Week 1: Network Edge, Network Core

인터넷 개요

What's the Internet?

Nuts and bolts View

Internet: network of networks

• host = end system = end node = termial = station
• protocols control sending, receiving of messages
• Internet standards: RFC, IETF

Service View

• infrastructure that provides services to applicaations
• provide programming interface to application

What's a protocol?

Protocols define format, order of messages sent and received among network entities, and actions taken on message transmission, receipt.

---

Network Edge (LAN)

Network Edge

hosts: clients and servers

access networks, physical media - wired, wireless communication links

 

Network Core

interconnected routers, network of networks

End System

end system = hosts: clients and servers

client host와 달리 server host는 core network의 라우터에 직접 연결되어 있다? -> (X)

How to connect end systems to edge router

• residential access net(home network), institutional access net(Ethernet), mobile access net..

Access Network

Access network: 인터넷 호스트(client, server)를 core network의 edge router로 연결해주는 네트워크

 

Digital Subscriber Line (DSL)

• data transmitted at different frequencies over dedicated line to central office
• use existing telephone line to central office DSLAM
• 구리 전화선을 이용하여 인터넷 연결 제공
• PC 통신 -> ISDN --HFC--> ADSL -> VDSL -> FTTH(fiber to the home)

Cable network

• data transmitted at different frequencies over shared cable distribution network
• HFC: hybrid fiber coax
• homes share access network to cable headened

Host: send packets of data

• 1 large msg -> multiple small packets (L bits each), pkt by pkt으로 독립적으로 전송
• link transmission rate = link bandwidth = link capacity = transmission rate R (bits / sec) bps

Dtrans (transmission delay, sec) = time needed to transmit L-bit pkt into link = L (bits) / R (bits/sec)

---

Network Core (WAN)

Packet Switching

Packet Switching: hosts break application-layer(L5) messages into packets.

• forward packets from one router to the next, across links on path from source to destination
• each packet transmitted at full link capacity
• statistical TDM - header 필요

Routing

• determines source-destination route taken by packets using routing algorithms (routing table)

 

Forwarding

• move packets from router's input to appropriate router output

 

Store-and-Forward

• L-bit 가 모두 들어올 때까지 기다린 후 처리 -> L-bit가 전송되어야 할 정보가 패킷 맨 앞 헤더에 있기 때문
• entire packet must arrive at router before it can be transmitted on next link
• one-hop trasnmission delay = L / R
• end-to-end delay = # of hops * (L / R) when assuming zero propagation delay

Queuing and Loss

If arrival rate to link exceeds transmission rate of link for a period of time:

• packets will queue at output buffer, wait to be transmitted on link
• packets can be dropped (loss) if memorry (buffer) fills up
• because in packet switching, there is no resource (link bandwidth = R) reservation before data delivery

 

Packet Switching 네트워크를 구성하는 라우터가 Store-and-Forward 방식으로 패킷을 처리하는 동작을 라우터와 L-bit 패킷 입장에서 서술 (패킷은 A 포트에서 들어와서 B 포트로 나간다고 가정)

1. L 개의 비트들이 A 포트의 buffer에 저장된다.
2. 패킷 헤더를 보고 목적지 주소로 보내기 위해 포워딩 테이블을 검색한다.
3. L-bit 패킷이 A에서 B로 스위칭 된다
4. L-bit 패킷이 B의 buffer(queue)에 한동안 저장된다.
5. L 개의 비트들이 signal로 변환 (transmit) 된다.

Packet Switching에서 메시지를 pkt으로 나누어 전송하는 이유

• 중간 라우터들이 동일한 msg에 속하는 패킷들을 동시에 전송하게 되어 결과적으로 end-to-end delay를 줄이기 위해
• 에러가 있는 링크에서 재전송되는 데이터량을 줄여 링크 사용 효율을 증대시키기 위해
• 여러 사용자가 하나의 링크를 공유할 때 한 명의 사용자가 링크를 오래 점유하는 것을 막기 위해

Circuit Switching

Circuit Switching: end-end resources allocated to, reserved for "call" btw source & destination

• before data delivery, there is call set-up delay -> during data delivery, no processing(routing) at intermediate node
• dedicated resources: no sharing, guaranteed performance, circuit segment idle if not used by call
• commonly used in traditional telephone network

FDM versus TDM

FDM : full rate

Synchronous-TDM: not full rate

Packet Switching VS Circuit Switching

 

Packet Switching	Circuit Switching
No Call set-up, simpler	Call set-up
Allocating link use on demand	Resource reservation
Resource sharing	Resource dedicated to a particular user
Packet delay or loss	Guaranteed service
Queueing delay	Call set-up delay
More users can be served	# of users at the same time is limited
Good for intermittent, bursty data	Good for apps which require guaranteed bandwidth
Dynamic use of BW	Fixed use of BW

Network Structure

End systems connect to Internet via access ISPs(Internet Service Providers)

• ISP example: T-mobile, SKT, KT, Orange etc.

Access ISPs in turn must be interconnected

Internet exchange point (IXP)

네트워크(ISP)를 연결하는 기술 중 하나로 third-party 회사에서 제공하는 링크를 이용해 다중의 ISP들이 연결되는 방식

• Typically in a stand-alone building with its own switches
• About 300 IXPs in the internet

Peer

네트워크(ISP)를 연결하는 기술 중 하나로 동일 계층의 ISP들끼리 payment 없이 직접 연결되는 방식. 상호 주고받는 트래픽양이 유사할 때 주로 사용

• To reduce cost -> when two ISPs peer, it is typically settlement-free(no cost)

Hierarchy structure

• Access ISPs in any given region connect to regional ISP
• Regional ISP then connects to global ISP
• There may be multiple competing regional ISPs in a region
• There will be a larger regional ISP to which smaller regional ISPs in the region connect -> Multi-tier hierarchy

Content Provider Network (CPN == CDN, e.g. Google, Microsoft)

• Major content providers may run their own private network to interconnect data centers
• to reduce payment to upper-tier ISPs and get greater control of services
• 마지막에 Tier-1 ISP와 연결되어야 한다

Point of Presence (PoP)

• customer ISP가 상위 provider ISP에 연결되는 지점으로 여러 router 들의 집합

Multi-home

• ISP(except for Tier-1 ISP) may choose to connect to two or more provider ISPs
• for reliability of service