Chapter 1: Introduction

🔹 1. What is the Internet?

• “Network of networks”: Interconnected ISPs (Internet Service Providers) and networks that enable global communication.

• Two perspectives:
• Nuts and Bolts View: Physical components (hosts, routers, links, protocols).
• Service View: Provides infrastructure for applications (web, email, VoIP, IoT, streaming, etc.).

1. 什么是互联网？

• “网络的网络”：互相连接的互联网服务提供商（ISPs）和其他网络，使得全球通信成为可能。

• 两种视角：
• 设备视角：物理组件（主机、路由器、链路、协议）。
• 服务视角：为各种应用（如网页、电子邮件、VoIP、物联网、流媒体等）提供基础设施。

📌 Key Components:

💡 Fun Fact: From web-enabled toasters to Fitbits — the Internet connects everything (IoT).

📌 关键要素：

💡 有趣知识点：从联网烤面包机到 Fitbit —— 互联网连接的是 万物（IoT）

🔹 2. What is a Protocol?

• Definition: A set of rules that govern communication between devices.

• Human analogy: “Hello”, “What time is it?”, handshakes.

• Network protocol example:

• Key elements:
• Message format
• Message order
• Actions taken on send/receive

🔹 2. 什么是协议？

• 定义：一组规定设备之间通信的规则。

• 人类类比：”你好“、“几点了？“、握手。

• 网络协议示例：客户端: "GET /index.html" → 服务器: "200 OK" + HTML

• 关键要素：
• 消息的格式
• 消息的顺序
• 发送/接收时所采取的操作

📌 Examples:

✅ RFCs (Request for Comments) = Official Internet standards documents.

✅ IETF = Organization that develops and promotes these standards.

📌 示例：

✅ RFC（请求评论） = 官方互联网标准文件

✅ IETF = 研发并推广这些标准的组织

🔹 3. Network Edge

➤ Hosts

• Clients → Request services (e.g., your laptop browsing a website).

• Servers → Provide services (e.g., Google’s server hosting Gmail). → Often located in data centers.

🔹 3. 网络边缘

➤ 主机

• 客户端 → 请求服务（例如，你的笔记本浏览网站）。

• 服务器 → 提供服务（例如，Google 的服务器托管 Gmail） → 常常位于 数据中心 中。

➤ Access Networks

🔧 Home Network: Cable/DSL modem → Router (with NAT, firewall) → WiFi/Ethernet → devices.

➤ 接入网络

🔹 4. Network Core

➤ Key Functions:

🔹 4. 网络核心

➤ 关键功能：

➤ Two Switching Techniques:

➤ 两种交换技术：

💡 Circuit Switching Techniques:

• FDM (Frequency Division Multiplexing): Each user gets a unique frequency band.

• TDM (Time Division Multiplexing): Each user gets time slots in sequence.

➤ Packet Switching: Store-and-Forward

• Transmission delay: d_trans = L / R → L = packet size (bits), R = link speed (bps)

• Queueing delay: Time waiting in router buffer.

• Entire packet must arrive before forwarding → causes delay but enables sharing.

➤ Example: Circuit vs Packet Switching

1 Gbps link, each user active 10% of the time at 100 Mbps

→ Circuit: Max 10 users (1 Gbps / 100 Mbps)

→ Packet: Can support 35+ users with negligible probability (>10 active at once) ≈ 0.0004

✅ Packet Switching Advantages:

• Better for bursty traffic

• No call setup

• More efficient resource use

❌ Packet Switching Drawbacks:

• Variable delay → bad for real-time apps

• Packet loss → requires congestion control and retransmission

🔄 Q: How to make packet switching “circuit-like”?

→ Use QoS (Quality of Service), traffic shaping, prioritization (covered later).

💡 电路交换技术：

• FDM（频分复用）：每用户分配一个独一无二的频率带宽。

• TDM（时分复用）：按照顺序为每用户分配时间片。

➤ 分组交换：存储-转发

• 传输延迟：d_trans = L / R → L = 分组大小（位），R = 链路速率（bps）

• 排队延迟：在路由器缓存中等待的时间。

• 整个分组必须到达后才能转发 → 会带来延迟，但允许共享。

➤ 例子：电路交换 vs 分组交换

1 Gbps 链路，每个用户 10% 时间的 100 Mbps 活动

→ 电路交换：最多支持 10 用户 (1 Gbps / 100 Mbps)

→ 分组交换：可支持 35+ 用户，同时有超过 10 个用户活动的概率极低 ≈ 0.0004

✅ 分组交换的优势：

• 更适合突发流量

• 没有呼叫建立的过程

• 更高效地利用资源

❌ 分组交换的劣势：

• 延迟可变 → 不适合实时应用

• 分组丢失 → 需要拥塞控制和重传

🔄 问题：如何让分组交换具有“电路交换”特征？

→ 使用 QoS（服务质量）、流量整形、优先级标记（后面会讲到）

🔹 5. Performance Metrics

➤ Four Sources of Delay

🔹 5. 性能指标

➤ 四种延迟来源

✅ Caravan Analogy (L=10 bits, R=10 bit/sec, d=100 km, s=100 km/hr):

• Time to transmit caravan = 120 sec

• Propagation time = 1 hr

• Total = 1 hr 52 min

✅ 车队类比（L = 10 位，R = 10 位/秒，d = 100 公里，s = 100 公里/小时）：

• 车队传输时间 = 120 秒

• 传播时间 = 1 小时

• 总计 = 1 小时 52 分

➤ Traffic Intensity: La / R

• L = packet size (bits)

• a = average packet arrival rate (packets/sec)

• R = link rate (bps)

➤ 流量强度：La/R

• L = 分组大小（位）

• a = 平均分组到达率（分组/秒）

• R = 链路速率（bps）

➤ Throughput

• Definition: Rate bits are delivered from sender to receiver.

• Bottleneck Link: The link with the lowest capacity on the path determines end-to-end throughput.
• e.g., If server link = 100 Mbps, client link = 10 Mbps → Throughput = 10 Mbps

• Shared bottleneck: If N users share link of rate R → per-user throughput ≈ min(Rc, Rs, R/N)

➤ Packet Loss

• Occurs when buffer (queue memory) fills up.

• Packets are dropped → may be retransmitted (e.g., TCP) or discarded (e.g., UDP).

• Detected via traceroute ( = timeout/loss).

➤ Traceroute

• Sends 3 packets with TTL=1,2,3,… → each router returns ICMP error

• Measures delay to each hop

• Reveals network path and latency spikes

➤ 吞吐量

• 定义：从发送端到接收端，数据以多少位每秒传输。

• 瓶颈链路：路径上传输的所有链路中，容量最低的链路决定端到端的吞吐量。
• 例如，如果服务器链路为 100 Mbps，客户端链路为 10 Mbps → 吞吐量 = 10 Mbps

• 共享瓶颈： 如果 N 个用户共享速率为 R 的链路 → 每个用户的吞吐量 ≈ min(Rc, Rs, R/N)

➤ 丢包

• 当缓存（队列内存）满了时发生丢包。

• 分组被丢弃 → 可能重新传输（如 TCP）或直接丢弃（如 UDP）。

• 通过 traceroute（ = 超时/丢包）进行检测。

➤ traceroute

• 发送 3 个包，TTL = 1,2,3,… → 每个路由器返回 ICMP 错误。

• 测量到每个跳转点的延迟。

• 揭示网络路径和延迟峰值。

👉 Example:

1 cs-gw (1ms)

2 border1… (1ms)

3 ...

8 62.40… (104ms) ← trans-oceanic jump!

🔹 6. Network Security (Critical!)

Original Internet Vision: “Mutually trusting users on a transparent network” → ❌ No security by design

🔥 Common Attacks:

✅ Defense Mechanisms:

⚠️ Security is needed at every layer: Application, Transport, Network, Link, Physical.

⚠️ 安全需要在每一层实现：应用层、传输层、网络层、链路层、物理层

🔹 7. Protocol Layers & Encapsulation (CORE EXAM TOPIC!)## 协议分层与封装

➤ Internet Protocol Stack (5 layers) ➤ 互联网协议堆栈（5 层）

✅ No Presentation/Session layers (unlike OSI).

→ Those services implemented in application layer if needed.

✅ 注意：互联网堆栈中没有表示层与会话层（不同于 OSI）

→ 如果有需要，这些服务会在应用层实现。

➤ Encapsulation (Take this seriously!)

• Each layer adds its own header (and sometimes trailer) to data from layer above.

• Analogous to Matryoshka dolls (Russian nesting dolls).

➤ 封装（请重视！）

• 每一层都在数据中添加自己的头部（有时包括尾部）。

• 类似于俄式套娃（嵌套娃娃）。

✅ Encapsulation at Each Node:

• Sender: Adds headers → downward through layers

• Router: Reads IP header → forwards → strips link header → adds new link header → sends

• Receiver: Removes headers bottom-up → delivers M to app

✅ 每个节点的封装过程：

• 发送端：添加头部 → 向下通过各层

• 路由器：读取 IP 头部 → 转发 → 剥离链路头部 → 添加新的链路头部 → 发送

• 接收端：自下而上剥离头部 → 将消息（M）递交给应用层

💡 Key Exam Question:

“What parts of the original message arrive at the destination?”

→ The entire payload M — headers are stripped away!

💡 关键考试问题：

➤ OSI Model (For Awareness)

❗ 重要：互联网堆栈 不使用 表示层与会话层 —— 合并到应用层中。

❗ Important: Internet stack does not use Presentation/Session — roll into Application layer.

🔹 8. Internet History (Timeline Summary)

💡 Key Takeaway: Internet evolved from research project → global utility → critical infrastructure

✅ Final Summary Checklist (Exam Must-Knows)

🧠 Exam Tips

• Draw the encapsulation stack with headers/footers — you’ll lose marks if you skip this!

• Traceroute output analysis: What do mean? Why does delay jump at hop 8?

• Compare circuit vs packet using the 100 Mbps user example.

• Remember: “The Internet didn’t plan for security — we’re fixing it now.”

• Use keywords like: bottleneck, store-and-forward, traffic intensity, encapsulation, tier-1 ISP, IXP, DoS, QoS

📘 Recommended Practice

• Try Wireshark capture (labs): Identify Ethernet, IP, TCP headers.

• Use traceroute from your computer.

• Practice the caravan analogy with different numbers.

• Watch Kurose & Ross’s interactive applets: http://gaia.cs.umass.edu/kurose_ross/interactive