The scale of network defines how many interfaces the network edge has. Ability to contain specific characteristics while network size is increasing is called scalability. This characteristics includes cost per port, reliability, traffic, and performance.
On one hand network scale can be increased simply by adding new hardware to the network (aggregation switches, core switches, network ports). However, cost-efficient scaling presumes control over the cost per port while increasing the scale. In real world scenario, traditional networks have tendency to scale power and cost faster than other fundamental network properties. If the network is running modular equipment and software it allows to use dynamic scalability. Dynamic scalability allows changing network’s scale while traffic is running.
While original Ethernet allowed only one connection between switches, which was manually configured, development of Ethernet architecture overcome that limitation. It is achieved with the help of automated discovery, which detects new equipment and learns about all connected devices in the network. If the traffic is load balanced, then in case of faulty link, all other links would be unaffected.
Since growing requirements need to be filled, modern data centers are created with the thought of the future size increase. This allows to maintain performance in future while achieving better economics. Economics of scale implies that big data center is better than multiple small ones. Applications are more efficient when computing equipment is tightly coupled in single data center, than multiple data centers use collective performance.