There are multiple meanings of "cluster", and I fear you may be confusing them...
In ArcGIS Server 10.1+ parlance, a "cluster" is a bound subset of servers within an ArcGIS Server site. The documentation contains this image, which shows how a subset of servers can be allocated to one aspect of GIS services.

In this diagram there are three (3) hosts running ArcGIS Server, all bound in one (1) site, with two (2) clusters (one for mapping services, and one for geoprocessing). Depending on the characteristics of the nodes (CPU cores & RAM), and the frequency of requests, this site could serve dozens or scores or hundreds of clients.
There are other ways to allocate services across servers, but clusters provide a convenient way to organize allocation of services to hosts in a large enterprise configuration.
In practice, you need to be careful to not have too many nodes in a single cluster, lest their internal communication impact overall throughput.
All of the nodes on which ArcGIS Server services are running must be licensed for ArcGIS.