CAN201 W4

This’s the note of introduction to networking Week4.
This week we will talk about the transport layer in networking system. 传输层

• Roadmap
  1. Transport-layer services
  2. Multiplexing and demultiplexing
  3. Connectionless transport: UDP
  4. Principles of reliable data transfer

Lecture

Transport-layer services

Transport services and protocols

• Provide logical communication between app processes running on different hosts 端与端之间
• Transport protocols run in end systems 在这里我们可以发现，之前讲的五层模型只出现于端处，在路由上只有三层，没有五层。
  • Send side: breaks app msg into segments, passes to network layer 发送端:将应用程序消息分成段，传递到网络层
  • Rcv side: reassembles segments into messages Rcv端:将段重组为消息
• Transport-layer protocols for Internet: TCP and UDP

---

• Network layer: logical communication between hosts 网络层:主机之间的逻辑通信。
• Transport layer: logical communication between processes: Relies on, enhances, network layer services. 传输层:进程之间的逻辑通信:依赖、增强网络层服务。

假如把信息的传递表示称Ann家12个孩子和Bill家12个孩子的通信，那么Transport protocol就是Ann和Bill家负责收集和分发12封信的服务，而Network-layer protocol则是家庭与家庭之间的邮政服务。

---

Internet transport-layer protocols

• Reliable, in-order delivery (TCP) 可靠，有序交付
  • Congestion control 拥塞控制
  • Flow control 流量控制
  • Connection setup 建立连接
• Unreliable, unordered delivery: UDP 不可靠，无序交付
  • 多路复用解复用
  • No-frills extension of “best-effort” IP “尽最大努力”IP的无虚饰扩展
• Services not available: 服务不可用 TCP和UDP都不可以
  • Delay guarantees 延迟保证
  • Bandwidth guarantees 带宽保证

---

Multiplexing and demultiplexing 多路复用和多路分用

发送端：Multiplexing从多个socket接收消息，然后打包成一个segment发送出去。
接收端：通过发送端封装上的头部信息把消息分发到每一个socket上。
How demultiplexing works

• Host receives IP datagrams 主机接收IP数据报
  • Each datagram has source IP address, destination IP address 每个数据报都有源IP地址、目的IP地址
  • Each datagram carries one transport-layer segment 每个数据报携带一个传输层段
  • Each segment has source, destination port number 每个段都有源端口号、目的端口号
• Host uses IP addresses & port numbers to direct segment to suitable socket 主机使用IP地址和端口号直接段到合适的套接字

---

Connectionless demultiplexing (UDP)

因为UDP是没有连接的，所以只要IP datagram里的目标IP地址和端口号是一样的，都会送到相同的socket中，不管这个datagram是否来自于同一个进程。

---

Connection-oriented demux

socket被分开了，UDP里只要接受的IP和port不变，就用一个socket接受。
但是TCP只要发送的IP和port不一样，即使接受的IP和port一样，也会重新开port。
threaded server: 线程多进程

---

Connectionless transport: UDP

UDP: User Datagram Protocol

Feature:

• Simple and straightforward
• Best effort
• Lost
• Connectionless
  • No handshaking
  • Each UDP segment handled independently of others: Out-of-order to APP
• UDP use:
  • Streaming multimedia apps
  • DNS
• Reliable transfer over UDP:
  • Add reliability at application layer 应用层解决
  • Application-specific error recovery 错误恢复手段

---

UDP: segment header

---

UDP checksum
Goal: detect “errors” in transmitted segment 检测传输段中的“错误”
Sender:

• Treat segment contents, including header fields, as sequence of 16-bit integers 将段内容(包括报头字段)作为16位整数序列处理
• Checksum: addition (one’s complement sum) of segment contents 校验和:段内容的添加(一个的补和) 负责检查传输有没有出错
• Sender puts checksum value into UDP checksum field 发送端将校验和值放入UDP校验和字段

Receiver:

• Compute checksum of received segment 计算接收报文段的校验和
• Check if computed checksum equals checksum field value: 检查计算的checksum是否等于checksum字段值:
  • NO - error detected
  • YES - no error detected. But maybe errors nonetheless?

在信息发送过来的时候会携带一个checksum，然后接收到之后拿16bits相加，多出来的第一位加到结尾。得出一个sum，然后在根据sum全部翻转过来，得到checksum。最后对比算出来的checksum和发送过来的checksum，假如一样说明没有问题，假如不一样就说明接受或发送出现了问题。

---

Principles of reliable data transfer

可靠数据传输的原理，参见书和ppt。
rdt1.0, rdt2.0, rdt2.1, rdt3.0

rdt 2.0是面对比特错误是使用的方法。假如收到的消息没有问题就发送ACK，有问题就发送NAK，然后发送者接收到返回的消息再继续进行下一阶段的发送，所以只需要两个状态就可以解决问题。
rdt2.0 has a fatal flaw!
What happens if ACK/NAK corrupted? 如果ACK/NAK损坏会发生什么

• Sender doesn’t know what happened at receiver! 发送方不知道接收方发生了什么
• Can’t just retransmit: possible duplicate 不能只是重传:可能重复

Handling duplicates:

• Sender retransmits current pkt if ACK/NAK corrupted
• Sender adds sequence number to each pkt
• Receiver discards (doesn’t deliver up) duplicate pkt

所以就多加了两个状态变成rdt2.1

在rdt2.2中去掉了NAK，只用ACK，假如checksum出现问题就反复回府之前的申请，直到可以进入下一个状态再继续。

rdt 3.0继续解决了丢包的情况：通过加上time设置。
利用率低，因为是停等协议。

---

Lab

# client
import json
from socket import *
import argparse


def _argparse():
    parser = argparse.ArgumentParser(description="This is description!")
    parser.add_argument('--ip', action='store', required=True, dest='ip', help='IP address')
    parser.add_argument('--input', type=str, required=True, dest='input', help='The path of input file')
    parser.add_argument('--port', type=int, required=True, dest='port', help='The port of server')
    return parser.parse_args()


def main():
    parser = _argparse()
    server_name = parser.ip
    server_port = parser.port
    clientSocket = socket(AF_INET, SOCK_STREAM)
    clientSocket.connect((server_name, server_port))
    clientSocket.send(parser.input.encode())
    feedback_command = clientSocket.recv(20480).decode()
    feedback = json.loads(feedback_command)
    print(feedback)

    clientSocket.close()


if __name__ == '__main__':
    main()


# server
import json
from socket import *
import argparse


def _argparse():
    parser = argparse.ArgumentParser(description="This is description!")
    parser.add_argument('--ip', action='store', required=True, dest='ip', help='IP address')
    parser.add_argument('--input', type=str, required=True, dest='input', help='The path of input file')
    parser.add_argument('--port', type=int, required=True, dest='port', help='The port of server')
    return parser.parse_args()


def main():
    parser = _argparse()
    server_port = parser.port
    server_socket = socket(AF_INET, SOCK_STREAM)
    server_socket.bind(('', server_port))
    server_socket.listen(2)
    print('The server is ready to receive')

    while True:
        connectionSocket, addr = server_socket.accept()
        receive_command = connectionSocket.recv(20480).decode()
        receive_message = json.loads(receive_command)
        print(receive_message)
        print(type(receive_message))
        connectionSocket.send(parser.input.encode())

    connectionSocket.close()


if __name__ == '__main__':
    main()