ATM* ATM Address Mapping ATM (Asynchronous Transf er Mode) is t he - PDF document

ATM* ATM Address Mapping � ATM (Asynchronous Transf er Mode) is t he � Rout er int erf ace (t o ATM link) has two addresses : I P and ATM address. swit ching and t ransport t echnology of t he B-I SDN (Broadband I SDN) archit ect ure (1980) � To rout e an I P packet t hrough t he ATM net work, � Goals: high speed access t o business and t he I P node: resident ial users (155Mbps t o 622 Mbps); (a) inspect s own rout ing t ables t o f ind next I P router address integrated services support (voice, dat a, video, (b) t hen, using ATM ARP table , f inds ATM addr of next rout er image) (c) passes packet (wit h ATM address) t o ATM layer � At t his point , t he ATM layer t akes over: (1) it det ermines t he interf ace and VC on which t o send out t he packet (2) if no VC exist s (t o t hat ATM addr) a SVC is set up * Kurose and Ross, “Computer Networking” ATM VCs ATM Physical Layer � Focus on bandwidth allocation f acilit ies (in � Two Physical sublayers : cont rast t o I P best ef f ort ) � ATM main role t oday: “ switched ” link layer f or I P- � (a) Physical Medium Dependent (PMD) sublayer over- ATM o (a.1) SONET/ SDH : t ransmission f rame st ruct ure (like a � ATM is a virtual circuit t ransport : cells (53 byt es) cont ainer carrying bit s); • bit synchronizat ion; are carried on VCs • bandwidt h part it ions (TDM); � in I P over ATM: P ermanent VCs (P VCs) bet ween I P • several speeds: OC1 = 51.84 Mbps; OC3 = 155.52 Mbps; rout ers; OC12 = 622.08 Mbps o (a.2) TI / T3 : t ransmission f rame st ruct ure (old � scalabilit y problem: N(N-1) VCs bet ween all I P t elephone hierarchy): 1.5 Mbps/ 45 Mbps rout er pairs o (a.3) unstructured : j ust cells (busy/ idle) ATM VCs ATM Physical Layer (more) � Switched VCs (SVCs) used f or short lived � Second physical sublayer connect ions � Pros of ATM VC approach: (b) Transmission Convergence Sublayer (TCS) : it o Can guarant ee QoS perf ormance t o a connect ion mapped adapt s P MD sublayer t o ATM t ransport layer t o a VC (bandwidt h, delay, delay j it t er) � Cons of ATM VC approach: � TCS Funct ions: o I nef f icient support of datagram t raf f ic; P VC solut ion o Header checksum generat ion: 8 bit s CRC; it prot ect s a 4- (one P VC bet ween each host pair) does not scale ; byt e header; can correct all single errors. o SVC int roduces excessive latency on short lived o Cell delineation connect ions o Wit h “unst ruct ured” P MD sublayer, t ransmission of idle o High SVC processing Overhead cells when no dat a cells are available in t he t ransmit queue

ATM Layer ATM Adapt ion Layer (AAL) [more] � ATM layer in charge of t ransport ing cells across � Dif f erent versions of AAL layers, depending on t he ATM net work t he service t o be support ed by t he ATM t ransport : � ATM layer prot ocol def ines ATM cell header o AAL1 : f or CBR (Const ant Bit Rat e) services such as f ormat (5byt es); circuit emulat ion � payload = 48 byt es; t ot al cell lengt h = 53 byt es o AAL2 : f or VBR (Variable Bit Rat e) services such as MP EG video o AAL5 : f or dat a (eg, I P dat agrams) ATM Layer ATM Adapt ion Layer (AAL) [more] � VCI (virt ual channel I D): t ranslat ed f rom link t o � Two sublayers in AAL: link; o ( Common P art) Convergence Sublayer (cpcs): � PT (P ayload t ype): indicat es t he t ype of payload encapsulat es I P payload (eg mngt cell) � CLP ( Cell Loss P riorit y) bit : CLP = 1 implies t hat o Segmentation/ Reassembly Sublayer (sar): t he cell is low priorit y cell, can be discarded if segment s/ reassembles t he CP CS (of t en quit e rout er is congest ed large, up t o 65K byt es) int o 48 byt e ATM segment s � HEC (Header Error Checksum ) byt e ATM Adapt at ion Layer (AAL) AAL5 - Simple And Ef f icient AL (SEAL) � ATM Adaptation Layer (AAL): “adapt s” t he ATM � AAL5 : low overhead AAL used t o carry I P layer t o t he upper layers (I P or nat ive ATM dat agrams applicat ions) o SAR header and t railer eliminat ed; CRC (4 byt es) moved t o CP CS � AAL is present only in end systems , not in o P AD ensures payload mult iple of 48byt es (LENGTH = P AD swit ches byt es) � The AAL layer has it s header/ trailer f ields , o At dest inat ion, cells are reassembled based on VCI carried in t he ATM cell number ; AAL indicate bit delineat es t he CP CS-P DU; if CRC f ails, P DU is dropped, else, passed t o Convergence Sublayer and t hen I P

Dat agram J ourney in I P-over-ATM ARP in ATM Net s (mor e) Net work � At Source Host : � (2) ARP Server: o (1) I P layer f inds t he mapping bet ween I P and ATM exit address (using ARP ); t hen, passes t he dat agram t o AAL5 o (2.a) source I P router f orwards ARP request t o server on dedicat ed VC (Not e: all such VCs f rom rout ers t o ARP o (2) AAL5 encapsulates datg and it segments t o cells; have same I D) t hen, down t o ATM o (2.b) ARP server responds t o source rout er wit h � I n the network , t he ATM layer moves cells f rom I P / ATM t ranslat ion swit ch t o swit ch, along a pre- established VC � Host s must register t hemselves wit h t he ARP server � At Destination Host , AAL5 reassembles cells int o original datg; Comments : more scaleable t han ABR Broadcast o if CRC OK, dat gramis passed up t he I P prot ocol. approach (no broadcast st orm). However, it requires an ARP server , which may be swamped wit h request s ARP in ATM Net s X.25 and Frame Relay � ATM can rout e cells only if it has t he ATM � Wide Area Net work t echnologies (like ATM); also, address bot h Virtual Circuit orient ed , like ATM o Thus, I P must translate exit I P address t o ATM address � X. 25 was born in mid ‘70s, wit h t he support of � The I P / ATM addr t ranslat ion is done by ARP the Telecom Carriers , in response t o t he (Addr Recogn P rot ocol) ARP ANET dat agram t echnology (religious war..) � Generally, ATM ARP � Frame relay emerged f rom I SDN t echnology (in t able does not st ore all ATM addresses : it must discover some of t hem lat e ‘80s) � Bot h X.25 and Frame Relay can be used t o carry I P datagrams ; t hus, t hey are viewed as Link � Two techniques : Layers by t he I P prot ocol layer (and are t hus o broadcast covered in t his chapt er) o ARP servers ARP in ATM Net s (mor e) X.25 � (1) Broadcast t he ARP request t o all destinations : � X.25 builds a VC bet ween source and dest inat ion f or each user connection � Along t he pat h, error control (wit h o (1.a) t he ARP Request msg is broadcast t o all ATM dest inat ions using a special broadcast VC; ret ransmissions) on each hop using LAP-B, a variant of t he HDLC prot ocol o (1.b) t he ATM dest inat ion which can mat ch t he I P � Also, on each VC, hop by hop f low control using address ret urns (via unicast VC) t he I P / ATM address credits ; map; o congest ion arising at an int ermediat e node propagat es t o source via backpressure � Broadcast overhead prohibit ive f or large ATM net s.

X.25 Fr ame Relay (mor e) � As a result , packet s are delivered reliably and in � Frame Relay implement s most ly permanent VCs sequence t o dest inat ion; per f low credit cont rol ( aggregat e f lows ) guarant ees f air sharing � P ut t ing “ intelligence into the network ” made � 10 bit VC I D f ield in t he Frame header sense in mid 70s (dumb t erminals wit hout TCP ) � Today, TCP and pract ically error f ree f ibers f avor � I f I P runs on t op of FR, t he VC I D corresponding pushing t he “ intelligence to the edges ”; moreover, t o dest inat ion I P address is looked up in t he local gigabit rout ers cannot af f ord t he X.25 processing VC table overhead � As a result , X.25 is rapidly becoming extinct � FR swit ch simply discards f rames wit h bad CRC (TCP ret ransmit s..) Frame Relay Frame Relay -VC Rat e Cont r ol � Designed in lat e ‘80s and widely deployed in t he � CI R = Commit t ed I nf ormat ion Rat e, def ined f or ‘90s each VC and negot iat ed at VC set up t ime; cust omer pays based on CI R � FR VCs have no error control � Flow (rate) control is end to end ; much less processing O/ H t han hop by hop credit based f low � DE bit = Discard Eligibilit y bit in Frame header cont rol o DE bit = 0: high priority , rat e compliant f rame; t he net work will t ry t o deliver it at “all cost s” o DE bit = 1: low priority , “marked” f rame; t he net work discards it when a link becomes congest ed (ie, t hreshold exceeded) Fr ame Relay (mor e) Frame Relay - CI R & Frame Marking � Designed t o interconnect corporat e cust omer � Access Rate : rat e R of t he access link bet ween LANs source router (cust omer) and edge FR switch (provider); 64Kbps < R < 1,544Kbps � Each VC is like a “ pipe ” carrying aggregat e t raf f ic bet ween t wo rout ers � Corporat e cust omer leases FR service f rom a � Typically, many VCs (one per dest inat ion rout er) public Frame Relay net work (eg, Sprint or ATT) mult iplexed on t he same access t runk; each VC has own CI R � Alt ernat ive, large cust omer may build Private Frame Relay net work. � Edge FR swit ch measures t raf f ic rat e f or each VC; it marks � (ie DE < = 1) f rames which exceed CI R (t hese may be lat er dropped)