I NetworksII HistoryIII Key IdeasIII-A Packet SwitchingIII-B End-to-End Error and Flow ControlIII-C InternetworkingIII-D Multiple AccessIII-E DNS:Domain Name SystemIII-F P2P: Peer-to-peerIII-G VOIP:Voice over IPIV ProtocolsV TechnologyV-A Communication LinksV-B SwitchesReferences1EE228a - Lecture 2 - Spring 2006InternetJean Walrand, Scribed by Bonnie ZhuAbstractIn today’s class, we cover the topic of the Internet with emphasis on review of TCP/IP. ( A good reference is [1] and a slightlymore readable but less comprehensive is [2]I. NETWORKSCommunication networks are usually defined by their size and complexity. From top-down, we have• Wide area networks(WAN) These networks connect computers and other terminals over large distances. They often requiremultiple communication connections, including microwave radio links and satellite. WAN consists of the back-bone networksthat are connected via high capacity fiber optics cables.• Metropolitan area networks(MAN) These networks are interconnected within a WAN and are spread around , say, a town orcity. This kind of network is a high speed network using optical fiber connections.• Local area networks(LAN) These networks are components of a MAN to connect computer and other terminals distributedin a localized area. The connection is usually cable (twisted pair copper) or fiber. Figure 1 shows the possible componentsof a LANFig. 1. The possible components of a LANII. HISTORYTable I lists the milestones in the development of the Internet.TABLE IMILESTONES IN DEVELOPMENT OF INTERNET1962 L. Kleinrock proposes Packet Switching1966 L. Roberts proposes architecture to Darpa1969 First demonstration of packet switching: 4-node Arpanet1969 S. Croker introduces RFCs (managed by J. Postel)1972 R. Kahn proposes an open architecture: Inter-networking with stateless routers, best effort, no control plane1973 Kahn and V. Cerf propose early ideas of IP (32-bit address) and TCP (end-to-end ACKs with a window scheme)1973 R. Metcalfe invents Ethernet1982 IGP and EGPLate 1970s, early 1980s Berkeley develops BSD, a modified implementation of UNIX that includes TCP/IP1983 Arpanet switches to TCP/IP1983 P. Mockapetris invents DNS1988 Van Jacobson fixes TCP1985 1995 Internet supported by NSF and other agencies1993? T. Berners-Lee invents WWW1995 Internet is privatized2000 WiFi,VoIP, P2P2III. KEY IDEASWe will address some of the key ideas to the development of Internet.• Packet Switching• End-to-End error and flow control• internetworking• Multiple Access• DNS• P2P• VoIPA. Packet SwitchingFigure 2 illustrates the idea of packet switching. Packets (units of information carriage) are individually routed between nodesover data links which might be shared by many other nodes. Once all the packets forming a message arrive at the destination, theyare recompiled into the original message. This contrasts with circuit switching, which sets up a dedicated connection between thetwo nodes for their exclusive use for the duration of the communication. Circuit-switching is ideal when data must be transmittedquickly and must arrive in the same order in which it’s sent. This is the case with most real-time data, such as live audio andvideo. Because packets follow different routes, packet switching introduces different delays and jitter. It is more efficient androbust for data that can withstand some delays in transmission, such as e-mail messages and Web pages.In Packet Switching, the intermediate routers are required to know the networking structures, i.e., how the connections aremade between huge numbers of servers and clients. This turns out to be an issue for graph encoding, as well as for the applicationsin DNS. In contrast to traditional methods relying on the canonical ordering of a connected plannar graph, orderly spanning treesare much more adaptable to create efficient and easy encodings [3], [4].Fig. 2. Packet SwitchingB. End-to-End Error and Flow ControlThe purpose of end-to-end error and flow control is to ensure complete data transfer. As shown in Figure 3, node A canhave up to N packets for node B that have not been acknowledged. By selecting N, A adjusts the rate of transmissions to enableB handle all the incoming data without buffer-overflow thus limit congestion . This is particularly important where the sendingdevice is capable of sending data much faster than the receiving device can receive it.Error control is implemented as two separate functions: error detection and retransmission. As depicted in Figure 3, if ACKis late, i.e., an error is detected, node A retransmits packets.Fig. 3. End-to-End Error and Flow ControlC. InternetworkingThe goal of internetworking is to connect different networks. As illustrated in Figure 4, for a packet travels through twonetworks, only local addresses are involved when it’s within one network, then router that connects two networks looks at IPaddress to transfer it to the other network. Once the packet is delivered into the other network, only local address are used again.3Fig. 4. InternetworkingD. Multiple AccessThe scheme allows temporary access to the network by individual users, on a demand basis, for the purpose of transmittinginformation while sharing one medium. (e.g.,WiFi,Ethernet,..). As illustrated in Figure 5, if there is a collision detected, thenrandom waiting time will be allocated after collision. Or to avoid collision, when the network is detected to be busy, the schemeensures that it will wait till idle before more transmission. Note that WiFi uses collision avoidance whereas wired schemes usuallyuse collision detection.Fig. 5. Multiple AccessE. DNS:Domain Name SystemThe Domain Name System(DNS) translates domain names into IP addresses, such as google.com into 216.239.57.99 . Thedirectory of DNS is a distributed database. The mapping between domain name and IP address can depend on source address andalso can be done to multiple addresses.F. P2P: Peer-to-peerThe purpose of P2P scheme is to allow all peers become servers thus to utilize the computing power and bandwidth of theparticipants in the network. The examples can be a case that each peer knows the addresses of a set of peers: his friends or acase that to find a file, peer A asks his friends who in turn ask their friends, and so on, until someone, B, says he has the file(can limit search); A then asks B for the file. Many variations are possible. The key idea is that there is no specialized server;every client becomes a
View Full Document
Unlocking...