What Is Packet Switching?

Packet switching is a method of digital data transmission in which messages are broken into smaller units called packets, each routed independently through a network and reassembled at the destination. It contrasts with circuit switching, which establishes a dedicated physical path for the entire duration of a communication session. Packet switching emerged from research in the early 1960s and became the foundational mechanism of the modern internet, enabling efficient sharing of network bandwidth among many simultaneous users.

The concept was developed independently by Paul Baran at RAND Corporation and Donald Davies at the UK National Physical Laboratory during the mid-1960s. DARPA subsequently funded the ARPANET, which became the first large-scale operational packet-switched network, with its first successful test in October 1969. The history of how packet switching evolved from message switching and early telegraphy concepts is surveyed in the IEEE paper on the evolution of packet switching, which traces technical lineage from the 1960s through the late 1970s.

Packet Structure and Forwarding

Each packet consists of a header and a payload. The header carries addressing information, sequence number, and error-detection fields that routers use to determine where to send the packet next. The payload contains a fragment of the original data. When a router receives a packet, it examines the destination address in the header and forwards the packet along the best available path toward that destination. Because packets from the same message may take different routes, they may arrive out of order, and the receiving system must resequence them before delivering data to the application. This independence of routing allows the network to route around congested or failed links without interrupting the data transfer.

Store-and-Forward and Virtual Circuit Models

Two principal modes of packet switching have been used in practice. In the store-and-forward model, each intermediate node receives an entire packet, buffers it briefly, and then forwards it when the next link is free. This approach is simple but introduces queuing delay at each hop. In virtual circuit switching, the network establishes a predetermined path through the network before any packet is sent, and all packets follow that path in order. Virtual circuits combine some properties of circuit switching (predictable delay, ordered delivery) with the statistical multiplexing benefits of packet switching. Frame Relay and ATM (Asynchronous Transfer Mode) are notable virtual circuit technologies that were widely deployed in wide-area networks during the 1990s.

Internet Protocols and Packet Switching

The TCP/IP protocol suite, standardized through IETF RFC 793 for TCP and RFC 791 for IP, codifies how packets are formatted, addressed, and delivered across interconnected networks. IP (Internet Protocol) handles addressing and routing at the network layer, while TCP (Transmission Control Protocol) provides reliable, ordered delivery at the transport layer by acknowledging received packets and requesting retransmission of lost ones. UDP (User Datagram Protocol) offers a lighter alternative for applications where speed takes priority over reliability. These protocols collectively turned packet switching from a laboratory experiment into the universal data-transfer mechanism for the global internet, as DARPA's ARPANET development demonstrated when TCP/IP was adopted as the standard in 1983.

Applications

Packet switching has applications in a range of fields, including:

• Internet data communications for web, email, and streaming services
• Voice over IP telephony and video conferencing
• Wide-area corporate networks and virtual private networks
• Mobile cellular data networks including 4G LTE and 5G
• Industrial control systems and remote SCADA communications