"non-blocking architecture"

 

It's a popular marketing term that refers broadly to the ability of a switch to handle independent packets simultaneously. For example, suppose a packet is traveling from port A to port B when a new packet arrives on port C. A ``nonblocking'' switch will accept and process the new packet before it completes the previous transfer. If the new packet is destined for port A or B (which are currently busy), the switch will queue the incoming packet until the destination port becomes available. Of course, the queue is finite; even a nonblocking switch must eventually reject packets (i.e., must eventually block).