What does scaling mean (horizontal and vertical)?

Scaling: definition and meaning

A system is scalable if it does not buckle under growing usage but handles more load with additional resources. This applies to websites as much as to online stores, customer portals or software-as-a-service products. Good scalability also means releasing resources again when demand falls and thereby saving costs.

There are two ways to do it. Vertical scaling enlarges the existing server: more processor cores, more memory, faster disks. Horizontal scaling spreads the work across several identical servers that cooperate. Both ways have advantages and drawbacks, and in practice they are often combined.

How vertical and horizontal scaling work

Vertical scaling is simple because the application hardly changes. However, there is an upper limit, larger servers become disproportionately expensive, and the switch often requires a restart. The server also remains a single point of failure as long as no second instance is on standby.

Horizontal scaling has hardly any limit and increases reliability, but it requires a suitable architecture: a load balancer distributes requests, and the application must not keep data locally on one server. With autoscaling, cloud platforms start additional instances automatically under load and shut them down again afterwards.

Real-world example: ticket sales for a city festival

A regional organizer sells tickets for a city festival. When advance sales open at ten o’clock, thousands of interested people visit the site within minutes, and then demand subsides. A server sized permanently for that would cost money all year round; a small one collapses on the big day.

With horizontal scaling, the web servers run as several instances behind a load balancer. They are started in advance for the sales launch or added through autoscaling and reduced again in the evening. Important: the database that allocates the tickets must be able to grow too, otherwise the bottleneck simply shifts.

Benefits, limits and common mistakes

Scaling creates reserves for growth and load peaks but costs money and planning. A common mistake is to enlarge only the web servers while the database, interfaces to third-party systems or inefficient code are the real bottleneck. Equally expensive: paying permanently for a peak load that occurs only a few hours a year.

Nano Concepts analyzes where your system actually hits its limits, plans suitable scaling strategies and implements load tests, autoscaling and optimizations—so that performance and cost stay in sensible proportion and your application can grow with the business without being permanently oversized.

• Measure the actual bottleneck before scaling up
• Build the application so that it needs no local session data
• Include the database, cache and file storage in the planning
• Run load tests before expected peaks
• Set cost limits for automatic scaling