Packet switching

Packet switching is a switching method in which a message is divided into small units called packets, each carrying a header (source, destination, sequence number), and the packets are sent through the network independently and then reassembled at the destination. Links are shared by many users on demand (statistical multiplexing). The nodes use store-and-forward: a packet is fully received, checked, queued and then forwarded to the next node.

Operation

1. The source splits the message into packets and adds headers.
2. Each packet is given to the first router, which stores it in a buffer and checks it for errors.
3. The router looks at the destination address (or virtual circuit number) and its routing table and forwards the packet on the best link.
4. The process repeats at every router until the packet reaches the destination.
5. The destination reorders packets by sequence number and rebuilds the message.

        +--R1---R2--+
 A -----+           +----- B
        +--R3---R4--+
 Packets P1,P2,P3 can take different routes
 and arrive: P2 P1 P3 -> reordered at B

Two approaches

• Datagram: each packet is routed independently, no setup; packets may arrive out of order (Internet, IP).
• Virtual circuit: a logical path is set up first, and all packets follow it in order (X.25, frame relay, ATM).

Features

• Efficient link use for bursty data; no wasted idle bandwidth.
• Alternative routes if a link fails.
• Delay is variable because of queuing; packets can be lost when buffers overflow.
• Packet header is overhead.

Delay in one hop = transmission time (L/R) + propagation + queuing.