Packet Switching

Packet switching sends data in small, discrete packets that can be routed independently across a shared network. Unlike circuit switching (dedicated path for the whole session), packets from many flows share links; each packet carries addressing info and is forwarded hop-by-hop. The network can use capacity flexibly—no single call monopolises a path—so aggregate throughput scales with statistical multiplexing. Tradeoff: variable latency and need for buffering and congestion control when many flows compete. The basis of the internet and most digital communication; the same idea applies to task queues, batch jobs, and any system that multiplexes work over shared resources.

When demand exceeds a single path's capacity, packetise: break work into units, add routing or routing keys, and let a shared fabric deliver them. Design for variable latency (buffers, timeouts) and congestion (backpressure, load shedding). In product and ops: think in "packets" of work that can be distributed and parallelised instead of one long-lived pipeline.

Packet switching sends data in discrete, routable units over shared links, enabling statistical multiplexing and scalable throughput. Variable latency and congestion require buffering and control. Use it when building systems that multiplex work or data over shared capacity.