Network protocols TCP and UDP

First of all, I want to emphasize that the TCP/IP protocol is a protocol cluster. It includes many protocols, UDP is just one of them. The reason why it is named TCP/IP protocol is that TCP and IP protocols are two very important protocols, so they are named after them.

The difference between the two protocols is that when using them in practice, you only need to remember that "TCP generally does not cause packet loss when sending data in a normal connection (excluding other factors in the upper and lower layers), while UDP packet loss is very common."

1 The following focuses on the differences between TCP protocol and UDP protocol

• Connection-based vs. connectionless;
• Requirements on system resources (more for TCP, less for UDP);
• The UDP program structure is relatively simple;
• Stream mode vs. datagram mode;
• TCP guarantees data correctness, but UDP may lose packets;
• TCP guarantees data order, UDP does not.

Through the above analysis, some people may think that since TCP is a reliable transmission protocol, it must be better than UDP, but this is not the case.

Here is an example to illustrate this problem:

TCP is used when reliable transmission is necessary at the transport layer. Since it is link-oriented and has mechanisms such as sequence control and retransmission control, it can provide reliable transmission for applications.

On the other hand, UDP is mainly used for communications or broadcast communications that require high-speed transmission and real-time performance. Let's take an example - when making a call through an IP phone, if TCP is used, the data will be resent if it is lost in the transmission diagram, but this will not smoothly transmit the caller's voice, resulting in an inability to communicate normally. With UDP, it will not be resent. Therefore, there will be no problem of significant delay in the arrival of the sound. Even if some data is lost, it will only affect a small part of the call. In addition, UDP is also used instead of TCP in multicast and broadcast channels.

TCP (Transmission Control Protocol) is a connection-oriented protocol, which means that before sending or receiving data, a reliable connection must be established with the other party. A TCP connection must go through three "conversations" to be established. The process is very complicated. I will only briefly describe the simple process of these three conversations:

• Host A sends a connection request packet to host B: "I want to send you data, can I?" This is the first conversation;
• Host B sends a data packet to host A to agree to the connection and request synchronization (synchronization means that one host sends and the other receives, coordinating work): "OK, when will you send?" This is the second conversation;
• Host A sends another data packet to confirm Host B's request for synchronization: "I'll send it now, you continue!" This is the third conversation.

The purpose of the three "conversations" is to synchronize the sending and receiving of data packets. After three "conversations", host A officially sends data to host B, which is also a prerequisite for ensuring reliable transmission.

6 UDP (User Data Protocol)

1. UDP is a connectionless protocol. The source and the terminal do not establish a connection before transmitting data. When it wants to transmit, it simply grabs the data from the application and throws it on the network as quickly as possible. At the sending end, the speed at which UDP transmits data is only limited by the speed at which the application generates data, the computer's capabilities, and the transmission bandwidth; at the receiving end, UDP puts each message segment in a queue, and the application reads one message segment from the queue each time.
2. Since no connection is established when transmitting data, there is no need to maintain the connection status, including the sending and receiving status, so a server can transmit the same message to multiple clients at the same time.
3. The header of a UDP packet is very short, only 8 bytes, which has very little additional overhead compared to the 20-byte packet of TCP.
4. Throughput is not regulated by the congestion control algorithm, but is only limited by the rate at which the application software generates data, the transmission bandwidth, and the performance of the source and terminal hosts.
5. UDP uses best-effort delivery, which means reliable delivery is not guaranteed, so the host does not need to maintain a complex link state table (which contains many parameters).
6. UDP is message-oriented. The sender's UDP adds a header to the message handed over by the application and then delivers it to the IP layer. It neither splits nor merges the messages, but preserves the boundaries of these messages. Therefore, the application needs to choose an appropriate message size.

We often use the "ping" command to test whether the TCP/IP communication between two hosts is normal. In fact, the principle of the "ping" command is to send a UDP data packet to the other host, and then the other host confirms receipt of the data packet. If the message whether the data packet has arrived is fed back in time, then the network is connected.

The ping command is used to detect whether hosts can communicate with each other. If a host cannot be pinged, it means that a connection cannot be established with the host. The ping command uses IP and Internet Control Message Protocol (ICMP), so it does not involve any transport protocol (UDP/TCP) and application. It sends an ICMP echo request message to the destination host.

ICMP protocol stipulates that the destination host must return an ICMP echo reply message to the source host. If the source host receives the reply within a certain period of time, it is considered reachable.