Load Balancing.' It’s admitting that some computers are shiny new super-machines while others are dusty old boxes from 2010.

The manager gives each waiter a score based on their muscle. If Waiter A is a champion lifter, they get the heavy lifting, while the trainee gets the light tea and biscuits.

It prevents the 'faff' of a slow machine trying to do a big job, ensuring the whole cafe runs at the speed of the fastest, not the slowest.

Does the manager just assume the strong ones never get tired?

Even a champion lifter has limits. If you give the strongest waiter a table of twenty people ordering complicated cocktails, they’ll be stuck for ages while the trainee breezes through ten simple teas.

We use 'Least Connections' to fix this. The manager stops obsessing over raw power and looks at active tasks. If the super-machine is already 'busy' with a massive data crunch, the manager won't pile more on.

It’s about staying nimble. You watch who’s actually drowning in work to keep the faff to a minimum.

Can the manager make sure a customer stays with the same waiter?

That’s a brilliant point. If you’re halfway through ordering a massive seafood platter and the manager shuffles you to a new waiter, you’d have to start your story from scratch. That’s a right faff.

To stop this, we use 'Sticky Sessions.' The manager gives you a little numbered ticket—like a cloakroom tag—so they remember exactly which waiter was already looking after you.

It ensures you stay with your original server until you're done. Without it, the website would 'forget' you’re logged in or empty your shopping basket every time you clicked a new page.