COMPUTER NETWORKS CS 571
Popular in Course
Popular in ComputerScienence
This 22 page Class Notes was uploaded by Juvenal Beahan on Friday October 23, 2015. The Class Notes belongs to CS 571 at University of Kentucky taught by Zongming Fei in Fall. Since its upload, it has received 5 views. For similar materials see /class/228215/cs-571-university-of-kentucky in ComputerScienence at University of Kentucky.
Reviews for COMPUTER NETWORKS
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 10/23/15
FTP RFC 959 I Generalized le transfer 7 Allow transfer of arbitrary files 7 Accommodate different rue types 7 Convert between heterogeneous systems Data types Word len s Rules for file names 7 User 1ogin Interface I Protocol actions include 7 List contents of directory 7 change to adifferent working directory 7 Retrieve afile 7 Put a file File transfer Protocol FTP I TCPIP standard is File Transfer Protocol FTP I General purpose protocol 7 Operating system and hardware independent 7 Transfers arbitrary files 7 Accommodates file ownership and access restrictions I Predates TCPIP adapted to TCPIP later ftp client commands I Twostep process 7 Launch ftp 7 Connect to remote host Connect my oiyes using user account on remote host ome FTP servers provide anonymous FTP I rp client interface from BSD UNIX is de icto standard 7 Many commands archaic and no longer used tenex carriage control 7 Most often used cd dir 1s get put 7 Otheruseful crpwd1cd Twoway file transfer I get from FTP server to local host put to FTP server from local host Default uses same name on both hosts p client allows speci cation of different names mget mput transfer multiple les 7 UNIXlike wildcard expansion File name translation I File name syntaxes may be incompatible I UNIX 128 character mixed case DOS 83 character upper case I Some names may not be legal in all systems I BSD p allows rules for lename translation File types and transfer modes Many different styles of le typing UNIX untyped may hold anything 7 Macos strongly typed ftp does two types oftransfer 7 Text with appropriate translations to maintain integrity 7 Binary no translation whatsoever FT P mess ages FTP messages from clients to servers are in the farm of cammand emarne m PORT n1n2 3n4n n Each message from server includes a threedigit decimal number 7 226 Transfer complete 7 221 Goodbye Convenient for computer and human recognition Verbuse mode shows messages qmet mode suppresses mess es FTP clientserver model Remote server accepts control connectzon from local clzent 7 Client sends commands to server 7 persists through entrre session Server creates data connection for data transfer 7 one data connectron for each transferred tile 7 Data transferred either way FTP clientserver model HP client In lmlIM HIHI Port numbers at the server mmm 7 Control 21 7 Data 20 Port numbers at the client WM MW 7 1 ornrnand connectto the port number speci ed otherWise use the same port number for the control connectron TFTP Using separate data connections Separates commands from data Client can send commands during data transfer Closed connection indicates end of file slmpler 7 Based on UDP 7 File transfer only no directory listing 7 N0 authorization Can be used in UDPonly system Requires less code than FTP Often used for bootstrap eg ROMbased diskless system 39 TrivialFile TransferProtocol TFTP much than FTP IPv6 435 Current version ofIP version 4 is 20years old IPv4 has slioWn remarkable ability to move to new technologies IETF has proposed entirely new version to address some specific problems Success ofIP IP has accornrnodated dramatic changes since original desi 7 Basic pnnciples sull appropriate today Many new types ofnardware fr wtens to a few tens ofrnillions ofcornputers 7 s eed 7 from 56Kbps to lcbps 7 Increased frame size in hardware Motivation for change Address space 7 32 Int address space allows for over a mllllon networks But most are Class c and too small for many organizations 7 2M Class B network addresses already alrnost exhausted and exhausuon was firstpredictedto occur a couple ofyears ago Type ofservice rent applicauons have different requirements for delivery reliability and speed 7 Current 11gt has type ofservice that39s not o en irnplernented Name and versions number Preliminary versions called IP 7 Next Generatlon IPng Several proposals all called IPng one was selected and uses next available version numba 6 Result is IP vemon 6 IPv6 New features Address size IPv6 addresses are 128bits Header format entirely differsn E a l 5quot m a ers Additional information stored in optional exteasion headers followed by data video ow labels and o applications to establish appropriate connections Exteasible new features can be added more sily IPv6 datagram format Minimal Illllliiiil llllll aaso H ndm IPV6 base header format IPv6 base header format cont Contains less information than IPv4 header NEXT E DER points to rst extension header FLOW LA BEL used to associate datagrams belonging to a flow or communication between two applications T ra ic clam Speci c path Routers use FLOW LABEL to forward datagrams along prearranged path saunas Anonzss DEsTmn39Lcm ADDRBS IPv4 datagram header format IPV6 NEXT HEADER a 4 3 VERS SERVICE TVPE TOTAL LENGTH mamlncnnon FLAGS FRAGMENT OFFSET my um nus TO LIVE TYPE HEADER cnacxsuu quotEXETCP sounc IP ADDRESS m UESTINATIDN IF ADDRESS VF OF39HDNS MAY as OMITTED FADDING ease Header Rama Hauler m Dquot BEGIN MEX aouT NEXT up MING OF DAIA W Fragmentation Parsmg I Pv6 headers 39 Base head is xed Size 39 40 Octet I Fragmentation information kept in separate 7 NEXT HEADER heldrn hase header de nes type ofheader extension header 7 Extension headers appear at end of xedrslze hase header I Each fragment has base header and inserted I Some extensions headers are variable sized 7 NEXT HEADER heldrn extension header de nes type fragmentauon hea er a HEADER LEN held gwes srze of extension header I Entire datagram including original header may be fragmented ONE 09 MORE DP IONS Fragmentation Tinggmfinri pin Untrlnmcnmbll Fmg i vim Nlnlilr m umgmannm Frag 2 pin Handr M Uiilrnqmnmabla mu 1 F m Humv J in Fragmentation and path MTU IPv6 rource riot intermediate routers responsible for fragmentation r Routers simply drop datagrams larger than network MTU 7 Source must fragment datagram to reach destination Source determines path MT U r Smallest MTU on any network between source and destination 7 Fragments datagram to fit within that MTU 39 Uses path MTUdircovery 7 Source sends probe message ofvarious sizes until destination reached 7 Must be dynamic path may change during 39 rarns transmission ofdatag Use of multiple headers Ef ciency header only as large as necessary Flexibility can add new headers for new features Incremental development can add processing for new features to testbed other routers will s 39p those headers IPv6 addressing 128bit addresses Includes network prefix and host suf x No address classes pre xsuf x boundary can fall anywhere Special types ofaddresses 7 mm Single destination computer 7 minim multiple destinations possibly not at same site 7 cluster collection of computers with same prefix datagam is delivered to one out of c uster IPv4 broadcast avors are subsets of multicast Cluster addressing allows for duplication ofservices IPv6 address notation 128bit addresses unwieldy in dotted decimal requires 16 um ers r 105 220136100255 255 255 255 0 01812814010255 255 Groups of 16bit numbers in hex separated by colons culuri hmadeczmal or colon hex r 69DC 8864 Zerocompression series ofzeroes indicated by two colons r FFOC Bl PVG address with 96 leading zeros is interpreted to hold an IPv4 address 7 IPv4compatible IPv6 address 128 96 33 81 r EPv4mapped IPv6 address FFFF128 96 33 81 Summary IPv4 basic abstractions have been very successful IPv6 carries forward many of those abstraction but all the details are changed 7 128bit addresses 7 Base and extension headers 7 Source does fragmentation 7 New types of addresses 7 Address notation EndtoEnd Arguments in System Design Example Flle transfer Distinguish communication subsystems and end teps applications 7 Read from disk ment appeals to application requirements and 7 Program ask communication system to transmit provides a rationale for moving function upwardin a 7 Move data in the network layered system closer to the application that uses the 7 Transfer datato appheataon Ian 7 erte to e Funetions plaeed atlow levels ofa system may be redundant or of I Reinforce each of the steps using little value when eompared Wth the eost ofprovldlng them at that Duphcm to M low level a Timeout andretry a The runetion in question ean eompletely and eoneetly be 7 Redundant for em daemon implemented only Wth the knowledge and help ofthe applieation Y standing at the end points ofthe eommunieation system 7 Crash recovery Endtoend check and retry cs 5 c 571 l 71 Pelfomlance Cost at lower level I It is too simplistic to conclude that lower levels I Performing functions at a lower level may cost should play no part in obtaining reliability more 7 Example dropping one message of each hundred 7 since the lower level subsystem is common to many messages sent applications those applications that do not needthe e The probability that all packets ofa le arrive correctly function will pay for it anyway decrease exponentially with the le length 7 The lower level subsystem may not have as mu Key idea lower levels need not Provide Perfect information as the higher levels so it cannot do thejob 39 l reliability do no I deeoff based on performance c 571 a c 571 Other examples Identifying the ends I Guaranteed delivery I Voice communicationas anexample a E d d k h b h k 3 quot 5 P Y W rEndhostto endhost 7 People to people I Secure transmission of data keys Clear when ltpasses into appheation a uthentlclty ofmessage stlll needto be checked by the applieation Duplicate message suppression FIFO message delivery Transaction management I Different levels in the protocol architecture Other areas 7 Backup of disk using tapes 7 RISC c 571 s c 571 Bridges and Extended LANs LANs have physical limitations eg 2500m Connect two or more LANs with a bridge 7 accept and forward strategy 7 level 2 connection does not add packetheader Bridges and Extended LANs 32 Learning Bridges An example Do not forward when unnecessary Maintain forwarding table lmmr H mm 4 mimlrr n v w v v 7 Even Segment 1 List Segmentz Llsl Brmge boots v Learn table entries based on source address v Table is an optimization need not be complete Y Always forward broadcast frames V sends to X U v 2 v Xsendstow Uiv Z V X Wundslol UV w Z Y X 3 Spanning Tree Algorithm Algorithm Overview Problem loops Each br1dge has un1que 1d e g B1 B2 B3 a total order is defined among ids Select bridgewith smallest id as root Select bridge on each LAN closest to root as designated bridge use id to break ties Each bridge forwards frames E over each LAN for which it Br1dges run a d1str1buted spanmng tree algor1thm is the designated bridge and 7 select which bridges actively forward the LAN towards the root 7 developed by Radra Perlman 7 now IEEE 802 1 speci cation 5 Algorithm Details e id for whatthe sending bndge bellevesto be rootbndge y e distance hops from sending bridgeto root bridge d e id forbndge sendingthe mes g Bridges exchange configuration messages v d Each bridge records current best configuration message for each port 7 Goodness ofconfigumtion messages in this order Ruut d r distaneeturuutd smderldX Initially each bridge believes it is the root myid 0 myid Algorithm Details cont Ifreceive abetter configuration than myid 0 myid a bridge leams it should not be the root Determine whether the bridge is a designated bridge for a LAN 7 choose the best configuration among myid o myld and configurations reco ed or eachport Assume itis R D s Add 1 to D ifthe best configuration is NOT myid o myld e For each port o er an the one with the best configuration compare its best configuration 1 d X Wth R D myld IfR D myid is better thanthe muteris the designatedbndge fur the LAN at thepurt Thebridge shuuld furwardmessage R D myid tr thatpurt lf Y dx is better itis net the desi hatedbndgefurthatLAN Du nut furwar message R D myid tr atpurt Algorithm Detail cont en learn not root stop generating config messages e in steady state unlymut generates cun guratlun messages ated ndge stop forwarding config messages e in steady state unly designatedhndges furw d er messages Root continues to perio ically send con Results b fig messages Fo a ndge only use the ports e The unemththehesteunligurataun e Thuse parts eunneeted With LAN ufwhieh this hndge is the designated ndge lfany bridge doesnotreceiye config message a er a period oftime it starts genemtmg config messages claiming to be L1m1tatlons of Brldges I 0 not scale 7 spanning tre algonthm does not scale 7 broadcast does not scale Do not accommodate heterogeneity Caution beware oftranspar nc e Applicati or transport Erotocols may make the assumption that it is connected to another ostwtth a single LAN and not aspect that the frames may get dropped at bridge exp erience long latency Di erences between bridges and routers Broadcast and Multicast I Forward all broadcastmulticast frames 7 current practice I Learn when no group members downstream I Accomplished by having each member of group G send a frame to bridge multicast address with G in source eld IP Internet Concatenation of Networks Internetworkin g M Wequot Outline Best Effort Service Model Global Addressing Scheme c 571 l c 571 2 S ervice Model I Connectionless dztagrambased I Besteffort delivery unreliable service 7 packets are as rdmteucra 7 packets are delivered out of order 7 duplicate copies ofa packet are delivered WE TO LIVE W E 7 packets can be delayed for a long time IP datagram header format N P in umzm EGINMNG or mu c 571 3 c 571 a IP datagram header elds VERS version of currently 4 H LEN header high in units of32 bits Several options can be addedto P header SERVICE TYPE sender39s preference for low latency high 7 Record route reliability rarely used 7 Sourceroute IP datagram options TOTAL LENGTH total octets in datagram e Timestamp 1D NT FL GS GMENT OFFSET used with fragmentation ad with no options has H LEN eld value 5 data TTL zzme to lzve decremented in each router datagram discarded begins immediately after DESTJNATION IP ADDRESS when m 0 ticus added between DESTJNATION IP ADDRESS TYPE type ofprotocol carriedin datagxam e g TCP UDP and data in mump1 0 bits HEADER CHECK s complement of 1 s complement sum Header with 96 bits of options has H LEN eldvalue 8 SOURCE DESTIP ADDRESS IP addresses oforzgmal source and ulzzmaze destination c 571 s c 571 5 Global Addresses Properti ES 7 globally unique 7 hierarchical network host Dot Notation 7 10 3 2 4 7 128 96 33 SI 7 192 12 69 77 a A a 4 a a New Hrs 21 r c cs 57 IP Addressing One key aspect of virtual network is single uniform address format Can39t use hardware addresses because dltfereat technologies have diffaalt address format Address format must be independmt of any particular hardware address format Seadlrlg host puts desu39rratlor intanet address in packet Destination address car be inta preted by arry inta mediate route Routequot examine address and forward packet on to the destlhatlor C557 2 IP address hierarchy Each IP address is divided into a pre x and a suf x 7 Pre x identifies network to which cornputa is attached network num er 7 s rx rdeau39 es cornputaquot wlthlr that network host umbaquot or ho st zddre ss Address format makes routing ef cient cs 57 Designing the format of IP addresses 1P designers chose 32bit addresses Allocate some bits for pre x some for suf x 7 Large pre x srrrall suf x many networks few hosts per networ e srrrall pre x large suf x few networks many hosts per network Because ofVaIiety oftechnologies need to allow for both large and small networks C5571 lo Classes of addresses cont ms ul3n c a class A tilmm V class a lamamm class c Ila ml Class D Class E cs 57 mulhcasl address veselven my tulure use Dotted decimal notation 32 I Binary Number EgulvalenlDotled Decimal ULUUD w Utlll lll l L L39 ll D ll OCHKHH llll quotWW 5 H mm39 m Glll l39ll39llquot lll HWUDDJ Mu 39WDHWQ HJOUHDOU l C5571 l2 Address Classes at a glance Networks and hosts in each class While dotted decimal makes separating network address from host address easier determining class is not so obvious Look at first dotted decimal number and use this table Address Bilsln MaximumNumbev Eitsln MaximumNumhevOl Class Pre x otNelworks Sultix HostePerNelwotk A 7 128 24 16777216 Class ange or Values v o Waugh 2 a IA 16166 16 65536 a 128 through 191 C 21 2007152 B 256 C 92 through 223 D 224 through 239 E 2 10 through 255 Lb 13 CSS71 14 Example Spec1al IP addresses Tm 39I39l m39z b Prefix Suf x TypeOtAddress Purpose 1 g alto alto lhisnompuler used during bootstrap 128100 12amoz 1282116115 12821123J network allos network idenlillesanetwork network ale directed broadcast broadcast on specified net allls all16 Iimlledbmdcasl broadcaston local net WW Wimzm 127 any loopback tesllng 39 r 1 4 l39 l D 1116 m mm mm 1925er CSS71 16 Routers and IP addressing IP address depends on network address What about routers connected to two networks IP address speci es an interface or network attachment point not a computer Router has multiple IP addresses one for each interface 08571 17 Routers and IP addressing Ehemet 131 10800 1 1311aems 223246129 223 240 129 17 k 5 mmrr 73 u e 11 WAN 7810 08571 18 Multi homed hos s Hosts that do not forward packets can also be connected to multiple networks Can increase reliability and performance Multihomed hosts also have one address for each interface Should not be used as a router as 571 19 Packets Packets serve same purpose in intemet as frames on LAN Each has aheader Routers fonnerly gateways forward between physical networks Packets have aunifonn hardereindependent format 7 lnclud s header and data 7 Can t use format from any partarular hardware Encapsulated in hardware frames for delivery across each physical network as at 20 What about reliability Routing table Reltable tteltvery provided by transport layer Network layer IP can detect and repurt errors without actually fmrtg thern Ha Ha H Natl rm f to 707 u a Network layer focuses on datagram delivery W 7 Application layer not interested in differentiating among delivery problems at intermediate routers Dam quotm quotup ne a re 2 and am quota r rum um um a as 571 2r 4 Default routes Routing table kept small by listing destination networks rather than hosts Can be further reduced through default route 7 Entry used ifdestination network not explicitly listed in routing table a E g UK carnpus edge routeruses default routes for all otrcarnpus networks as 571 2 Forwarding destination address and next hop Destination address in P datagmm is always ultimate destination Router looks up nexthop address and forwards datagram Network interface layer takes two pammeters r P datagram a Nexthop address Nexthop address never appears in P datagram Datagram Forwardlng I Strategy 7 every datagram contains destination s address 7 ifdirectly connected to destination network then forward to host 7 ifnot directly connected to destination network then forward to some router 7 forwarding table maps network number into next hop 7 each host has a default router 7 each router maintains aformrding table cs 57 25 Maximum Transmission Unit I Every hardware technology speci cation includes the de nition of the maximum size 0 the fmme I Called the maximum transmission unit MTU I Any datagram encapsulated in a hardware frame must be smaller than the MTU for that hardware cs 57 26 MTU and datagram transmission I IP datagrams can be larger than most hardware MTU s 7 IP 2 6 l 7 Ethernet 1500 7 Token ring 2048 or 4096 I Source can simply limit IP datagram size to be smaller than local MTU cs 57 27 Fragmentation and Reassembly I Each network has some MTU Strategy 7 fragment when necessary MTU lt Datagram 7 try to avoldfragmentatlon at source host 7 do not recover from lost fragments cs 57 22 Example lHHl cs 57 29 Fragmentation cont Each fragment is an independent datagram 7 Includes all headcrficlds 7 Bitinhcadcrindicatcs datagram is afr ment 7 Other fields have information for reconstructing original datagram 7 FRAGME or SE gives original location offragmcnt Router uses local MTU to compute size ofeach fragment Puts part of data from original datagram in each fragment Puts other information into header cs 571 an Fragmentation cont mar2 nen I m amt IFHdH new t as m n Flags Fragnlent Offset Flags 3 bits 7 Front reseryearnuettieo 7 seeena tilt DF Denvtrragnent 1dm39t tragnent onny gnent 7 mm at ME More Fragment tnnere rragnente 07m tragrnem otfset 13 bits indicates where this 39agment is located In the original I datagram measured In the unit of8 octets 7 The aret tragnent nae erreet zero The total length is the header lengthsegment length 7 The total length rm packetslt Mm not the length of data in m packet cs 57 a Data gram reassembly Reconstruction of original datagram is called reassembly U timate destination perfome reassembly Fragme s may ye out of order header bit identi es fragment containing end of data from original datagram E g Fragment 3 identified as last fragment as m as Fragment loss IP may drop fragment w at happens to original dalagram 7 Destination drops ennre original datagram How does destination identify lost fragment 7 Sets timer wtth eaeh fragment 7 Lftimer expires before all 39agments arrive fragment assumed lost 7 Datagram dropped Source higher layer protocol assumed to retransmit cs 57 24 Fragmenting a fragment Fragment may encounter subsequent network with even smaller MTU Router fragments the fragment to t Resulting subfragments look just like original fragments except for size No need to reassemble hierarchically subfragments include position in orzgznal agram as m 25 Ethernet 26 I Adapted from PetersonampDavie Book and other universities Ethernet Overview History 7 developed by Xerox PARC in midrl9705 e roots in Alohapacketrradlo network 7 standardized by Xerox DEC and Intel in 1978 7 similar to IEEE 802 3 standard 7 mu eaccess r collision detection I Frame Format Desi Sn auur auur Typ e Ethernet cont Addresses 7 unique 48bit unicast address assigned to each adapter 7 example80e4b12 r broadcast all ls r multicast Bandwidth lOMbps lOOMbps leps Length 2500m 500m segmenw with 4 repeaters Problem Distributed algorithm that provides fair access Multiple Access Communications 7 Broadcast networks I Also referred to as multiple access networks I All information is received by allusers routing is not necessary A at addressing scheme is sufficient to indicate which user a given packet is destined Medium access control protocol is required to orchestrate the transmissions from the various users Main concern is interference fonn otherusers Transmit Algorithm I lfline is idle r sendimmediatel 7 upper bound message size of1500 bytes 7 must wait 9 6us between backtoback frames I lfline is busy 7 wait until idle and transmit immediately 7 called liperszstem special case ofpr erszstzm Algorithm cont If collision 7 Jam for 32 bits then stop transmitting frame 7 minimum frame is 64 bytes header 46 bytes of data 2ndtime 05121024or153 6us 3rd time o5121o24153 6 204 8 256 o 307 2 or358 4us rim time kx 51 2us for randomly selected 160 2w 7 1where mmmlOn give up after several tries usually 16 exponential backoff CmAmpmldm Sxmplx ed lavchart fax kamxmng ame thnv httfhrJrnrt tuttown gtuh mu W t Inter packet Gap Time LAN Standard IEEE 8023 The anz Preqeet Mndel In addmun xtxsnut suf erendy fur anude tn wartundl ed furmeyear and mumhnbegm am February Wu 5 u net rk q net and thenbegmtu transrnrt mmedmely mmustbe me me be as 7 IEEE anz standards aetually predated the ISO standards packets Emma netwmk 5E mewmg mass can 7 spht the data lmk layer mtu twu drfferent sublayers drshngursh that the nnurnaekethas ended and Lugxcal Lmk Cuntml LLC and Medra Access renare te reeerve a new naeket The standard Cuntml MAC requrres th anude wart atleastthe de nedmter MD at My 52 52333 mmes W Standarnze these that em be generened e LLC is the upperlayerufthe IEEE anz datahnk ayer and 15 eernhnnnte all N prutuenls xsnlatedthnse that mustrenarn speci c 7 AC prutueuls are speer e tn the LAN usmg thern y EthemeL Tukm ng etc 1 osx Mudel and Fruject anz nemmam Lmnndel Other laych myquot mm M um hum V thl v m m MN quotm V qual laquot n n wnhm nutn nun mtnmu Project 802 OS Model 2 Ln slumlnnls The IEEEVXEIZ 3 farndy efspeet eatmns LAN Standard 7 IEEE anz 3 EEEEDZCnegmes mm x mm x mlMM H ttmhtst H mm t LAN Standard 7 Ethernet name forums foxEEErIIZ 3 1rd mx attemtstsnaards mc trtttn ut tsm cmmdmmvhwpulw M dmew dm enamwmhmmnmmnmwm IixmmmL WQquhwW gnmvh ant Mgt2 mm a cat pmlmm 4 Mlmmum Frarne Length vs Maxtrnurn Netwerk Length The angnal an a maxxmum dstanee was dengnedtn aperate ntlEIMb Mlmmum Frarne Length ps hrzsnn meters When and wanees are made rd cenari mamas Maugham mum mm Befure the sender ean sendthe entare ame nut the pzapagmandelayaf 2 mxcmsecands Attthpshtspmprgndnn stgna1 travels threugn thenetwurk and reaehes the detny enanstz bxtsa164bytes whchwas seteeted as39he rmmmum and Eng Wm W szimmm mum hm wmmmm xftherets anether stgnal atthe end ufthe netwnrk The rersnntstn a ww a sendng staunntnhe able tn detect a 5 we uu s m 5 pasnhIe ca mmsbefme the sewing stanancarnpletedthe The Sen needs the Upp m mty in Shun 39113 sendinng fume smdmg quhe ame and send 212m sequence te Thehngest e nrttetresmrsennngstmntnnndm mfunn ten taca xsmaccuxedxs39hemmdmp pxwpagmanumela thernthestneenng mun Themund mp ttrn plus required te send thejam sequence shuuld he less than the tarne Caculahmufmundm a a unnume t Dmmcjgjfmmmgmmmm needed furthe senderte sendtherntntrnurn harne shee ureteemertyrnnetersnerseednd 512 bits The sender needs tn be aware ufthe culllsmn hefure ttts the late thatts hefure lthas sent the enttr rrarne 8 transmlt a ame shuuld he greaterthan ZD Thls lmplles that mlmmum tameneeded td txansmlt a ame shuuld he ZD Nuts mlmmum tame needed td txansmlt a ame mlmmum ame length m hrts taansmrt speed m hrtsseednd Mlmmum Frame Length and the Maxlmum Netwnrk Length Maximum netwnrklength leen a mlmmum ame length the maxlmum length at an Ethemet netwurk ean he ealenlated hum mlmmum ame length and the txansmlt speed Example a gyen mlmmum frame length the faster the h39znsmlt speed the shdrter the manmum Ethernet netwurk length A History of Speeds ofEthemet Introduction Outline Statistical Multiplexing InterProcess Communication Network Architecture Performance Metrics Building Blocks I Nodes PC specialpurpose hardware os s r switches I Links coax cable optical ber 7 pointtopoint III CI 7 multiple access M Switched Networks I A network can be de ned recursively as 7 two more networks connected by two or more nodes 7 two 0 mo nodes connected by a link or U 55 r Strategies I Circuit switching carry bit streams a original telephone network I Packet switching storeandforward messages 7 Intem et Addressing and Routing I Address bytestring that identi es a node a usually unique I Routing Vs Forwarding I Types ofaddresses a unicast nodespeci c r broadcast all nodes on the network 7 multicast some subset ofnodes on the network Multiplexing I TimeDivision Multiplexing T DM I FrequencyDivision Multiplexing FDM Switch 1 S 39th2 ch Statistical Multiplexing Ondemandtimedivision Schedule link on aperpacket basis Packets from different sources interleaved on link Buffer packets that are Cuntzndzng for the link Buffer queue over ow is called Cungesnun InterProcess Communication Tum hosttohost connectivity into processtoprocess at F tween what applications expect and what the underlying technology provides IPC Abstractions I RequestReply I StreamBased dismbmed le systems 7 video sequence offrarnes 7 digital libraries web 7 video applications onrdemand video Videoconferencing What Goes Wrong in the Network I Bitlevel errors electrical interference I Packetlevel errors congestion I Link and node failures I Messages are delayed I Messages are delivered outoforder I Third parties eavesdrop Layering I Use abstractions to hide complexity I Abstraction naturally lead to layering I Alternative abstractions at each layer Application programs Requestreply Message stream channel channel Hosttohost connec ivity H ardware Protocols I Building blocks of a network architecture I Each protocol object has two different interfaces 7 SenICE mter zce operations on this protocol 7 peerrmrpeer mter zce messages exchanged with peer I Term protocol is overloaded e specification ofpeertopeer interface 7 module that implements this interface Interfaces H051 Has 2 Protocol Machinery I Protocol Graph 7 mostpeertopeer communication is indirect e peertopeer is direct only at hardware level R aniai Ivvi pli ia Machinery cont I Multiplexing and Dernultiplexing demux key I Encapsulation headerbody Internet Architecture I De ned by Internet Engineering Task Force IETF I Hourglass Design I Application Vs Application Protocol FTP HTTP ISO Architecture Enaiialt Enaiialt iiiiiiin in Mint Performance Metr1cs Bandwidth 7 data transmitted per time unit 7 link versus endrtormd r notation KB 2 bytes Mbps105bitspersecond I Latency 7 time to send message from point A to point B r onerway Versus roundrtnp time RTT 7 components Latency Propa ation Transmit Queue Propagation Distance c Transmit Size Bandwidth Bandwidth versus Latency Delay X Bandwidth Product I Relative importance I Amount of data in ight or in the pipe 7 1byte 1ms vs 100ms dominates 1Mbps vs 100Mbps 7 25MB leps vs 100Mbps dominates 1ms vs looms 39 Example39 looms X 45Mbps 560KB I Throughput Daisy 7 RTT dominates Throughput TransferSiZe Tran5ferTime Bandwidth TransferTime RTT 1Bandwidth gtlt TransferSiZe 7 lMB le to lGbps link as 1KB packet to lMbps link
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'