‘ IP + 2 = Ipsilon IP plus a little bit more... lp|NtwkCfdtldPptry P91

. . 3» \ ’ “ ' Be the premier provider of solutions for high performance IP networks. Ipsilon Networks Conf|dent|aI and Proprietary Page 2

IP w ;<»»~ » = ~» o Phenomenal growth of IP lnternetworking 0 IP is the only choice for large networks 0 No single vendor dominance; strong vendor suppo 0 Users & vendors heavily invested in IP lpsilon Networks Condentlal and Propr|etary Page 3

Why is the IP user frustrated? o Router price/performance falling behind switches o ATM’s complexity far exceeds Ethernet’s o Switches don’t scale - no attention to routing o Networks not open for research innovation, end user customization, 3rd party enhancements lps|Ion Networks Conf|dent|a| and Proprietary Page 4

The Innovation Gap - - 4 \ “ 1 1 ‘ .1§J*’~‘,€W i""m'$*§§§“*% 0 More innovative ideas & skills for network evolution than ever 0 Researchers can’t afford to build their own large networks 0 Vendors, not owners or researchers, control the evolution of networks 0 Large networks evolve slowly anyway 0 Evolution ceases; revolution starts to look good lpsilon Networks Confidential and Proprietary Page 5

o Focus on IP to simplify network architecture 0 Use switch hardware for revolutionary pricel performance 0 Let users avoid complex, proprietary & immature ATM protocols & products 0 Leverage innovation in the broad research community |p$l|Ol'l Networks Conf|dent|al and Proprietary Page 6

What does the IP user need? - Y \ _ ~ »r@,,:"W»**, vZ§%;s§§§§<%¢/:§e1;?%:‘;‘”?%¢;A: ~ 2 ~~ Vxi ;\@w$>>»~ ‘:' ' 1/Y x-M . ;;\_\_ V V -<=-_ .5. . -. . 0 Routing 0 Price/Performance 0 Simplicity 0 Openness lps Ion Netwo ks Co fde tal a d Proprietary Page 7

IP . R t. a 0 Routing is not just for traversing LAN media types 0 Routers allow scaling: >> Address & topology management >> Administrative boundaries >> Security management >> etc. 0 Lower layers can aggravate, but not alleviate, routing problems lpsilon Networks Condential and Propr|etary Page 8

Price/Performance - .-: .. . :,,;;:,( ~ »».¢\¢~»~“=»»~»; -»,=;>.~-; - » \ 5//::'A'4 7 A _ 4% 4‘ V »~":i“‘&"?';颧;,<:>,. - ‘ t ~ \~ \ *~s\\.==.\'/:7 W 0 Ethernet is cheap(!), but not fast enough o Workstations can use > 100Mbps; PCs are catching up 0 Connectionless multi-protocol routers are difficult and expensive 0 Switches are cheap and easy 0 ATM hardware leads in price/performance >> 25Mbps switched ATM cost == 1x switched Ethernet >> 155Mbps switched ATM cost == 2x switched Ethernet lpsilon Networks Confidential and Proprietary Page 9

» _ ‘ L 1 » $9; ‘7;'§», ,~ / § \ * »/, ‘fa, ;,’,y km 2 \(,»,Me - _»1z-5-1.-e : ._ 6000 — 1000 — O Performance g1§~M*“» *~ \>\ .-1. ._ 1, £3‘ av; 5°°° eeée I Shared Ethernet 4.=f=?‘=>%Z; -'=II- '6‘ we ,,...- P J x 3000 - I Switched Ethernet 1e@10 r%ee ti; 5%‘ 2 O O O I ATM 16@25 5 Egg *4 ii ,0 ,; '>.‘“l/V’ $7? <\vwv»<\¢|> U Multi-Protocol Routers =1, ‘ ATM 16@155 lpsilon Networks Confidential and Proprietary Page 10

Where’d the simplicity go? o Layering for maximum compatibility leads to complexity o Virtual LANs, LAN Emulation mean more work for routers o ATM defines new, complex, address space - needs lots of administration o ATM defines heavyweight connection protocols lpsllon Networks Conf dent|a| and Proprietary Page 11

Ethernet vs. ATM Forum » IP ARP LAN Emulation IP Q.2931 ARP SSCOP ‘Ethernet ‘A ’ ATMV Ethernet Stack ATM Forum Stack lpsilon Networks Confidential and Proprietary Page 12

Open Networks: The Promise 1 - we " 0 Open platforms allow innovation by customers, 3rd parties, researchers 0 Open interfaces allow customers choice of implementation 0 Open competition leads to continual pricel performance improvement 0 Reference implementations greatly improve interoperability |ps|lon Networks Confidential and Proprietary Page 13

O en Networks: The Reality o Desktops & servers have become open platforms o Hubs, bridges, routers, switches have become complex, closed, proprietary systems o Network vendors seek lock-in through proprietary interfaces, esp. management o Standards usually behind market need; typically no reference implementation 0 The Mega-Hub is the new mainframe lpsilon Netwo ks Confidential and P p etary Page 14

The Solution » ~ 0 Marry router slw with switch hlw 0 Not just a layering of routing above switching; route management replaces switch management 0 Open protocols, open platform for router hlw & slw to enable innovation lpsilon Networks Confidential and Proprietary Page 15

Strategy I ,~ 4 -_ "’”’”¢»”“5.i’ ‘ \' I1 ‘ o Use public domain routing software; leverage oversupply of ATM hardware o Publish protocol specifications, management interfaces & reference implementations for all components 0 Deploy on commodity, open systems o Encourage commoditization of industry; foster competition to grow market o Create interoperability products for ATM Forum standards, Ethernet, etc. lpsilon Networks Confidential and Proprietary Page 16

, T ~- = 0 Soaleable networks through routing 0 Leading edge price/performance - ATM hlw 0 Lower cost of ownership - simple, familiar administrative model 0 Choice of vendors, easier customization - open routing platform lpsilon Networks Confidential and Proprietary Page 17

Alternatives Fast Cheap Simple Routeable Open Ipsilon + + + ++ U + + Ethernet - + + + + - - - Switches Newbridge + ? - + - VIVID Fore + ? ? ? - Cisco + - - - - + + I - Fusion Synoptics + ? - - - Lattis ATM Forum + - - - - Standard Ipsilon Networks Confidential and Proprietary Page 18

ITM VS. ATM .» ,. =_, ,. :1‘ »_ . 5 ,;,;,,.* ,, ¢',11_1:¢;11*->< I , '>-<1 . ‘ t / }i‘\\\\“1‘<’%;§*,€§:v 1'~>1;“/ 14*"-"*°°**~;"M“: 1 >’-.;=;=;a:a=&;.;2;a;:,»; ...;.=~e:=.:===.-':=:-.-==a'" - ~ . _ \~ ¢ K »:Yf,*§»'1f’“‘f.=‘~.’~€‘l=2§’3§““""/-‘i‘i ' 1' » W 1&5.‘ \:\:>M».%»~>,“‘%;%l:%?-.'§.>?C.'_' ‘ii; o ITM uses ATM hardware 0 ITM uses IP focused software protocols 0 Both ends must be ITM, e.g., switch & hub or switch & NIC o ITM uses router software instead of ATM’s connection management software o ITM has PVCs just like ATM lpslon Networks Co fidentiala d Proprietary Page 19

ITM vs. ATM -- - - . - -.»:-1=»~.>s>=»=-a»'>»;>E>2=-=2-::»a:a:a:a=a=a=a=a=a=a: /w‘"‘=‘;JyTZ’ ’ €"";¢ : ' - ' ', ' *1 ‘ /" 2. I?" i\\\»~;‘;; ,=:.;=;=;=;=;:;, -s;:;._a;a=s,e’ _.-;:=.-;av:::_.,s-.¢ .= _ .. ~ ~ *%?w><>*~§z»>>%'5~»~»*»-@ A7) 1" J :=.=:;:¢»_:\.;a%;:e»;:?i:'s;zea:.":'"2- IP ARP LAN Emulation Ip Q2931 CARP SSCOP AAL5 AAL5 -ATM ATM " ITM Stack ATM Forum Stack lpsilon Networks Confidential and Proprietary Page 20

ATM Workgroup . . , =.'=:..;=_<-=-;:_ 1,, _=—'@;¢2~~<;* “:1 »> ~»~ 1 ¢ " K\ *1./4 an ‘M .4" , =-<=¢<-=->1»>¢=si.E-E*‘<-2.=-=-=.».=.-.=.. >s:1»:».==s'=:=s I:-- . Lollipop Router Lan Emulation Server amz EEEE x L <*~ HE L ‘ lpsilon Networks Confidential and Proprietary Page 26

1*}: " 1 -1- ' 0 IP world is frustrated 0 IP can join forces with switching 0 Networks should be open lpsilon Networks Confidential and Proprietary Page 28

‘ ‘ * ;/ ~, 3 , W /Y" I '-.§ \;,. 7* _. \:\ ~ 2-I =-.v=-='“‘:“"“’=;'~‘ = F3 215 .....£ . W Big ATM Backbone o v /Lollipop Router @ gm z ).= ;.,,,. Ki‘ lee Q) Q) lpsilon Networks Confidential and Proprietary Page 24 EE-'2“ E525 EEEE i ‘* L ‘*- L <*

Big ITM Backbone as. ’ ..,.1,,.Z.,1..é:.‘£§;_ xi .- ‘E>59>'e;?E:E:E:E:E:E:E:E:E:E:=:E:E=k \ 4% i - 1 =_ i @ @ 22:" lpsilon Networks Confidential and Proprieta B510 -—1l L <"- L. <1 Cl.L ry Page 25 ' g»,“w<>~\~ » ,77;1>~r=r>z;":(2;;~,§>;r,z>w;¢¢zs~zmss<\§»1@<4=\-$\V_*>ms;,.g»;;;;;;g:...=;;.;;;.; _;.~=;.==:;;", 1 4 »: ~; v '

' Backbone 4 Q 1 , ‘ -. ..-:.fsv.za=ziiEEI.- '5E-:5;2-;=;====.=.=._..-... lpsilon '1‘ W 7*» > l RY: 3.1//4; mi‘ /AQ ~=;== ,> »<<¢>¢=2: ~\ ‘J, ¢ *3 :§>s1;» $”I!.~’.;~ w » .. '3 ,4 i;;5:§¢»>§ rw~=<= -, D . \ \\ " Q. E525 -- <- <= Page 21 Networks Confidential and Proprietary

=.“».\» Small ATM Backbone ’“"‘ *A* W" » A- we er, “9;¢:4 1,.»--..-7»-».»» w \ \ A an ’ >i?'””‘ » T "> /é’) §”.’5i‘I§ \ 7< £3 3 r W 1 M : » :5:5:5:§=E'E:g§:!w:-::wca-:&m4E§§ 5E:E:E:E$:5;s>m:-:e==='-= 2 2 I -W; V »,=-=m£- ’ Router Router - ‘R E5! ig ;;:; L 1 lpsilon Networks Confidential and Proprietary Page 23

Segmented Backbone $2 Q >11 aw ~. 5 2282;? 4; ,A ~~: . > r ;££A&“\%;/,,%/@s4m.T?<,iL,'ii‘ V r \\~ < » € % w ’ e ~AM,,, »,> Q Router 4. .». W‘ *5 ’ <\ lpsilon Networks Confidential and Proprietary Page 22

ITM Workgroup V . '=..=..~;-;;=-5.5, '"";1;‘;»§»a:1'~;A~z'/<%»(§§<#¢:/:/x 1*=:x§::<(:<*;:. .,:;;\,;.. - \ »:::»1s\:$“ 1 ‘~ ,~> ¢~/z' ~,i<<<<*~““ ' ‘ I 1,; , ,,_., V . -» / 1" » . ,. :1; r ‘ " / muiu ‘—.3‘i,' L lpsilon Networks Confidential and Proprietary Page 27

@1pSwiich Proposol 1" '*__ Hal-HS“ "*b‘ 4_._ s- ,.\. I I The broken promise of ATM I ATM user bcrcklosh in ATM vendor nightmares I ATM Forum vs. IETF I ATM: layer 2 or loyer 3? I Neiwork lcryer Trends I Where IP is woy oheod I lpSwiic:h: besi of boih worlds I The business

The Broken Promise of ATM I ATM: seamless WAN,LAN,publi<:,privale? I Public addressing differs from privale (E.164 vs‘.‘NSAP) A IWhaf abouf privafe/public securily? Where’s fhe firewall? I Public ATM hobbled by regulafion, lack of compefilion I Privale ATM sfill immalure - esp. managemenf 8. rouling (P-NNI); sfandards sfill lacking I Too many new concepfs & oplions fo compefe wifh LAN fechnologies

ATM User Backlash I LAN users will reject complexity of configuration and management; e.g. NSAP address management, LAN emulation servers, virtual routers, NBMA ARP, etc. I Users will reject lock-in to vendor specific “value added ATM" whether in protocol or management arena I Users don't want to enable another Cisco; vendors all want to be the next Cisco I Why bother with ATM to the desktop when Fast Ethernet gives more bandwidth with no new hassles?

ATM Vendor Nightmares I Everyone’s doing ATM - lots of competition I Most vendors underestimate software effort I ATM Forum 8. IETF keep making the software harder; also promote competing models I Lots of vendors doing simplified, or value added, software frameworks; start pushing single-vendor-network strategy I Interoperability takes back seat to time-to- market I Interoperability of management frameworks is nowhere to be found

ATM Forum vs. IETF I ATM 8. TCP/IP should be each other’s dream come true I ATM Forum dominated by telecom providers & h/w vendors - no understanding of LAN & lnternetworking requirements I IETF dominated by researchers with no respect or stomach for standards process; ATM is ‘packet shredder’ 8. ‘revenge of X.25’ I ATM Forum keeps innovating upwards in the protocol stack I IETF moving to less 8. less requirements of lower layers

ATM: Layer 2 or 3? I Current standards promote ATM as a data- link layer (I2), like X.25, ISDN, SMDS, LAN bridges I These protocols notorious for their mismatch with network layer protocols - lP,lPX, etc. I Requires routers (L3) to tie it all together I Complexity acceptable for WAN, but not LAN I User management investment is in network layer (incl. security) I What it ATM switches were routers instead of bridges (L3 instead of L2)?

Network Layer Trends I Novell IPX has the installed base I TCP/IP has the growth I Not much else matters - strong support of IP by Microsoft, Apple, Novell I lPng on horizon - potential to unity IP 8. IPX I lP evolving to support Q08 guarantees

Where IP is way ahead I Routing: intra-domain, inter-domain, policy QOS, & multicast routing - well researched approaches, standards deployed 8. in refinement I Security: protocols under standardization, key management frameworks being defined, firewalls known technology I Multicast traffic management - protocols in use, standards under way, existing multi- media applications (ATM very weak here) I Auto-configuration: standards deployed, getting even better with lPng I Mobility support - none yet with ATM

lpSwitch: Best of Both Worlds I Marry semantics of routing with speed of ATM hardware I Hybrid connectionless/connection based model: connections for speed & Q03 control I Simple new protocol for connection setup, close to ARP in function I No new address space to manage for ATM I Acknowledge need for different routing protocols in different parts of network hierarchy I New “flat” routing for leaf networks - eliminates need for virtual LAN model

|pSwilch: The Business Core lechnology is new conneclion prolocol 8. various rouling prolocol soflware Freely available specificalions, available reference implemenlalion lo caplure research communily 8. aempl de-faclo sfalus 2 possible company slruclures - be syslems company which resells h/w, or be s/w company which sells lo syslems companies Long lerm: be leader in deploymenl of innovalive lechnologies in lhe TCP/IP markel - mullimedia, mullicasl, mobilily, elc.