A closer look at network structure:

 network edge: applications and hosts  network core: Network Overview ❍ routers ❍ network of networks  access networks, physical media: communication links

Introduction 1-1 Introduction 1-2

The network edge: The network edge:

 client/server model ❍ client host requests, receives service from always-  end systems (hosts): on server ❍ run application programs ❍ e.g. Web browser/server; email client/server ❍ e.g. Web, email ❍ at “edge of network”  why such a popular model?

Introduction 1-3 Introduction 1-4

The network edge: Internet Services Models

 Connection-oriented service

 Connectionless service  peer-peer model: ❍ minimal (or no) use of dedicated servers  Applications ❍ e.g. Gnutella, KaZaA ❍ FTP, Internet Phone, Web, Internet radio, ❍ SETI@home? email

Introduction 1-5 Introduction 1-6

1 Connection-oriented service Connectionless service

 Goal: data transfer between end systems Goal: data transfer between end systems  handshaking: setup (prepare for) data transfer ahead of time ❍ same as before!  TCP - Transmission Control Protocol  UDP - User Datagram Protocol [RFC 768]: ❍ Internet’s connection-oriented service ❍ ❍ reliable, in-order byte-stream data transfer connectionless • loss: acknowledgements and retransmissions ❍ unreliable data transfer ❍ flow control: • sender won’t overwhelm receiver ❍ no flow control ❍ congestion control: ❍ • senders “slow down sending rate” when network congested no congestion control

What’s it good for? Introduction 1-7 Introduction 1-8

A Comparison The Network Core

App’s using TCP:  mesh of interconnected routers  HTTP (Web), FTP (file transfer), Telnet  the fundamental (remote login), SMTP (email) question: how is data transferred through net? App’s using UDP: ❍ circuit switching: dedicated circuit per  streaming media, teleconferencing, DNS, call: telephone net Internet telephony ❍ packet-switching: data sent thru net in discrete “chunks” Introduction 1-9 Introduction 1-10

Network Core: Circuit Switching Circuit Switching: FDM and TDM Example: FDM End-end resources 4 users reserved for “call”  link bandwidth, switch frequency capacity  dedicated resources: no sharing time  circuit-like (guaranteed) TDM performance  call setup required frequency

 must divide link bw into pieces... time Introduction 1-11 Introduction 1-12

2 Network Core: Packet Switching Packet Switching: Statistical Multiplexing

10 Mb/s C each end-end data stream divided into packets A Ethernet statistical multiplexing  user A, B packets share network resources  each packet uses full link bandwidth 1.5 Mb/s  resources used as needed B queue of packets resource contention: waiting for output  aggregate resource demand can exceed amount available link ❍ what happens if bandwidth is not available?  congestion: packets queue, wait for link use D E  store and forward: packets move one hop at a time ❍ Node receives complete packet before forwarding Sequence of A & B packets does not have fixed pattern  statistical multiplexing.

Introduction 1-13 Introduction 1-14

Packet switching versus circuit switching Packet switching versus circuit switching

Packet switching allows more users to use network! Is packet switching a “slam dunk winner?”

 1 Mb/s link  Great for bursty data  each user: ❍ resource sharing ❍ 100 kb/s when “active” ❍ simpler, no call setup ❍ active 10% of time N users  Excessive congestion: packet delay and loss 1 Mbps link ❍ protocols needed for reliable data transfer,  circuit-switching: congestion control ❍ 10 users  Circuit Switching = Guaranteed behavior  packet switching: ❍ good for which apps? ❍ with 35 users, probability > 10 active less than .0004 Introduction 1-15 Introduction 1-16

Packet-switching: store-and-forward Packet-switched networks: forwarding

L  Goal: move packets through routers from source to destination R R R ❍ we’ll study several path selection (i.e. routing) algorithms (chapter 4)  Takes L/R seconds to Example:  datagram network: transmit (push out)  L = 7.5 Mbits ❍ destination address in packet determines next hop packet of L bits on to ❍  R = 1.5 Mbps routes may change during session link or R bps ❍ analogy: driving, asking directions  delay = 15 sec  Entire packet must  virtual circuit network: arrive at router ❍ each packet carries tag (virtual circuit ID), tag determines next hop before it can be ❍ fixed path determined at call setup time, remains fixed thru call transmitted on next ❍ routers maintain per-call state link: store and forward  delay = 3L/R Introduction 1-17 Introduction 1-18

3 Network Taxonomy Access networks and physical media Telecommunication Q: How to connect end systems networks to edge router?  residential access nets Circuit-switched Packet-switched  institutional access networks networks networks (school, company)  mobile access networks Networks Datagram FDM TDM Keep in mind: with VCs Networks  bandwidth (bits per second) of access network? • Datagram network is not either connection-oriented  shared or dedicated? or connectionless. • Internet provides both connection-oriented (TCP) and connectionless services (UDP) to apps. Introduction 1-19 Introduction 1-20

Residential access: point to point access Residential access: cable modems

 Dialup via modem  HFC: hybrid fiber coax ❍ up to 56Kbps direct access ❍ asymmetric: up to 30Mbps downstream, 2 to router (often less) Mbps upstream ❍ Can’t surf and phone at same  network of cable and fiber attaches homes to time: can’t be “always on” ISP router  ADSL: asymmetric digital subscriber line ❍ homes share access to router ❍ up to 1 Mbps upstream (today typically < 256 kbps)  deployment: available via cable TV companies ❍ up to 8 Mbps downstream (today typically < 1 Mbps) ❍ FDM: 50 kHz - 1 MHz for downstream 4 kHz - 50 kHz for upstream

0 kHz - 4 kHz for ordinary telephone Introduction 1-21 Introduction 1-22

Cable Network Architecture: Overview Cable Network Architecture: Overview

Typically 500 to 5,000 homes

cable headend cable headend

home home cable distribution cable distribution network (simplified) network (simplified)

Introduction 1-23 Introduction 1-24

4 Company access: local area networks Wireless access networks

 company/univ local area  shared wireless access network (LAN) connects network connects end system end system to edge router to router router  Ethernet: ❍ via base station aka “access point” base ❍ shared or dedicated station  wireless LANs: link connects end ❍ 802.11b (WiFi): 11 Mbps system and router  wider-area wireless access ❍ 10 Mbs, 100Mbps, ❍ provided by telco operator mobile Gigabit Ethernet ❍ 3G ~ 384 kbps hosts  LANs: chapter 5 • Will it happen?? ❍ WAP/GPRS in Europe Introduction 1-25 Introduction 1-26

Home networks Physical Media

Typical home network components: Twisted Pair (TP)  Bit: propagates between  ADSL or cable modem  two insulated copper transmitter/rcvr pairs  router/firewall/NAT wires  physical link: what lies ❍ Category 3: traditional  Ethernet between transmitter & phone wires, 10 Mbps  wireless access receiver Ethernet ❍ Category 5: point wireless  guided media: 100Mbps Ethernet to/from laptops cable router/ ❍ signals propagate in solid cable modem firewall media: copper, fiber, coax headend wireless  unguided media: access Ethernet ❍ signals propagate freely, point e.g., radio Introduction 1-27 Introduction 1-28

Physical Media: coax, fiber Physical media: radio Fiber optic cable: Coaxial cable:  signal carried in Radio link types:  glass fiber carrying light electromagnetic spectrum  terrestrial microwave  two concentric copper pulses, each pulse a bit conductors  no physical “wire” ❍ e.g. up to 45 Mbps channels  high-speed operation:  bidirectional  bidirectional  LAN (e.g., Wifi) ❍ high-speed point-to-point ❍ 2Mbps, 11Mbps   propagation environment baseband: transmission (e.g., 5 Gps) effects:  wide-area (e.g., cellular) ❍ single channel on cable  low error rate: repeaters ❍ reflection ❍ e.g. 3G: hundreds of kbps ❍ legacy Ethernet spaced far apart ; immune to ❍ obstruction by objects  satellite  broadband: electromagnetic noise ❍ interference ❍ ❍ multiple channel on cable up to 50Mbps channel (or multiple smaller channels) ❍ HFC ❍ 270 msec end-end delay ❍ geosynchronous versus low altitude Introduction 1-29 Introduction 1-30

5 Internet structure: network of networks Tier-1 ISP: e.g., Sprint

 roughly hierarchical Sprint US backbone network  at center: “tier-1” ISPs (e.g., UUNet, BBN/Genuity, Sprint, AT&T), national/international coverage ❍ treat each other as equals Tier-1 providers also interconnect Tier-1 Tier 1 ISP at public network providers NAP access points interconnect (NAPs) (peer) privately Tier 1 ISP Tier 1 ISP

Introduction 1-31 Introduction 1-32

Internet structure: network of networks Internet structure: network of networks

 “Tier-2” ISPs: smaller (often regional) ISPs  “Tier-3” ISPs and local ISPs ❍ Connect to one or more tier-1 ISPs, possibly other tier-2 ISPs ❍ last hop (“access”) network (closest to end systems)

local local ISP Tier 3 local local Tier-2 ISPs ISP ISP ISP ISP Local and tier- Tier-2 ISP pays Tier-2 ISP Tier-2 ISP also peer Tier-2 ISP Tier-2 ISP tier-1 ISP for privately with 3 ISPs are Tier 1 ISP Tier 1 ISP connectivity to each other, customers of NAP NAP rest of Internet interconnect higher tier  tier-2 ISP is at NAP ISPs customer of connecting Tier 1 ISP Tier 1 ISP Tier-2 ISP Tier 1 ISP Tier 1 ISP Tier-2 ISP tier-1 provider them to rest local Tier-2 ISP Tier-2 ISP of Internet Tier-2 ISP Tier-2 ISP ISP local local local ISP Introduction 1-33 ISP ISP Introduction 1-34

Internet structure: network of networks How do loss and delay occur?

 a packet passes through many networks! packets queue in router buffers  packet arrival rate to link exceeds output link capacity local  packets queue, wait for turn local ISP Tier 3 local local ISP ISP ISP ISP packet being transmitted (delay) Tier-2 ISP Tier-2 ISP Tier 1 ISP NAP A

Tier 1 ISP Tier 1 ISP Tier-2 ISP B local Tier-2 ISP Tier-2 ISP packets queueing (delay) ISP local local local free (available) buffers: arriving packets ISP ISP ISP Introduction 1-35 dropped (loss) if no free buffers Introduction 1-36

6 Four sources of packet delay Delay in packet-switched networks

3. Transmission delay: 4. Propagation delay:  1. nodal processing:  2. queueing  R=link bandwidth (bps)  d = length of physical link ❍ check bit errors ❍ time waiting at output   ❍ determine output link link for transmission L=packet length (bits) s = propagation speed in 8 ❍ depends on congestion  time to send bits into medium (~2x10 m/sec) level of router link = L/R  propagation delay = d/s

transmission Note: s and R are very A propagation transmission different quantities! A propagation B nodal processing queueing B nodal queueing Introduction 1-37 processing Introduction 1-38

Nodal delay “Real” Internet delays and routes d = d + d + d + d nodal proc queue trans prop  What do “real” Internet delay & loss look like?  Traceroute program: provides delay

 dproc = processing delay measurement from source to router along end- ❍ typically a few microsecs or less end Internet path towards destination. For all i: ❍ sends three packets that will reach router i on path  dqueue = queuing delay towards destination ❍ depends on congestion ❍ router i will return packets to sender  dtrans = transmission delay ❍ sender times interval between transmission and reply. ❍ = L/R, significant for low-speed links 3 probes 3 probes

 dprop = propagation delay 3 probes ❍ a few microsecs to hundreds of msecs

Introduction 1-39 Introduction 1-40

“Real” Internet delays and routes Packet loss

traceroute: gaia.cs.umass.edu to www.eurecom.fr Three delay measements from  queue (aka buffer) preceding link in buffer gaia.cs.umass.edu to cs-gw.cs.umass.edu 1 cs-gw (128.119.240.254) 1 ms 1 ms 2 ms has finite capacity 2 border1-rt-fa5-1-0.gw.umass.edu (128.119.3.145) 1 ms 1 ms 2 ms 3 cht-vbns.gw.umass.edu (128.119.3.130) 6 ms 5 ms 5 ms 4 jn1-at1-0-0-19.wor.vbns.net (204.147.132.129) 16 ms 11 ms 13 ms  when packet arrives to full queue, packet 5 jn1-so7-0-0-0.wae.vbns.net (204.147.136.136) 21 ms 18 ms 18 ms 6 abilene-vbns.abilene.ucaid.edu (198.32.11.9) 22 ms 18 ms 22 ms 7 nycm-wash.abilene.ucaid.edu (198.32.8.46) 22 ms 22 ms 22 ms trans-oceanic is dropped (aka lost) 8 62.40.103.253 (62.40.103.253) 104 ms 109 ms 106 ms 9 de2-1.de1.de.geant.net (62.40.96.129) 109 ms 102 ms 104 ms link 10 de.fr1.fr.geant.net (62.40.96.50) 113 ms 121 ms 114 ms  lost packet may be retransmitted by 11 renater-gw.fr1.fr.geant.net (62.40.103.54) 112 ms 114 ms 112 ms 12 nio-n2.cssi.renater.fr (193.51.206.13) 111 ms 114 ms 116 ms previous node, by source end system, or 13 nice.cssi.renater.fr (195.220.98.102) 123 ms 125 ms 124 ms 14 r3t2-nice.cssi.renater.fr (195.220.98.110) 126 ms 126 ms 124 ms 15 eurecom-valbonne.r3t2.ft.net (193.48.50.54) 135 ms 128 ms 133 ms not retransmitted at all 16 194.214.211.25 (194.214.211.25) 126 ms 128 ms 126 ms 17 * * * 18 * * * * means no reponse (probe lost, router not replying) 19 fantasia.eurecom.fr (193.55.113.142) 132 ms 128 ms 136 ms

Introduction 1-41 Introduction 1-42

7 Protocol “Layers” Organization of air travel

Networks are complex! ticket (purchase) ticket (complain)  many “pieces”: baggage (check) baggage (claim) ❍ hosts Question:

❍ routers Is there any hope of gates (load) gates (unload) organizing structure of ❍ links of various runway takeoff runway landing media network?

❍ applications airplane routing airplane routingairplane routing Or at least our discussion ❍ protocols of networks? ❍ hardware,  a series of steps software

Introduction 1-43 Introduction 1-44

Layering of airline functionality Why layering?

Dealing with complex systems: ticket (purchase) ticket (complain) ticket  explicit structure allows identification, baggage (check) baggage (claim baggage

gates (load) gates (unload) gate relationship of complex system’s pieces runway (takeoff) runway (land) takeoff/landing ❍ layered reference model for discussion airplane routing airplane routing airplane routing airplane routing airplane routing  departure intermediate air-traffic arrival modularization eases maintenance, updating of airport control centers airport system

Layers: each layer implements a service ❍ change of implementation of layer’s service transparent to rest of system ❍ via its own internal-layer actions ❍ e.g., change in gate procedure doesn’t affect ❍ relying on services provided by layer below rest of system

Introduction 1-45  layering considered harmful? Introduction 1-46

source Encapsulation Internet protocol stack message M application segment Ht M transport datagram H H M  application: supporting network applications n t network frame Hl Hn Ht M link ❍ FTP, SMTP, STTP application physical  transport: host-host data transfer Hl Hn Ht M link Hl Hn Ht M ❍ TCP, UDP transport physical  network: routing of datagrams from source to destination switch network ❍ IP, routing protocols  link: data transfer between neighboring link H H M network elements destination n t network Hn Ht M Hl Hn Ht M link H H H M ❍ PPP, Ethernet M application l n t physical H M transport physical  physical: bits “on the wire” t Hn Ht M network router Hl Hn Ht M link physical

Introduction 1-47 Introduction 1-48

8