If the Internet is the answer, then what was the question?

Transcription

1 If the Internet is the answer, then what was the question? EE122 Fall 2011 Scott Shenker Materials with thanks to Jennifer Rexford, Ion Stoica, Vern Paxson and other colleagues at Princeton and UC Berkeley 1

2 Administrivia No sections next week (Monday is a holiday) Enrollment: admitted 20 additional seniors We are now at our legal limit; there is nothing I can do Tuesday section: moved to Cory 521 Need feedback on the index cards Can do it anonymously on Piazza Or send to me Plug computers arrived! Now we just have to get them to work..(go Yahel!) Get instructional account forms in section 2

3 Outline for today s class The telephone network boring Why does the Internet use packet switching? Less boring, but hardly earthshattering Important life lessons This will change your life. Internet history From my somewhat heretical viewpoint 3

4 Telephone network Alexander Graham Bell 1876: Demonstrates the telephone at US Centenary Exhibition in Philadelphia

5 Telephone network uses circuit switching Establish: source creates circuit to destination Nodes along the path store connection info And reserve resources for the connection If circuit not available: Busy signal Transfer: source sends data over the circuit No destination address, since nodes know path Continual stream of data Teardown: source tears down circuit when done 5

6 The switch in circuit switching incoming links Node outgoing links How does the node connect the incoming link to the outgoing link? 6

7 Circuit Switching With Human Operator 7

8 Modern switches Almon Brown Strowger ( ) 1889: Invents the girl-less, cuss-less telephone system -- the mechanical switching system

9 Timing in Circuit Switching Host 1 Host 2 Switch 1 Switch 2 time 9

10 Timing in Circuit Switching Host 1 Host 2 Switch 1 Switch 2 Circuit Establishment propagation delay between Host 1 and Switch1 time 10

11 Timing in Circuit Switching Host 1 Host 2 Switch 1 Switch 2 Circuit Establishment Transmission delay propagation delay between Host 1 and Switch1 time 11

12 Timing in Circuit Switching Host 1 Host 2 Switch 1 Switch 2 Circuit Establishment Transmission delay propagation delay between Host 1 and Switch1 time 12

13 Timing in Circuit Switching Host 1 Host 2 Switch 1 Switch 2 Circuit Establishment Transmission delay propagation delay between Host 1 and Switch1 propagation delay between Host 1 and Host 2 time 13

14 Timing in Circuit Switching Host 1 Host 2 Switch 1 Switch 2 Circuit Establishment Transmission delay propagation delay between Host 1 and Switch1 propagation delay between Host 1 and Host 2 Transfer Information time 14

15 Timing in Circuit Switching Host 1 Host 2 Switch 1 Switch 2 Circuit Establishment Transmission delay propagation delay between Host 1 and Switch1 propagation delay between Host 1 and Host 2 Transfer Circuit Teardown Information time 15

16 Sharing a link incoming links Node outgoing links How do the black and orange circuits share the outgoing link? 16

17 Circuit Switching: Multiplexing a Link Time-division Each circuit allocated certain time slots Frequency-division Each circuit allocated certain frequencies time frequency time 17

18 Time-Division Multiplexing/Demultiplexing Frames Slots = Time divided into frames; frames into slots Relative slot position inside a frame determines to which conversation data belongs E.g., slot 0 belongs to orange conversation Requires synchronization between sender and receiver Need to dynamically bind a slot to a conversation If a conversation does not use its circuit capacity is lost! 18

19 Strengths of phone system Predictable performance Known delays No drops Easy to reason about Supports a crucial service What about weaknesses? 19

20 Weakness #1: Not resilient to failure Any failure along the path prevents transmission Entire transmission has to be restarted All or nothing delivery model 20

21 Weakness #2: Wastes bandwidth Consider a network application with: Peak bandwidth P Average bandwidth A How much does the network have to reserve for the application to work? The peak bandwidth What is the resulting level of utilization? Ratio of A/P 21

22 Smooth vs Bursty Applications Some applications have relatively small P/A ratios Voice might have a ratio of 3:1 or so Data applications tend to be rather bursty Ratios of 100 or greater are common Circuit switching too inefficient for bursty apps Generally: Don t care about factors of two in performance But when it gets to several orders of magnitude. 22

23 Weakness #3: Designed Tied to App Design revolves around the requirements of voice Not general feature of circuit switching But definitely part of the telephone network design 23

24 Weakness #4: Setup Time Every connection requires round-trip time to set up Slows down short transfers In actuality, may not be a big issue TCP requires round-trip time for handshake No one seems to mind. This was a big issue in the ATM vs IP battle But I think it is overemphasized as a key factor 24

25 What if we wanted a resilient network? How would we design it? This is the question Paul Baran asked. 25

26 Paul Baran Baran investigated survivable networks for USAF Network should withstand almost any degree of destruction to individual components without loss of endto-end communications. On Distributed Communications (1964) Distributed control Message blocks (packets) Store-and-forward delivery 26

27 What about a less wasteful network? How would we design it? This is the question Len Kleinrock asked.. Analyzed packet switching and statistical multiplexing 27

28 Taxonomy of Communication Networks Communication networks can be classified based on the way in which the nodes exchange information: Communication Network 28

29 Taxonomy of Communication Networks Communication networks can be classified based on the way in which the nodes exchange information: Communication Network Broadcast Communication Network 29

30 Broadcast Communication Networks Information transmitted by any node is received by every other node in the network Usually only in LANs (Local Area Networks) E.g., WiFi, Ethernet (classical, but not current) E.g., lecture! What problems does this raise? Problem #1: limited range Problem #2: coordinating access to the shared communication medium Multiple Access Problem Problem #3: privacy of communication 30

31 Taxonomy of Communication Networks Communication networks can be classified based on the way in which the nodes exchange information: Communication Network Switched Communication Network Broadcast Communication Network 31

32 Taxonomy of Communication Networks Communication networks can be classified based on the way in which the nodes exchange information: Communication Network Switched Communication Network Broadcast Communication Network Circuit-Switched Communication Network 32

33 Taxonomy of Communication Networks Communication networks can be classified based on the way in which the nodes exchange information: Communication Network Switched Communication Network Broadcast Communication Network Circuit-Switched Communication Network Packet-Switched Communication Network 33

34 Packet Switching Data sent as chunks of formatted bit-sequences (Packets) Packets have following structure: Header Data Trailer (sometimes) Header and Trailer carry control information (e.g., destination address, checksum) Each packet traverses the network from node to node along some path (Routing) based on header info. Usually, once a node receives the entire packet, it stores it (hopefully briefly) and then forwards it to the next node (Store-and-Forward Networks) 34

35 Packet Switching Node in a packet switching network incoming links Node Memory outgoing links 35

36 Packet Switching: Multiplexing/Demultiplexing How to tell packets apart? Use meta-data (header) to describe data No reserved resources; dynamic sharing Single flow can use the entire link capacity if it is alone This leads to increased efficiency 36

37 Simple Example: M/M/1 Queue Consider n flows sharing a single queue Flow: random (Poisson) arrivals at rate λ Random (Exponential) service with average 1/µ Utilization factor: ρ = nλ/µ If ρ >1, system is unstable Case 1: Flows share bandwidth Delay = 1/(µ - nλ) Case 2: Flows each have 1/n th share of bandwidth No sharing Delay = n/(µ - nλ) Not sharing increases delay by n 37

38 Taxonomy of Communication Networks Communication networks can be classified based on the way in which the nodes exchange information: Communication Network Switched Communication Network Broadcast Communication Network Circuit-Switched Communication Network Packet-Switched Communication Network Datagram Network 38

39 Datagram Packet Switching Each packet is independently switched Each packet header contains full destination address No resources are pre-allocated (reserved) in advance Leverages statistical multiplexing Gambling that packets from different conversations won t all arrive at the same time, so we don t need enough capacity for all of them at their peak transmission rate Assuming independence of traffic sources, can compute probability that there is enough capacity 39

40 Timing of Datagram Packet Switching Host 1 Host 2 Node 1 Node 2 Packet 1 propagation delay between Host 1 and Node 1 40

41 Timing of Datagram Packet Switching Host 1 Host 2 Node 1 Node 2 transmission time of Packet 1 at Host 1 Packet 1 propagation delay between Host 1 and Node 1 41

42 Timing of Datagram Packet Switching Host 1 Host 2 Node 1 Node 2 transmission time of Packet 1 at Host 1 Packet 1 propagation delay between Host 1 and Node 1 Packet 1 processing delay of Packet 1 at Node 2 Packet 1 42

43 Timing of Datagram Packet Switching Host 1 Host 2 Node 1 Node 2 transmission time of Packet 1 at Host 1 Packet 1 Packet 2 Packet 3 propagation delay between Host 1 and Node 1 Packet 1 Packet 2 Packet 3 processing delay of Packet 1 at Node 2 Packet 1 Packet 2 Packet 3 43

44 Datagram Packet Switching Host C Host A Host D Node 1 Node 2 Node 3 Node 5 Host B Node 4 Node 6 Node 7 Host E 44

45 Datagram Packet Switching Host C Host A Host D Node 1 Node 2 Node 3 Node 5 Host B Node 4 Node 6 Node 7 Host E 45

46 Datagram Packet Switching Host C Host A Host D Node 1 Node 2 Node 3 Node 5 Host B Node 4 Node 6 Node 7 Host E 46

47 Taxonomy of Communication Networks Communication networks can be classified based on the way in which the nodes exchange information: Communication Network Switched Communication Network Circuit-Switched Communication Network Packet-Switched Communication Network Broadcast Communication Network A hybrid of circuits and packets; headers include a circuit identifier established during a setup phase Datagram Network Virtual Circuit Network 47

48 5 Minute Break Questions Before We Proceed? 48

49 If you were building a network. Which would you choose? Circuit switched? Packet-switched? Let s review the strengths and weaknesses 49

50 Advantages of Circuit Switching Guaranteed bandwidth Predictable communication performance Not best-effort delivery with no real guarantees Simple abstraction Reliable communication channel between hosts No worries about lost or out-of-order packets Simple forwarding Forwarding based on time slot or frequency No need to inspect a packet header Low per-packet overhead Forwarding based on time slot or frequency No headers on each packet 50

51 Disadvantages of Circuit Switching Wasted bandwidth Bursty traffic leads to idle connection during silent period Unable to achieve gains from statistical multiplexing Blocked connections Connection refused when resources are not sufficient Unable to offer okay service to everybody Network state Network nodes must store per-connection information Unable to avoid per-connection storage and state This makes failures more disruptive! Connection set-up delay No communication until the connection is set up Unable to avoid extra latency for small data transfers 51

52 Packet-Switching vs. Circuit-Switching Reliability advantage: since routers don t know about individual conversations, when a router or link fails, it is easy to fail over to a different path Efficiency advantage of packet-switching over circuit switching: Exploitation of statistical multiplexing Deployability advantage: easier for different parties to link their networks together because they re not promising to reserve resources for one another Disadvantage: packet-switching must handle congestion More complex routers (more buffering, sophisticated dropping) Harder to provide good network services (e.g., delay and bandwidth guarantees) 52

53 Choosing a Design for the Internet If you cared about: Resilience Bursty applications Ease of interconnection Which would you choose? Not so fast.. Hindsight is great But there were important reasons to choose differently 53

54 The paradox of the Internet s design As we will discuss next lecture, one of the main design goals is to support a wide range of apps These applications have different requirements Shouldn t the Internet support them all? 54

55 Diversity of application requirements Size of transfers Bidirectionality (or not) Latency sensitive (or not) Tolerance of jitter (or not) Tolerance of packet drop (or not) Need for reliability (or not) Multipoint (or not).. 55

56 Diversity of application requirements Size of transfers Bidirectionality (or not) Latency sensitive (or not) Tolerance of jitter (or not) Tolerance of packet drop (or not) Need for reliability (or not) Multipoint (or not).. 56

57 What service should Internet support? Strict delay bounds? Some applications require them Guaranteed delivery? Some applications are sensitive to packet drops No applications mind getting good service Why not require Internet support these guarantees? 57

58 Important life lessons People (applications) don t always need what they think they need People (applications) don t always need what we think they need Flexibility often more important than performance But typically only in hindsight! Example: cell phones vs landlines Architect for flexibility, engineer for performance 58

59 Applying lessons to Internet Requiring performance guarantees would limit variety of networks that could attach to Internet Many applications don t need these guarantees And those that do? Well, they don t either (usually) Tremendous ability to mask drops, delays And ISPs can work hard to deliver good service without changing the architecture If the Internet had focused on voice applications early, it might have made different choices 59

60 Internet History

61 Timeline 1961 Baran and Kleinrock advocate packet switching 1962 Licklider s vision of Galactic Network 1965 Roberts connects two computers via phone 1967 Roberts publishes vision of ARPANET 1969 BBN installs first IMP at UCLA IMP: Interface Message Processor 1971 Network Control Program (protocol) 1972 Public demonstration of ARPANET 61

62 The beginning of the Internet revolution Kleinrock s group at UCLA tried to log on to Stanford computer: His recollection of the event We typed the L Do you see the L? Yes, we see the L. We typed the O Do you see the O? Yes, we see the O. Then we typed the G and the system crashed! 62

63 Timeline continued invented 1972 Telnet introduced 1972 Kahn advocates Open Architecture networking 63

64 The Problem Many different packet-switching networks Only nodes on the same network could communicate 64

65 Kahn s Rules for Interconnection Each network is independent and must not be required to change (why?) Best-effort communication (why?) Boxes (routers) connect networks No global control at operations level (why?) 65

66 Solution Gateways 66

67 Kahn s vision Kahn imagined there would be only a few networks (~20) and thus only a few routers He was wrong Why? Imagined gateways would translate between networks We think of it as all routers supporting IP 67

68 Timeline continued FTP introduced 1974 Cerf and Kahn paper on TCP/IP 1980 TCP/IP adopted as defense standard 1983 Global NCP to TCP/IP flag day 198x XNS, DECbit, and other protocols 1984 Janet (British research network) 1985 NSFnet (picks TCP/IP) 198x Internet meltdowns due to congestion 1986 Van Jacobson saves the Internet (BSD TCP) 68

69 Unsung hero of Internet: David D. Clark Chief Architect Great consistency of vision Kept the Internet true to its basic design principles Authored what became known as the End-to-end principle (next lecture) Conceives and articulates architectural concepts Read his Active Networking and End-To-End Arguments Perhaps the only irreplaceable Internet pioneer 69

70 Timeline continued 1988 Deering and Cheriton propose multicast 1989 Birth of the web.tim Berners-Lee 70

71 Why did it take physicist to invent web? Physicists are the smartest people in the world? Computer scientists were trying to invent nirvana Well, actually Xanadu (Ted Nelson) More generally, CS researchers focused on hyptertext Again, users didn t need what we wanted to invent Think about it: a paper on the web design would have been rejected by every CS conference and journal In general, the CS research community is great at solving well-defined problems, but terrible at guessing what users will actually use 71

72 Timeline continued Search engines invented (Excite) 199x ATM rises and falls (as internetworking layer) 199x QoS rises and falls 1994 Internet goes commercial 1998 IPv6 specification 1998 Google reinvents search 200x The Internet boom and bust 2011 EE122 enrollment suggests boom is back! 72

73 Next Lecture(s) Monday is a holiday Wednesday: Internet Priorities and Principles HW #1 assigned on Wednesday If you haven t already: Take the survey Have a good holiday! 73