CIS5930 Internet ComputingOriginsDefinitionGlobal ConnectivityInternet PenetrationGrowth of the Internet (1)Growth of the Internet (2)Internet StructureSome Nodes are Highly ConnectedPower LawRandom Networks versus Scale-Free NetworksInternet ConnectivityRobustness to FailuresThe Internet Under AttackRethinking the Design of the InternetA Summary of Requirements for Today’s Internet (1)A Summary of Requirements for Today’s Internet (2)End-to-end versus Network Centric ViewsApplication and Network Boundaries are BlurringCIS5930Internet Computing1. Internet BasicsProf. Robert van EngelenCIS 5930 Fall 2006 - Internet Basics 201/13/19Origins1968: First proposal for ARPANET1971: ARPANET in use1973: Redesign of lower-level protocols leads to TCP/IP1983: Birth of InternetNSF constructs the University network backbone NSFnetBerkeley TCP/IP implementation for 4.2BSD in public domain1980s: Rapid growth of NSFnet for broad academic useAdditional networks (Usenet, Bitnet) merge1991: Tim Barners-Lee introduced HTTP and HTML1990s: WWW, public access, and commercializationToday: 1.04 Billion users (Internet World Stats)CIS 5930 Fall 2006 - Internet Basics 301/13/19DefinitionAn internet isA collection of interconnected networksThe Internet isA particular internet whichUses TCP/IP protocolsIs globalIs hardware and network independentSupports commonly-used applications (WWW, mail, rlogin, …)Uses publicly available standards (RFC, W3C)The Internet is not just the WebCIS 5930 Fall 2006 - Internet Basics 401/13/19Global ConnectivityThe Internet is all about global connectivityNetworks connected togetherEthernet, FDDI, Wi-Fi, token ring, AppleTalkEach may use a local communication protocol that sits below TCP/IPYou’re on the Internet when one of the networks is connected to a central backbone, e.g. an ISPCIS 5930 Fall 2006 - Internet Basics 501/13/19Internet PenetrationCIS 5930 Fall 2006 - Internet Basics 601/13/19Growth of the Internet (1)The net total number of domains on the Internet as of 2006(number of existing domains + new registrations - expired domains)010,000,00020,000,00030,000,00040,000,00050,000,00060,000,00070,000,0001998 1999 2000 2001 2002 2003 2004 2005 2006Source: ZooknicCIS 5930 Fall 2006 - Internet Basics 701/13/19Growth of the Internet (2)A high percentage of active domains are .com010,000,00020,000,00030,000,00040,000,00050,000,00060,000,00070,000,0001998 1999 2000 2001 2002 2003 2004 2005 2006010,000,00020,000,00030,000,00040,000,00050,000,00060,000,00070,000,00012006.com.net.org .biz.info.eduCIS 5930 Fall 2006 - Internet Basics 801/13/19Internet StructureQuestions:Is the Interneta random graph?Does it matter?The Internet appearsto grow randomly(no single entityhas oversight)CIS 5930 Fall 2006 - Internet Basics 901/13/19Some Nodes are Highly ConnectedMap of some 100 000 Internet routersand the physical connectionsbetween themImage source: www.physicsweb.org/articles/world/14/7/9You are hereBill is thereA few highly connectednodes (hubs) keep the network togetherCIS 5930 Fall 2006 - Internet Basics 1001/13/19Power LawMany man-made and naturally occurring phenomena are distributed according to a power-law distribution (a.k.a. Zipf and Pareto)City sizesIncomesWord frequenciesEarthquake magnitudesCompare:Very few people are famous, most aren’tA few highly connected nodes keep the Internet network together, most are at the fringe of the networkWe say that the Internet as a network exhibits scale-free propertiesThe router connections and hyperlinks to network nodes are distributed according to the power lawScale-free models view networks as dynamical systems that self-assemble and evolve over time with preferential attachmentCIS 5930 Fall 2006 - Internet Basics 1101/13/19Random Networks versus Scale-Free NetworksRandom networks are homogeneous: most nodes have approximately the same number of links k ~ <k>, where <k> is the average value(a) Example N=10 and <k>=p=0.2(c) Probability P(k) that a node has k links has peak at <k>(e) A random network with N=130Scale-free networks are inhomogeneous(b) Network continually grows: a new node (red) connects to two existing nodes in the network (black) at time t+1(d) Power law: P(k) ~ k-g at large k, g > 1(f) Scale-free network with N=130More than 60% of nodes (green) can be reached from the five most connected nodes (red) compared with only 27% in the random networkCIS 5930 Fall 2006 - Internet Basics 1201/13/19Internet Connectivity(a) Web pages are hyperlinked, where each page haskout number of outgoing URLskin incoming references; large kin means high popularity (and higher Google rank)(b) The Internet is a network of routers that navigate packets of data from one computer to another, where groups of routers and computers form "domains"(c) The probability that a Web page has kin (blue) or kout (red) links follows a power law distribution(d) The degree distribution (green) of the Internet at the router level, where k denotes the number of links a router has to other routersCIS 5930 Fall 2006 - Internet Basics 1301/13/19Robustness to FailuresExtreme robustness to failures is rooted in the inhomogeneous topology of the networkPercolation theory: a random network will break into tiny clusters when a critical fraction, fc, of nodes is removedThis prediction does not hold for scale-free networks: 80% of the nodes have to fail before the Internet network fragmentsBut if the most-connected nodes are removed then the scale-free network will break at a small fcCIS 5930 Fall 2006 - Internet Basics 1401/13/19The Internet Under AttackSource: SANS ISCOrigin of port scansCIS 5930 Fall 2006 - Internet Basics 1501/13/19Rethinking the Design of the InternetLack of trust is driving a rethinking of the Internet’s designThe simple model of the early Internet - a group of mutually trusting users attached to a transparent network - is gone foreverMany forces are pushing the Internet away from the end-to-end arguments that guided the initial Internet designEnd-to-end applications solely define behaviors, network is just a communication medium (simple, extensible, and network-independent)Emerging requirements add mechanisms to the core of the network:Operation in an untrustworthy world: enforce “good” behavior at the centerMore demanding applications: core network support for streaming mediaISP service
View Full Document