If a network is described as “scalable”, it can adapt to changing performance requirements. Hyperscale servers are small, simple systems that are designed for a specific purpose. To achieve scalability, the servers are networked together “horizontally”. This means that to increase the computing power of an IT system, additional server capacity is added. Horizontal scaling is also referred to as “scaling out”.
The alternative solution – vertical scaling or scaling up – involves upgrading an existing local system, for example, by using better hardware: more RAM, a faster CPU or graphics card, more powerful hard drives, etc. In practice, vertical scaling usually comes first. In other words, local systems are upgraded as far as is technically feasible, or as much as the hardware budget permits, at which point the only remaining solution is generally hyperscale.