Transcript Chapter 2
Review of Underlying Network Technologies Chapter 2 Network Communication Internet is not a new kind of physical network Method of interconnecting physical networks Set of conventions for using networks Communication networks can be divided into two basic types Connection-oriented (circuit-switched) Connectionless (packet-switched) Connection-Oriented Forms a dedicated connection or circuit between two points Like U.S. telephone system Guarantees capacity Once circuit is established, no other network activity will decrease the circuit’s capacity Disadvantage is cost Circuit costs are fixed, regardless of use Connectionless Data to be transmitted is divided into packets Usually only a few hundred bytes Carries identification information Allows concurrent communication Multiple computers over shared medium Disadv: As activity increases, given pair of computers receives less of the network capacity Who Wins? Connectionless! Despite not being able to guarantee capacity Wins via cost and performance Sharing network bandwidth means fewer connections are required Performance ok since we can build high speed network hardware Throughout the text, network means connectionless network WANs and LANs WAN: spans large geographical distance Sometimes called long-haul networks Usually do not have any distance limit Slower: 1.5 Mbps to 2.4 Gbps More delay: few ms to several tenths of a second Usually consists of a series of packet switches interconnected by long distance comm lines Extend network by new switch and comm line Computers added by attaching to a packet switch LAN: spans short geographical distance Fast: 100 Mbps to 10 Gbps Less delay: few tenths of ms to 10 ms Each computer usually contains a Network Interface Card Connects machine directly to network Network is “dumb”; interface devices in the computers do the work Every computer attached to a network has a unique address* Sender must know recipient’s address Hardware technology specifies address scheme *Will change this statement slightly soon! Ethernet Technology Ethernet is a packet-switched LAN Invented by Xerox in early 1970’s Standardized by Xerox, Intel, and DEC in 1978 IEEE standard number 802.3 Many variants exist Original design known as 10Base5 Uses coaxial cable approximately ½ inch in diameter and up to 500 meters long Cable is completely passive 10 Mbps Original wiring scheme Has been superseded Thin-wire Ethernet Known as 10Base2 Some original Ethernet disadvantages: Transceiver has non-trivial cost Transceivers located with cable Cable difficult to install (thick shield; hard to bend) Thinnet cable thinner, cheaper, more flexible Computer has both the host interface and connection circuitry Easy to connect and disconnect (no technician) Less protection from interference; shorter distances; fewer connections per network Twisted Pair Ethernet Known as 10Base-T Popular, current technology Uses conventional unshielded copper wire Cheaper and easier to install Each computer connects to a hub over 4 pairs of wires Only 2 pairs of wires used Same communication capability as thick or thin Ethernet; just alternate wiring scheme Fast Ethernet Thick, thin, twisted pair: 10 Mbps Faster processors Ethernet became bottleneck Developed 100Base-T (100 Mbps) Uses same twisted pair wires Gigabit Ethernet Known as 1000Base-T (1 Gbps, copper)) 1000Base-X – uses fiber optics Fiber is much faster Developing 10 and 40 Gbps Ethernet technologies 10/100/1000 Ethernet Allows compatibility with either 10Base-T, 100Base-T, or 1000Base-T Can use for computer interfaces or hubs Computer with 10/100/1000 interface card can attach to any of the 3 configurations Hardware automatically detects speed No hardware or software reconfiguration required Power over Ethernet Small amount of power Sent over same copper cable Power does not degrade data transmission Can power small devices with one cable Two facts about increased capacity (1) Few computers can sustain 1 Gbps data rate (2) New versions did not change standards Max packet size same as for 10Base-T Higher-speeds not optimized for highest possible computer-to-computer throughput Allows more stations and more total traffic Ethernet properties Shared bus that supports broadcast All stations connect to single shared channel All stations receive every transmission Uses best-effort delivery Sender gets no information about packet delivery Distributed access control No central authority to grant access to shared channel Carrier Sense Multiple Access with Collision Detect (CSMA/CD) Host listens before transmit; sends if idle Maximum packet size limits transmission time Must observe minimum idle time between sends Collision detection and recovery Signals travel at approx 70% speed of light Stations can begin transmitting simultaneously (or almost simultaneously) Results in a collision Each station monitors cable while transmitting If detect collision: host stops, waits, retries Uses binary exponential backoff policy Sender delays random time; doubles on second collision; quadruples on third collision; and so on If network busy, retransmit attempts quickly spread over a reasonably long period Wireless Ethernet Several wireless standards related to Ethernet 802.11b (Wi-Fi) Up to 11 Mbps; usually 2.5 – 4 Mbps 802.11a and 802.11g Up to 54 Mbps 802.16 (Wi-Max) Above three for point-to-point or with base 802.16 is for point-to-point only 802.11n 540 Mbps 802.11i Standard for security Ethernet addressing and frames Each computer has a 48-bit address Checked in interface hardware, not computer’s CPU When sending, specifies destination: Single, broadcast, or multicast Addresses associated with the interface hardware Move/change interface; new machine physical address Transmitted data viewed as a frame Variable length; 64 octets to 1518 octets Besides data, contains: Preamble, destination and source addresses Frame type Cyclic Redundancy Check (CRC) Ethernet bridges Bridge Connects two Ethernets; passes frames Operates on packets vs signals Bridge follows CSMA/CD rules, so (almost) arbitrary number can be added Ethernet switch is like a multi-port bridge Adaptive or learning bridges Can decide which frames to forward Usually very sophisticated and robust Important point: Bridges hide the interconnection details Set of bridged segments acts like single Ethernet Asynchronous Transfer Mode ATM is connection-oriented Designed for extremely high speed data switching Operate at gigabit speeds Needs complex, expensive hardware One or more high speed switches Optical fiber for all connections Uses fixed-size frames called cells Connection-oriented networking Not like packet-switched networks Must first establish a connection to the destination computer ATM switch finds path from sender to receiver Waits on remote computer to accept the request Local ATM switch selects identifier for the connection and passes it to the computer Computer sends using identifier When done, connection must be broken Wide Area Point-to-Point Networks WANs formed by leasing data circuits Digital circuits initially used for digitized voice Data came later, so data rates are not powers of 10 Are powers of 64 Kbps due to PCM 8000 samples/sec; each sample 8 bits T1: 1.544 Mpbs; T2: 6.312; T3: 44.736; T4: 274.760 Lower data rates use copper; higher data rates need fiber circuits Point-to-point “network” When system connects exactly two computers “Network” is a stretch; viewed as such for consistency Main point: do not need hardware addresses Dialup IP Another example of a point-to-point network Typically from modem (residence) to ISP TCP/IP view: Placing call is like running a wire Connection is made; stays as long as needed Summary Reviewed several network hardware technologies General idea: TCP/IP protocols are extremely flexible Almost any underlying technology can be used to transfer TCP/IP traffic