Differences and Similarities between TCP and UDP

The transport layer of the TCP/IP model has two different protocols: the UDP User Datagram Protocol and the TCP Transmission Control Protocol.

Similarities

  • Both TCP and UDP are protocols that operate at the transport layer.
  • Communication using both TCP and UDP requires open ports.

Differences

TCP

  • TCP is a connection-oriented protocol that provides full-duplex communication. It requires the establishment of a connection before data transmission, and the data payload is relatively larger.
  • TCP provides reliable delivery services, ensuring reliable communication using services like flow control and congestion control.
  • The minimum TCP header size is 20 bytes, and the maximum header size is 60 bytes, including source port, destination port, sequence number, acknowledgment number, data offset, control flags, window, checksum, urgent pointer, options, and other information.
  • TCP only supports one-to-one communication.
  • Communication in TCP is byte-oriented.
  • TCP ensures the order of data transmission by assigning a sequence number to each byte of the data stream within the TCP connection.
  • TCP provides features such as checksum, acknowledgment response, sequence number, timeout retransmission, connection management, flow control, and congestion control.
  • TCP is suitable for applications that require reliable transmission, such as file transfers.

UDP

  • UDP is connectionless, meaning that a connection does not need to be established before sending data, and the data payload is relatively smaller.
  • UDP provides best-effort delivery, meaning that it does not guarantee reliable delivery and does not use flow control and congestion control.
  • The UDP header is 8 bytes long, including source port, destination port, length, and checksum information.
  • UDP supports unicast, multicast, and broadcast functionalities, enabling one-to-one, one-to-many, many-to-many, and many-to-one data transmission.
  • UDP is message-oriented, and it does not merge or split the data segments received from the application layer. Instead, it preserves the boundaries of these segments and passes them down to the IP layer after adding the header.
  • UDP does not ensure the order of data transmission, and the application layer program needs to control the order by adding sequence numbers to the data segments.
  • UDP adds minimal functionality on top of the IP datagram service, mainly adding port functionality and error detection.
  • UDP is suitable for real-time applications such as network telephony, video conferencing, and live streaming.

Related Issues

Achieving Reliable Transmission with UDP

The transport layer cannot ensure the reliable transmission of data, and it can only be implemented at the application layer. This can be done by emulating the reliability transmission features of TCP, such as implementing acknowledgment mechanisms, retransmission mechanisms, and window confirmation, but at the application layer. Open-source programs like RUDP, RTP, and UDT use UDP to achieve reliable data transmission. Additionally, there are protocols like KCP that balance the reliability of TCP with the speed of UDP.

Differences between Short and Long TCP Connections

Short connections: When a Client sends a message to a Server, the Server responds to the Client, and then the process is completed with a single read and write. At this point, either party can initiate a close operation, but it is usually the Client that initiates the close operation. Short connections usually involve only one read and write operation between the Client and the Server. Managing short connections is relatively simple, as any existing connection is useful and does not require additional control measures. However, frequent client requests can result in wasted time and bandwidth on TCP connection establishment and closure operations.

Long connections: After a Client and Server complete a read and write operation, the connection between them does not close actively. Subsequent read and write operations continue to use this connection. Long connections can eliminate many TCP setup and closure operations, reduce waste, and save time. They are more suitable for clients that frequently request resources. However, as the number of client connections increases, the server load will also increase. In such cases, strategies such as closing long-unused connections using the LRU algorithm, and limiting the number of client connections may be necessary to reduce the load.

Solutions for TCP Packet Sticking and Unpacking

Since TCP is byte-oriented and cannot understand the business data of the higher layer, it is unable to guarantee that data packets are not split or recombined at the lower level. This issue can only be resolved through the design of the upper-layer application protocol stack.

  • Fixed-length message: The sending end encapsulates each data packet with a fixed length (padding with 0 if necessary). This way, the receiving end naturally splits each data packet by reading a fixed length of data from the receive buffer.
  • Setting message boundaries: The server separates message content from the network stream based on message boundaries and uses a line feed character at the end of the packet for separation, as is done in the FTP protocol.
  • Splitting messages into message headers and bodies: The message header includes a field that indicates the total length of the message (or the length of the message body).

Daily Question

https://github.com/WindrunnerMax/EveryDay

References

https://zhuanlan.zhihu.com/p/108822858 https://www.cnblogs.com/liangyc/p/11628148.html https://blog.csdn.net/m_xiaoer/article/details/72885418 https://blog.csdn.net/pangyemeng/article/details/50387078 https://blog.csdn.net/quiet_girl/article/details/50599777 https://blog.csdn.net/liuchenxia8/article/details/80428157 https://blog.csdn.net/qq_40732350/article/details/90902396 https://www.cnblogs.com/fundebug/p/differences-of-tcp-and-udp.html