What Is a Packet?

A packet is a formatted unit of data transmitted across a network, carrying both the information to be delivered and the control metadata needed to route and reassemble it at the destination. Packets are the fundamental transmission units of packet-switched networks, including the Internet, where they travel independently through a network of routers and may take different paths before being reassembled at the receiving end. The packet model contrasts with circuit-switched communication, where a dedicated channel is held open for the duration of a session; packet switching instead allows network resources to be shared dynamically among many concurrent flows.

The packet concept emerged in the early 1960s from work by Paul Baran at RAND Corporation and Donald Davies at the UK National Physical Laboratory, and was adopted as the basis for ARPANET in the late 1960s. Today, the Internet Protocol (IP) defines the standard packet format used on the global Internet, with RFC 791, the IETF specification for the Internet Protocol, establishing the IPv4 packet structure that has carried the majority of Internet traffic for decades.

Packet Structure and Header Fields

Every packet consists of a header and a payload. The header contains control fields that enable routers and receiving hosts to process and deliver the packet correctly. In an IPv4 packet, the minimum 20-byte header includes the source and destination IP addresses, a total length field, a time-to-live (TTL) counter that is decremented at each hop to prevent infinite routing loops, a protocol field identifying the upper-layer protocol (TCP, UDP, ICMP, or others), and a checksum over the header for integrity verification. The payload immediately follows the header and carries the encapsulated data from upper-layer protocols, such as a TCP segment or a UDP datagram, which in turn encapsulate application-layer content such as HTTP messages or DNS queries. IPv6, defined in RFC 8200 from the IETF, extends the address space to 128 bits, simplifies the base header to 40 bytes, and moves optional fields into extension headers that appear between the base header and the payload.

Packet Switching

In a packet-switched network, each packet is routed independently based on its destination address and the current state of the network's routing tables. Routers examine the destination IP address in the header, consult their forwarding tables, and forward the packet toward the next hop on the path to the destination. This store-and-forward model means each router receives an entire packet, checks its integrity, and then transmits it onward, introducing a small delay at each hop. Packet switching delivers high network utilization because capacity is used only when packets are actually in transit, and it provides resilience because packets can be routed around failed links. The ScienceDirect overview of packet structure describes how intermediate devices process packets independently, enabling the statistical multiplexing that makes shared network infrastructure efficient.

Packet Fragmentation and Reassembly

When a packet must traverse a network segment with a maximum transmission unit (MTU) smaller than the packet's size, the packet must be fragmented. IPv4 routers may split an oversized packet into smaller fragments, each carrying a fragment offset field in the header so the destination host can reassemble them in order. The identification field in each fragment matches the original packet, and the "more fragments" flag indicates whether additional pieces follow. In IPv6, fragmentation is handled only by the sending host rather than intermediate routers; the sender performs path MTU discovery using ICMPv6 to learn the minimum MTU along a path before transmitting, avoiding the overhead of mid-path fragmentation.

Applications

Packets are the basis for communication in a wide range of systems, including:

• Internet routing and data delivery across local area and wide area networks
• Voice over IP (VoIP) and video conferencing, where real-time packet delivery governs quality of service
• Content delivery networks, where packets from distributed servers reach users with low latency
• Industrial control systems, where deterministic packet delivery over Ethernet-based protocols supports automation
• Network security analysis, where packet capture and inspection underpin intrusion detection and traffic monitoring