What is a load balancer?

Load balancer: definition and role

A load balancer is an upstream component that receives visitor requests and distributes them across several identical servers. Outwardly, only a single address appears; behind it, two, ten or a hundred servers work together. This increases the total available capacity without a single machine having to grow ever larger.

It also protects against outages: the load balancer performs regular health checks and stops sending requests to a server that does not respond. It also enables maintenance without interruption, because individual servers can be taken out of the pool one after another and updated.

How a load balancer distributes load

There are several methods of distribution. With round robin, the servers take turns; with least connections, the server with the fewest active connections receives the next request. More powerful servers can be weighted higher so they take a larger share.

Technically, a distinction is made between layer 4 load balancing, which looks only at addresses and ports, and layer 7 load balancing, which evaluates the content of the request, such as the address of a page. Many load balancers also handle TLS encryption and can route users back to the same server through session persistence (sticky sessions).

Real-world example: a discount week in a DIY online store

An online retailer of DIY supplies runs its store on a single server. During a discount week with a newsletter campaign, the site becomes slow and occasionally goes down. The retailer then sets up two more servers with an identical store installation and puts a load balancer in front of them.

Requests are now spread out, and if a server needs maintenance or fails, the store keeps running. For this to work, sessions, shopping carts and uploaded images had to be stored centrally in the database and shared storage rather than on a single server. This preparation is often the biggest effort of the changeover.

Benefits, limits and what to watch for

A load balancer increases performance and reliability, simplifies maintenance and bundles encryption in one place. But it does not solve every bottleneck: if the database is the choke point, additional web servers help little. It must also not become a weak point itself, which is why it is operated redundantly or as a managed service of the provider.

Nano Concepts plans the right architecture, sets up load balancers in the cloud or on your own servers and adapts your applications so that they run reliably on several servers—including monitoring, alerting and load tests before the next promotion period, so your store can go into the peak with confidence.

• Run at least two identical servers behind the load balancer
• Configure health checks so that they test the real application
• Do not store sessions, uploads and caches locally on one server
• Run the load balancer itself redundantly or as a managed service
• Test load behavior with a load test before promotions