💡 왼쪽 원문을 읽으면서 오른쪽에 따라 써보세요. Tab 키로 힌트를 받을 수 있습니다.

원문 렌더가 준비되기 전까지 텍스트 가이드로 표시합니다.

This post is based on the textbook Computer Networking: A Top-Down Approach (6th Edition) by James Kurose and Keith Ross.

1. The Network Core

The network core is a mesh of **packet switches** and **links** that interconnect end systems. There are two fundamental approaches to data transfer:

- **Packet Switching**

- **Circuit Switching**

2. Packet Switching

2.1 Basic Concept

Application-layer messages are divided into smaller chunks of data called **packets** for transmission. Each packet travels through communication links and packet switches (routers, link-layer switches) to reach its destination.

2.2 Store-and-Forward Transmission

Most packet switches use **store-and-forward** transmission.

> The switch must receive the **entire packet** before it can begin transmitting the first bit of the packet on the outbound link.

Source Router Destination

| | |

|==== Packet 1 ====>| |

| |==== Packet 1 ====>|

|==== Packet 2 ====>| |

| |==== Packet 2 ====>|

Transmission Delay Calculation

When the packet size is `L` bits and the transmission rate is `R` bps:

- Time to traverse one link: `L/R` seconds

- Total transmission delay across `N` links: `N * L/R` seconds (due to store-and-forward)

**Example**: Packet size `L = 10,000 bits`, transmission rate `R = 2 Mbps`, 2 links

Total delay = 2 * (10,000 / 2,000,000) = 0.01 seconds = 10ms

2.3 Queuing Delay and Packet Loss

Each packet switch has multiple links connected to it, and maintains an **output buffer** (or **output queue**) for each link.

+----------------------------+

Input --->| Output Queue (Buffer) |----> Output Link

Link | [pkt3][pkt2][pkt1] ---> | (R bps)

+----------------------------+

^ Queuing delay occurs!

Queuing Delay

The time a packet waits in the queue before being transmitted on the output link. It varies depending on network congestion.

Packet Loss

Since buffer space is finite, a packet arriving at a full queue will be **dropped**. This is **packet loss**.

Scenario: Buffer size = 3 packets

Time 1: [pkt1][pkt2][pkt3] -> Buffer full

Time 2: pkt4 arrives -> Buffer overflow -> pkt4 dropped!

2.4 Forwarding Tables and Routing

For a packet to reach its destination, routers must select the appropriate output link.

1. Each end system has an **IP address**

2. The sender includes the **destination IP address** in the packet header

3. The router consults its **forwarding table** to determine the output link

4. Forwarding tables are automatically configured by **routing protocols**

Packet header: Destination IP = 121.7.106.83

Router forwarding table:

+------------------+-------------+

| Address Range | Output Link |

+------------------+-------------+

| 121.7.0.0/16 | Link 2 |

| 200.23.0.0/16 | Link 3 |

| Default | Link 0 |

+------------------+-------------+

-> Forward to Link 2

3. Circuit Switching

3.1 Basic Concept

**Before** communication begins, a **dedicated circuit** is established between the sender and receiver.

The traditional telephone network is a classic example of a circuit-switched network.

Circuit switching process:

1. Connection setup (circuit reservation)

2. Data transfer (using dedicated resources)

3. Connection teardown (circuit release)

Key Characteristics

- Guarantees a **constant transmission rate** for the duration of the connection

- Resources are reserved, so there is **no queuing delay**

- Resources cannot be used by other connections even when idle (possible resource waste)

3.2 FDM (Frequency Division Multiplexing)

The frequency band is divided into multiple **frequency bands**, each assigned to a connection.

Frequency

| +--------+

| | User 4 | Band 4

| +--------+

| | User 3 | Band 3

| +--------+

| | User 2 | Band 2

| +--------+

| | User 1 | Band 1

| +--------+

+--------------------> Time

- Each connection uses only its assigned frequency band

- FM radio stations are a good example of FDM

- Typical bandwidth: 4 kHz (telephone networks)

3.3 TDM (Time Division Multiplexing)

Time is divided into **frames** of fixed length, and each frame is divided into a fixed number of **time slots**, each assigned to a connection.

Frame 1 Frame 2 Frame 3

+--+--+--+--+ +--+--+--+--+ +--+--+--+--+

|S1|S2|S3|S4| |S1|S2|S3|S4| |S1|S2|S3|S4|

+--+--+--+--+ +--+--+--+--+ +--+--+--+--+

---------------------------------------------> Time

S1: User 1's slot, S2: User 2's slot, ...

- Each connection transmits data only during its time slot in each frame

- Transmission rate: If the link rate is `R` and the number of slots is `N`, each connection gets `R/N`

4. Packet Switching vs Circuit Switching Comparison

4.1 Quantitative Comparison Example

Link capacity: `1 Mbps`

| Method | Simultaneous Users |

| ---------------------------------- | ------------------ |

| Circuit switching (100 kbps each) | 10 |

| Packet switching (avg. 10% active) | 35 or more |

Why Packet Switching Can Accommodate More Users

Assume each user generates data at `100 kbps` but is active only 10% of the time.

- **Circuit switching**: Reserves `100 kbps` per user regardless of activity -> max 10 users

- **Packet switching**: The probability that more than 10 of 35 users are simultaneously active is less than 0.0004

Statistical Multiplexing in Packet Switching:

35 users, each active with 10% probability

Expected number of simultaneously active users = 35 * 0.1 = 3.5

-> 1 Mbps link can easily handle this

4.2 Advantages and Disadvantages

| Criterion | Packet Switching | Circuit Switching |

| ------------------- | ------------------------------- | --------------------------------- |

| Resource efficiency | High (statistical multiplexing) | Low (resource reservation) |

| Implementation | Simple | Complex (connection setup needed) |

| Delay guarantee | No guarantee | Constant delay guaranteed |

| Under congestion | Packet loss, increased delay | Connection rejection |

| Bursty traffic | Handles well | Inefficient |

4.3 Conclusion

Today's Internet uses **packet switching**.

- More efficient for bursty traffic

- Simpler and more cost-effective

- However, QoS guarantees for real-time services (voice, video) remain a challenge

5. A Network of Networks

5.1 Evolution of the ISP Hierarchy

End systems connect to the Internet through access ISPs. But how do access ISPs connect to each other?

Structure 1: Connect All Access ISPs to a Single Global ISP

Access ISP --+

Access ISP --+-- Global ISP

Access ISP --+

Unrealistic: One ISP would need to cover the entire world

Structure 2: Multiple Global ISPs Competing

Access ISP -- Global ISP A -- IXP -- Global ISP B -- Access ISP

- **IXP (Internet Exchange Point)**: A point where ISPs exchange traffic with each other

Structure 3: Multi-Tier Hierarchy

Access ISP -- Regional ISP -- Tier-1 ISP -- Regional ISP -- Access ISP

5.2 Modern Internet Structure

+---------------------+

| Tier-1 ISP |

| (AT&T, NTT, etc.) |

+------+------+-------+

| IXP |

+------------+ +------------+

| | | |

+----+----+ +---+---+ +----+----+

| ISP | | ISP | | ISP |

+----+----+ +---+---+ +----+----+

| | |

+----+----+ +---+---+ +----+----+

| ISP | | ISP | | ISP |

+---------+ +-------+ +---------+

Key Components

- **Tier-1 ISP**: Top-level ISPs with worldwide coverage (approximately 12)

- **IXP**: Direct traffic exchange points between ISPs (over 600 worldwide)

- **Peering**: Settlement-free traffic exchange between ISPs at the same tier

- **Content Provider Network**: Companies like Google and Microsoft build their own networks

Google's network strategy:

+--------------------------------+

| Google data centers (global) |

| Connected via private network |

| Direct connections to Tier-1 |

| and IXPs |

| -> Reduces intermediate ISP |

| costs |

| -> Direct control over |

| service quality |

+--------------------------------+

6. Summary

Network Core Key Comparison:

Packet Switching Circuit Switching

---------------- -----------------

Store-and-forward Dedicated circuit setup

Queuing delay occurs No queuing delay

Statistical multiplexing FDM / TDM

Packet loss possible Resource waste possible

Today's Internet Traditional telephone network

7. Review Questions

A transmission method where the packet switch must **receive all bits of a packet** before transmitting the first bit on the outbound link. This adds a transmission delay of `L/R` seconds at each link traversal.

Because the output buffer (queue) of a packet switch has finite space. When the arrival rate exceeds the transmission rate of the output link, packets queue up, and when the queue is full, newly arriving packets are **dropped**.

Thanks to **statistical multiplexing**. Since not all users transmit data simultaneously, packet switching does not reserve resources in advance but uses links only when needed. This allows the same resources to accommodate more users.