Computer network interview knowledge points

1. Computer network architecture model

• The physical layer ： establish 、 maintain 、 Disconnect the physical connection .
• Data link layer ： Establish logical connections 、 Address the hardware 、 Error checking and other functions .
• The network layer ： Logical address addressing , Realize path selection between different networks .
• Transport layer ： Define the protocol port number of data transmission , And flow control and error checking .
• The session layer ： establish 、 management 、 Terminate the conversation ,（ The application layer has been incorporated into the five layer model ）
  At one end of the session is the local host , The other end is the remote host .
• The presentation layer ： Presentation of data 、 Security 、 Compress .（ The application layer has been incorporated into the five layer model ）
  The format is ,JPEG、ASCll、EBCDIC、 Encryption format, etc .
• application layer ： An interface between network services and end users .

The network protocol is actually an end-to-end Communication rules , With these rules , The communication between both sides is meaningful .

2. TCP/IP Header format for

among , There are several important points ：

• Serial number (seq)： The serial number of the packet , Confirm the continuity of the package through the sequence number , Solve the disorder of packages .
• Response number (ack)： Confirmation sequence number for the above field , For example, the server receives a request from the client , It contains seq Serial number , When the server responds back , Will be carried out in ack = seq + 1 Field settings for , Indicates the cumulative data received .
• Window： Slide the window to use , Used to feed back the packet size that the receiver can handle next , Prevent unequal processing capacity of data packets between both parties , It mainly solves the problem of flow control .
• TCP Flag：TCP Packet type , To assist in TCP Stage processing , such as SYN To establish a connection ,FIN Indicates that the connection is closed .

3. TCP The state of circulation

4. TCP Three handshakes of 、 How about four waves ？ Why is it designed this way? ？

If we only do 2 A handshake establishes a connection , So for Server It's too easy to establish a connection , Basically, a client came , that Server The connection is about to be established . This situation will lead to too low connection cost ,Server The end is very tolerant of overload .

Four waves because TCP yes full duplex Of , There are two behaviors of data sending and receiving , Data inflow and outflow in both directions need to be closed .

5. What is semi connection ？ Related to it SYN What is the attack like ？

When the service receives a client request to connect SYN, Then respond to the client ACK and SYN after , The connection will be maintained to the semi connection queue . When the client replies again ACk after , The current connection will be maintained in the full connection queue .

SYN The attack was DOS A type of attack , Establish a connection by forging a large number of requests , Make the semi connected queue exceed the maximum capacity , Other normal requests cannot be processed .

6. Why does the initiator in the four waves need to wait TIME_WAIT Time to close the connection ？

TIME_WAIT Is a timing setting , stay 2*MSL（MSL Represents the lifetime of a packet in a network environment , It's usually 2 minute , Linux Li Wei 30s） After time, it will really CLOSED.

Why not close it immediately , It is mainly to allow the passive shutdown party to have enough time to receive the final Ack package , If not received , The passive party will resend Fin package , Re trigger the active party to send the last Ack package . In this case , We can try our best to ensure that the passive shutdown party closes the connection as soon as possible , After all, the active shutdown party needs to bear the main responsibility , So there will be TIME_WAIT I'm waiting for you .

Another reason is the fear that the current connection will be released immediately , There is a certain probability that the current connection ID will be reused ( Quintuples ), The old network packet was received at this time due to delay , There may be package confusion .

7. TCP What is the retransmission mechanism of ？

TCP All packets sent need to be sent by the receiver Ack Package response , If there is no response within a certain time , Then the sender will think that the packet did not arrive correctly , Retransmission action is required . This is it. TCP The retransmission mechanism .

TCP The retransmission mechanism in will have a timeout judgment , This timeout is not very accurate , Or it's not very standard , After all, different network environments , The arrival of the package will be different .

therefore TCP A sampling time is used , First, the time of a packet from sending to response confirmation under normal conditions is recorded , That is to say RTT（Round Trip Time） Time , Make some formula calculations according to this time , Get the value of timeout ：RTO（Retransmission TimeOut）

For the retransmission mechanism , There is another trigger mechanism . The above situation belongs to the sender to find out the sending situation , There is another situation that the receiver can detect . For example, the sender sent 1, 2, 3 My bag , But in fact, the receiver only receives 1 and 3, I haven't been able to receive 2 This package , Then the receiver will respond to three consecutive About 2 Of ack package .

When the sender receives such a continuous 3 individual ack After package , You know you need to retransmit 2 了 , There is no need to wait until 2 The timeout of was not acknowledged and triggered , Can be retransmitted in advance 2 This bag. .

8. introduce TCP Flow control of , What about sliding windows ？

TCP use The sliding window The flow is controlled , The so-called sliding window means that the sender and receiver maintain a window respectively , In this window, the corresponding packet will be maintained , To perceive the current data processing .

The window on the receiver side is called Receiving window , It specifically represents the current packet size that can be received , The formula is ： Current maximum acceptable buffer size - Currently received size , When the connection is established, it is generally 65535 byte .

After calculating the acceptable size , The receiver will set this value to TCP In the head Window Field , Then respond back to the sender , The sender will know the current packet size allowed to be sent .

The window on the sender side is called Send window , In normal logic , The sending window maintains the data to be sent , That is, the transmission data calculated according to the size of the reception window just fed back .

But because a packet needs to be sent with ACK Response is the complete process , So for these “ Sent unresponsive ” The data should also be included in the management of the sending window , And only really ACK Respond back , To continue the preparation of the next packet .

It should be noted that , If the sender receives Window The size is 0, It means that the current receiver is no longer capable of processing new packets , At this point, the sender will no longer send data , Until the receiver sends a Window announcement , Before continuing to send data .

But here's a situation to consider , The receiver failed to deliver the window notification to the sender due to network problems , Then the sender will be waiting . So for the sender , It will start Window detection action , The receiver is required to ACK Its current receive window size , If exceeded 3 The first discovery action , Then directly disconnect .

9. TCP How to perceive and control the congestion in the network environment ？

TCP The protocol Abstracts Congestion window （cwnd） The concept of , It will adjust dynamically according to the current degree of network congestion . Due to the consideration of congestion , The sending window we mentioned above cannot only consider the receiving window , Need to carry out min( Congestion window , Receiving window ) Your choice was sent .

10. MTU and MSS What is it? ？

MSS Express Network transmission data The maximum of , If MSS Plus the Baotou size , Indicates the maximum network transmission message ：MTU .

stay Internet In this kind of Internet , commonly MTU Defined as 576 byte , subtract TCP、IP The baotou 40 byte , You can get MSS = 536 Byte value ; And in a LAN like Ethernet , commonly MTU It'll be bigger ：1500 byte ,MSS by 1460 byte .

11. TCP What is the slow start of ？ What are the links ？

When the connection is established , When starting data transfer ,TCP The protocol stipulates that large packets cannot be sent at the beginning , This avoids problems in the network environment , Newly added connections exacerbate congestion . therefore , For new connections , You need to increase the amount of data bit by bit , That's what's called Slow start .

among , The calculation process of congestion window involved in slow start is as follows ：

• When you first set up a connection , Congestion window = 1
• Whenever you receive a ACK Packet time , Congestion window = Congestion window + 1, At this time, it increases linearly .
• Every time I pass a RTT, Congestion window = Congestion window * 2, At this time, it shows an exponential upward trend .

12. What is congestion avoidance ？

From the slow start algorithm , Every time I pass by RTT after , The growth rate of congestion window will become very severe , If there are no restrictions , Then the bandwidth will be full soon . therefore , TCP The protocol uses a threshold called slow start (ssthresh) The variable of （ Usually take 65535 byte ）. When cwnd( Congestion window ) Beyond this limit , Will enter the so-called Congestion avoidance Stage .

In the congestion avoidance phase , The calculation process of congestion window is as follows ：

• Every time I receive one ACK Packet time , Congestion window = Congestion window + 1/ Congestion window
• Every time I pass a RTT, Congestion window = Congestion window + 1

As can be seen from the above algorithm , After entering the congestion avoidance phase , The sending size of packets will increase linearly . In this way , bring TCP The transmission is very fast in the early stage , Then slowly lower , Reach the best value of the network .

13. When congestion happens ,TCP What will happen to ？

When congestion occurs , The calculation of congestion window is in different TCP The version will be different , There are mainly the following 3 Kind of version ：

Tahoe edition

Tahoe The version is TCP The earliest version of , When it finds that it needs to retransmit , That triggers RTO Timeout or sender receives three duplicates ACK Packet time , At this time, the action will be ：

• sshthresh( Slow start threshold ) = cwnd( Congestion window ) / 2
• cwnd Reset to 1
• Go back to the slow start phase

Reno edition

Reno The action of version is ：

• sshthresh( Slow start threshold ) = cwnd( Congestion window ) / 2
• cwnd( Congestion window ) = sshthresh( Slow start threshold ) + 3 * MSS （ take 3 A repetition ACK Take into account ）
• Enter the rapid recovery phase

among , The calculation of rapid recovery phase is as follows ：

• When the retransmitted packet is sent out , Received a retransmission packet ACK after ,cwnd( Congestion window ) = cwnd( Congestion window ) + 1
• When a new packet is received ACK after , The quick recovery process is now complete , be cwnd = sshthresh, Then return to the congestion avoidance phase

NewReno edition

NewReno It's right Reno Improvement , It mainly optimizes the rapid recovery phase , stay Reno In the version , What is considered is the loss of a packet . However , In practice , Send a data window , There may be a lot of packet loss .

In this case , Will trigger multiple times cwnd and ssthresh Halve , once cwnd Reduce to less than 3 when , That is, the sending window will appear less than 3 The circumstances of , This will no longer trigger 3 A fast retransmission , Can only rely on RTO Overtime , In general RTO The value of is relatively large （ Too small will often trigger retransmission ） Of , At this time, the whole transmission speed will be greatly reduced .

therefore NewReno The fast recovery phase will not end until all packets are confirmed , such cwnd and sshthresh It won't be easily reduced .

NewReno It mainly uses a recover Variable , As in the current data window , Maximum sequence number of possible packet loss . That is, if packet loss occurs , And greater than the current recover value , The value is updated .

When receiving the receiver's ack after , Will be carried out in ack_seq The judgment of the , If ack_seq > recover, At this point, you can end the rapid recovery phase ; If ack_seq < recover, It means that multiple packets are lost , Can't end the fast recovery phase yet . Through such control , To improve the throughput of the whole .

14. TCP What algorithms or mechanisms are there to improve efficiency ？

Nagle Algorithm

Nagle The algorithm combines multiple small packets into one segment , And wait until certain conditions are met , Send it together . The specific trigger conditions are as follows ：

• If the packet length reaches MSS, Is allowed to send
• If you include FIN, Is allowed to send
• If set TCP_NODELAY, Is allowed to send
• Not set TCP_CORK Option , If all the small packets sent （ Packet length less than MSS） All is confirmed , Is allowed to send

When none of the above conditions are met , But a timeout occurred （ It's usually 200ms）, Send immediately .

about TCP In terms of agreement , By default Nagle Algorithm , Reduce network load , Reduce network congestion , Improve network throughput .

Delay Ack( Delay confirmation )

stay TCP In the confirmation mechanism , It is possible not to each in the communication process TCP Data packets are processed separately ACK Package response , But when transmitting data , By the way, the ACK The message is sent with the packet , This can improve the utilization of network traffic .

If in a certain period of time （ commonly 40 ms） No packets to send , At this time, it will be carried out separately ACK Package response . This process is also called Delay Ack.

15. TCP How about sticking and unpacking in ？

TCP Is byte stream oriented transport , It will load and send some data according to the packet processing capacity of the receiver and the current network congestion , Plus there are Nagle This algorithm for integrating small packets exists . So for the receiver , The received data may be glued together , It may also be split , The so-called sticking and unpacking .

For sticking and unpacking , Common solutions are ：

• Add the length of the current packet to be transmitted at the header of the packet , Let the receiver cut according to the length .
• Encapsulate packets into fixed lengths , Not enough to make up 0, Let the receiver parse by a fixed length .
• Artificially adding boundaries to packets , For example, add special characters at the end of the packet , When parsing to special characters , The receiver thinks it has read a complete and meaningful data segment .

16. TCP Why is it reliable ？

TCP The connection and disconnection of both parties are carried out through mutual communication （ Three handshakes 、 Four waves ）. In the process of data transmission, response confirmation will be carried out 、 Over time retransmission 、 flow control 、 Congestion control 、 Congestion avoidance and other means to ensure the accuracy of transmission .

17. TCP and UDP The difference between

TCP It is a reliable connection for byte stream , and UDP It is data message oriented connection , Reliable connection is not guaranteed , But the transmission is relatively fast .TCP Often used in mail 、 File transfer, which requires high accuracy , and UDP It is often used in live video transmission .

18. TCP and UDP Common application layer protocols

• TCP application layer ：HTTP( Hypertext transfer protocol )、FTP( File transfer protocol )、SMTP( Mail transfer )
• UDP application layer ：DNS、TFTP（ Simple file transfer protocol ）、NTP（ Network time protocol ）

19. HTTP Why is it stateless ？

HTTP Is based on TCP Short connection to protocol , On request - Respond to communicate . Each request is independent , It has nothing to do with last time . Even though TCP It's stateful , But its state is for transmission , For example, message serial number 、 Send auxiliary information such as window size , These and HTTP Your request doesn't matter .

Http Although there are keep-alive Field control , But that is to improve transmission efficiency , Let the life cycle of this request connection be as long as possible , Not to establish connections frequently - Destroy connection . in addition ,cookie Conversation is just HTTP A supplement to , It is allowed to close or forge , It is not the communication dependence of the Protocol .

20. Input from the browser url To the whole process of displaying web pages

First , According to the domain name DNS Parsing , To get the IP Address . Get IP After the address, it will shake hands with the server three times , establish TCP Connect . Then we will follow HTTP Request for agreement - Response to transmit web content . Last ,TCP End the connection with four waves .

21. HTTP 1.0/1.1/2.0 as well as https The difference between

HTTP 1.0

Each request - The response will be established once TCP Connect , The server will be disconnected after processing TCP Connect . I added Connection: keep-alive To delay TCP Connection time , Try to make the request - The response uses the same connection

HTTP 1.1

• Default TCP Persistent connection , No declaration required Connection Field , But the same domain name is maintained TCP There won't be too many persistent connections , commonly 6 individual ;
• Introduce pipeline mechanism , The client can use the same TCP The connection sends multiple requests at the same time , The server responds in the order of the received requests ; But this is a pseudo pipeline pattern , Because after sending multiple at the same time , The client still needs to block and wait for all responses from the server before continuing with subsequent requests .
• Added PUT、DELETE、OPTIONS Etc

HTTP 2.0

• Use binary format instead of text format
• Real pipeline mode , Non blocking waiting for all responses , Equivalent to full duplex
• The message header will be compressed and then sent

HTTPS

HTTPS It's solved HTTP Secure transmission of , stay HTTP Under this layer of agreement SSL layer . That is, end-to-end encryption / Authentication , To ensure that the data will not be stolen or tampered .

HTTPS The process is as follows ：

22. ping Principle ？

ping Adopted ICMP agreement ,ICMP The agreement is used in IP host 、 Routing control messages between routers . Control message means that the network is not accessible 、 Whether the host can reach 、 Whether the route is available and so on . Although these control messages do not transmit user data , But it plays an important role in the transmission of user data .

23. What is? DOS attack ？

DOS： Denial of service , Its purpose is to make the computer or network unable to provide normal services . The most common DoS Attacks include computer network bandwidth attacks and connectivity attacks , image SYN Flood attack is also a kind of , It USES TCP The protocol sends a large number of semi connection requests , Consume server's CPU And memory resources .

---

Interested friends can search the official account 「 Read new technology 」, Pay attention to more push articles .
If you can , Just like it by the way 、 Leave a message 、 Under the share , Thank you for your support ！
Read new technology , Read more new knowledge .