What does scalable cloud infrastructure mean?
Scalable means an environment can grow along with demand; here you can read how that works and where the limits lie.
Scalable cloud infrastructure can grow or shrink with demand. That can be done by making one machine heavier (vertical) or by adding more machines (horizontal). Elasticity means this happens automatically and quickly.
Who is it for: For anyone building or purchasing a website, application or data environment who wants to understand what a provider means by scalability.
What is scalability?
Scalability is the ability of an environment to handle more or less work without you having to rebuild everything. An application with a hundred users should in principle also be able to work with ten thousand users, by adding capacity.
The concept says something about the infrastructure, but also about the application itself. An environment can offer ample capacity while the software is unable to use it.
Vertical and horizontal scaling
With vertical scaling you make a single machine more powerful, for example with more memory or a faster processor. That is simple, but there is a limit to what one machine can do, and a restart is often needed.
With horizontal scaling you add more machines and spread the work over them with a so-called load balancer. That goes further, but the application has to be designed for it. Data and sessions must then be shareable between the machines.
What is elasticity?
Elasticity is scaling that adjusts automatically to the load. When it is busy, capacity is added; when it gets quiet, it goes away again. It provides flexibility and you usually pay according to use.
Elasticity only works if the rules are set up properly. A late reaction causes delays, overly generous settings cause unnecessary costs. Also check whether there is an upper limit that you set yourself.
Which questions do you ask a provider?
Scalability is not a fixed property that all services deliver equally well. Test it against your own situation.
- Is scaling manual or automatic, and how quickly does it happen?
- Are there limits per account or per region, and can you raise them?
- Which components scale along: compute, storage and network, or not all of them?
- What are the cost implications of extra capacity and of data traffic?
Forms of scaling compared
| Concept | How does it work | What to look for |
|---|---|---|
| Vertical scaling | Making one machine more powerful | Upper limit per machine; sometimes an interruption is needed |
| Horizontal scaling | More machines side by side | Application must be designed for it |
| Elasticity | Automatic scaling up and down | Set good rules and limits |
| Capacity planning | Estimating in advance what you need | Remains necessary with fixed contracts or limited quotas |
Who is this less suitable for?
- A highly elastic environment is less suited to a stable, predictable load; fixed capacity is then easier to oversee.
- Horizontal scaling is less suitable if your application cannot be spread over several machines; then a redesign comes first.
- Automatic scaling is less suitable if you cannot keep an eye on costs, because spending can grow along with it.
Frequently asked questions
Is scalable the same as available?
No. Scalability is about capacity, availability is about continuing to work during failures. An environment can be one without the other.
Does my small website need to be scalable?
Usually not to a great extent. First look at whether there are peaks, such as promotions or seasonal busyness, and only then opt for more flexibility.
Does storage scale in the same way as compute?
Not in every case. Storage often grows along easily, while compute needs more coordination with the application. Ask per component how it works.
