Taching Computr Ntworking with th Hlp of Prsonal Computr Ntworks Rocky K. C. Chang Dpartmnt of Computing Th Hong Kong Polytchnic Univrsity Hung Hom, Kowloon, Hong Kong csrchang@comp.polyu.du.hk ABSTRACT This papr dscribs an xprintial larning approach to taching a foundational cours on Computr Ntworking. In additional to th traditional laboratory sssions and analytical problm-solving, I hav introducd a class projct basd on prsonal computr ntworks for th last thr yars. Each projct group sts up and owns an IP privat ntwork throughout th cours, and thy prform various xprimnts on it to vrify and tst th ntworking principls larnd from th txtbooks and lcturs. Th studnts fdbacks collctd so far ar xtrmly positiv. Not only thy can acquir practical skills during th procss, thy ar abl to bttr undrstand th abstract ntworking concpts and protocols through thir working xprinc with th prsonal computr ntworks. Catgoris and Subjct Dscriptors C.2.2 [Ntwork Protocols]: Applications - Routing Protocols; C.2.6 [Intrntworking]: Routrs - Standards Gnral Trms Algorithms, Masurmnt, Prformanc, Exprimntation Kywords TCP/IP, Exprimntal problm-solving, Prsonal computr ntworks 1. INTRODUCTION Computr ntworks and th Intrnt hav alrady bcom th most critical infrastructur today for information dissmination, businss transactions, human communications, computr gams, scintific computation, and vn national scurity. Morovr, computr ntworks ar ubiquitously dployd in many othr commrcial sctors, such as automobils, smart highways, smart clothing, smart appliancs, and Prmission to mak digital or hard copis of all or part of this work for prsonal or classroom us is grantd without f providd that copis ar not mad or distributd for profit or commrcial advantag and that copis bar this notic and th full citation on th first pag. To copy othrwis, to rpublish, to post on srvrs or to rdistribut to lists, rquirs prior spcific prmission and/or a f. ITiCSE 4, Jun 28 3, 24, Lds, Unitd Kingdom. Copyright 24 ACM 1-58113-836-9/4/6...$5.. coff shops, just to nam a fw. As a rsult, thr is a prssing nd to quip studnts on both undrgraduat and graduat lvls with a solid foundation in th fild in ordr to furthr fostr and flourish its dvlopmnt and applications. Howvr, in many ways Computr Ntworking ducation is still in an xploratory stag. For xampl, w hav sn for th last tn yars many diffrnt approachs adoptd by various txtbooks analytical [1], bottom-up [2], nginring [3], systm [4], balancd-viw [5], visual [6], and top-down [7]. Whthr ths approachs can ffctivly facilitat studnts larning ar yt to b sn and valuatd. As th contnt of th fild continus to incras in a vry rapid rat, it is also vry important to idntify a minimal st of cor principls to tach undrgraduat studnts, and this has bn discussd in th ACM Workshop on Ntworking Education last yar [8]. Morovr, thr is incrasing ffort on providing hands-on xprinc to studnts through mor traditional laboratory sssions/courss, implmntation of ntworking hardwar, simulation tools, and sockt programming. This papr considrs a foundational cours on Computr Ntworking for junior studnts in a typical Computr Scinc or Computr Enginring program. Th main thsis of this papr is that studnts larning in this subjct can b significantly nhancd if thy can intract with an actual computr ntwork in paralll to othr taching and larning activitis. In th nxt subsction, I will first prsnt th difficultis connctd to taching and larning this subjct. In sction 2, I dscrib my xprinc of using a class projct basd on prsonal computr ntworks to nhanc th taching and larning qualitis. In sction 3, I prsnt th studnt survy rsults to assss th ffctivnss of th approach. Finally, I conclud with som futur work in sction 4. 1.1 Idntifying larning and taching difficultis Why is it (not) so difficult to larn Computr Ntworking? is prhaps th first qustion that w as ducators should ask ourslvs (and our studnts). On of th obvious answrs is of cours to do with th rapid advancs in th fild. Only tn yars ago, on undrgraduat subjct and on postgraduat subjct wr prhaps sufficint to provid a rasonabl covrag. Today, howvr, an ntir undrgraduat program can b dvotd to Computr Ntworking, and its associatd subjcts, such as conomic and social issus. Many Computr Ntworking topics hav also quickly dvl- 28
opd into sparat subjcts, notably Wirlss and Mobil Ntworks, Optical Ntworks, and Ntwork Scurity. From th ducators point of viw, it also bcoms incrasingly difficult to tach th subjct ffctivly, partly bcaus th Intrnt s succss has attractd studnts with various backgrounds. Thus, it is hard to provid on class for all. Evn among studnts with similar ducational backgrounds, say Computr Enginring studnts, som of thm may hav takn profssional xaminations, such as Cisco s CCNA, bfor attnding a first cours on Computr Ntworking, whil othrs ar not vn abl to xpand th trm TCP/IP. Bsids th issus rlatd to th forvr changing natur of th fild (at last for now) and htrognity in studnts backgrounds, thr ar quit a numbr of obstacls to taching and larning Computr Ntworking ffctivly, as outlind blow. 1. Th principls undrlying Computr Ntworking ar intrinsically vry complx. Th layrd modl hlps undrstand and manag th complxity. But vry soon studnts find out that this layring approach has its own inadquacy. For xampl, th layrs ar not rally indpndnt of ach othr, and a ntwork layr could b dgnratd into a data-link layr,.g., IP ovr ATM. Thrfor, gtting th whol pictur corrctly is alrady a challng to many studnts. 2. Computr Ntworking concpts and protocols ar also vry abstract to many studnts. For on, thy cannot s typical ntworking quipmnts, and visualiz packts and protocols for thmslvs during lcturs. For xampl, talking about a LAN switch without sing on is alrady a hindranc for many studnts. Examining th krnl cod without a prior undrstanding of how th protocols work hlps vry littl. Computr animation allviats th problm to som xtnt, but it may not b abl to quip studnts with th ability of concptualizing othr mor abstract concpts on thir own. 3. Unlik Computr Programming and Computr Architctur courss, for xampl, rsourc provision for hands-on practical xprinc in Computr Ntworking is problmatic. A Computr Ntworking laboratory, if availabl, usually has a rady-to-us computr ntwork on which studnts can conduct various xprimnts. Howvr, studnts should also b xpctd to know how to st up a computr ntwork from scratch aftr taking th cours, vry much lik knowing how to writ programs aftr taking a Computr Programming class. But it is not quit possibl to provid adquat rsourcs to achiving that in many acadmic nvironmnts. 4. Many trminologis and acronyms ar introducd and usd in th fild, and som of thm ar vry similar,.g., ARP and ARQ. Wors yt, ths trminologis ar oftn not usd consistntly, spcially in th industry,.g., hubs, switchs, switching hubs, port switching hubs, and sgmnt switching hubs. Studnts, on th othr hand, ar mor usd to clar and formal dfinitions. For xampl, many studnts in my class ar vry prplxd by th trm round-trip tim whn it is first mntiond during th subjct ovrviw. 5. Som of th ntworking problms ar difficult to comprhnd and apprciat du to studnts common lack of practical xprinc. It is gnrally not difficult for undrgraduat studnts to undrstand th accss ntwork tchnology, bcaus almost all of thm hav xprinc of accssing th Intrnt via ISPs and LANs. Howvr, it is rlativly difficult to comprhnd th scalability issus in intr-domain routing and th importanc of traffic nginring. 6. Th squnc of covrag can also affct studnts larning significantly. Computr Scinc studnts ar prhaps mor comfortabl with a top-down approach whil nginring studnts may find th pur bottom-up approach mor logical. In ithr cas, it is bound to hav missing information, which is yt to b rvald latr, in ordr to complt th ntir storytlling of th Computr Ntworking intrnals. Unfortunatly, studnts may stumbl on ths missing information. Thrfor, a good squnc of covrag may follow a nonlinar path which is in contrast to a rigid topdown or bottom-up squnc. 2. PERSONAL COMPUTER NETWORKS On issu that clarly stands out from th discussion in th last sction is that a lack of practical and maningful xprinc with an actual computr ntwork could gratly hindr th undrstanding of th Computr Ntworking concpts. Thrfor, practical assignmnts ar usually supplmntd in a typical undrgraduat cours,.g., sockt programming assignmnts and laboratoris. Although sockt programming assignmnts and laboratory sssions ar usful, thy do not allow studnts to intract with th krnl of a computr ntwork. To fill this void, I initiatd a class projct thr yars ago to lt studnts xprimnt with a prsonal computr ntwork throughout th cours ach group is xpctd to work on thir own ntwork in paralll to th cours progrssion. Howvr, th amount of rsourcs rquird for supporting 17 studnts is obviously vry dmanding, in trms of both quipmnts (PCs, hubs, NICs, tc) and spac (spac is xtrmly costly in Hong Kong). Finally, I hav com to th conclusion that th only solution to rsolving th rsourc constraint is to bring th ntworking laboratory to whr studnts ar. That is, ach group is rsponsibl for finding thir own machins, OSs, ntworking quipmnt, and spac. Although it sounds lik a vry difficult task, all groups in th past thr yars wr abl to find all th rsourcs and finish th projct. Th projcts ar prformd on group basis. Th ntir projct consists of svral phass. Th first on is to find all th rsourcs to form IP privat ntworks which ar usually situatd in studnts homs. Th ntwork is st up in a month s tim. Thrfor, th studnts had to rad ahad and undrstand som of th basic ntworking concpts, such as IP subnts and NAT. Th first phas is concludd with a prsntation on th ntwork stup. Th scond phas is to obsrv how ntwork protocols work by conducting xprimnts on th prsonal computr ntwork. At th nd of th projct, ach group is askd to submit a writtn rport and mak a final prsntation. A sampl rport is availabl in [9]. 29
192.168..2 192.168..3 Hub 192.168..1 Subnt 4 11Mbps Wirlss (Ad-hoc) Ethrnt Ntwork/Subnt ID: 192.168.1./24 SSID: PAUL WEP: 1:3b:4d:2b:3:6a:7c:45:12:1d:2d:14:5 (14bits) KEITH-NOTEBOOK wlan: 192.168.1.2 turtl (Windows ME) th: 172.16..17 KEITH-NOTEBOOK th: 172.16..15 1Mbps Fast Ethrnt switch Subnt 1 1BasT Ethrnt Star Topology using Hubs Ntwork/Subnt ID: 172.16../17 ED_NOTEBOOK th: 172.16..16 192.168.1.1 Routr th: 172.16..1 192.168.1.2 wlan: 192.168.1.1 Paul (Rdhat Linux 8) ED_NOTEBOOK wlan: 192.168.1.3 th1: 172.16.128.2 Figur 1: A minimal prsonal computr ntwork. 1Mbps Fast Ethrnt switch 516L (Rdhat Linux 9) th: 172.16.128.15 2.1 Configuring privat IP ntworks A minimum st up, as dpictd in Fig. 1, includs two subnts an Ethrnt LAN and a PPP ntwork. A Linux systm is configurd as a routr that also prforms Ntwork Addrss Translation (NAT). Although th tst ntwork is small, it contains both point-to-point and broadcast ntworks. Morovr, th PPP s MTU can b manually configurd to catr for xprimnts on IP fragmntation and TCP throughput. Advancd studnts ar abl to configur mor complx ntworks and srvics, such as that in Fig. 2. Starting from last yar I hav askd studnts to configur an IPSc tunnl btwn two such privat ntworks. Thrfor, two or mor projct groups nd to work among thm to st up such tunnls. Through th procss of stting up th ntwork, studnts pick up various invaluabl practical skills of configuring ntworks, diagnosing ntwork problms, and finding solutions to thm. Ths skills cannot b taught adquatly in traditional laboratory sssions. In particular, th studnts will b familiarizd with th followings aftr this stag of projct: 1. IP addrssing schm 2. Th concpt of IP subnts and subnt masks 3. Dfault routr configurations 4. PPP ntwork stup 5. NAT stup using iptabls 6. Ntwork diagnosis tools, such as ipconfig, ping, and ntstat 7. Stting up various ntwork srvics, such as SSH, HTTP, TELNET, SAMBA 2.2 Ntwork xprimntation Th scond part is to rinforc th principls larnd in lcturs by prforming various xprimnts on th ntwork. In othr words, th hands-on xprimnts draw studnts closr to th Computr Ntworking principls, bcaus thy can now visualiz how protocols work for thmslvs. Thr ar svral typs of xprimnts. Th first on is simply to obsrv how ntwork protocols work,.g., thy can obsrv 1. How ARP rsolvs th targt MAC addrsss using data-link broadcast, Ntwork/Subnt ID: 158.132.45./24 Gatway: 158.132.45.2 Intrnt th2: 172.16.128.1 th1: 192.168..1 Subnt 2 1BasT Ethrnt Star Topology using Ethrnt Switch Ntwork/Subnt ID: 172.16.128./17 th: 172.16.128.1 1742L (Rdhat Linux 8) Subnt 3 1BasT Ethrnt Ntwork/Subnt ID: 192.128../24 kn-hall th: 192.168..15 Figur 2: A mor laborat prsonal computr ntwork. 2. How th NAT routr translats th intrnal IP addrsss and ports, 3. How a routr fragmnts an IP packt, 4. Th TCP thr-way handshaking mssags and th protocol valus, and 5. Th TCP s prsistnt bhavior in stting up a connction. Anothr st of xprimnts is rlatd to th masurmnt of th ntwork prformanc. For xampl, basd on th masurmnt rsults, studnts can invstigat 1. Th ffct of fram collisions on an Ethrnt ntwork s prformanc, 2. Th rlationship btwn MTU and ntwork throughput, 3. Th ffct of IP fragmntation on th ntwork throughput, 4. Th ffct of packt fragmntation on th ntwork prformanc, 5. Th rlationship btwn buffr sizs and TCP throughput, and 6. Th DNS dlay in rpatd quris. Th last st of xprimnts concrns ntwork faults and misconfigurations. For xampl, studnts will obsrv that 1. Th stats of a TCP connction ar not affctd by a nonprsistnt ntwork outag, 21
Assignmnts Projct for practical skills Projcts for und r standing M id-t r m t s t s 7 6 5 3 2 1 Agr Agr N u t r a l Figur 3: Effctivnss of assssmnt componnts for a larg class in 21. 2. Th protocol intractions as a rsult of configuring duplicat addrsss on th sam subnt, and 3. Th ffct of various hacking activitis on th ntworks. 3. EVALUATION This sction prsnts studnts fdbacks rcivd from an undrgraduat cours on Computr Ntworking in th last thr yars. Th class in 21 was a larg on, consisting of 168 studnts. Th classs in 22 and 23, on th othr hand, wr small, consisting of 54 and 46 studnts, rspctivly. Th mtrics of th valuation ar basd on th impact of th group projcts on th studnts larning, and th chang of studnts intrsts in th subjct at th nd of th cours. 3.1 Th larning outcoms Bsids th fdbacks on th group projct, I also includ thir fdbacks on th assignmnts and tsts for th purpos of comparison. Thrfor, th following four statmnts wr prsntd to th studnts, and thr wr fiv possibl answrs to ach qustion: Strongly agr, agr, nutral, disagr, and strongly disagr. Th fdbacks ar chartd in Figs. 3-5. Th numbrs of rsponss rcivd wr 133 (79%), 52 (96%), and 41 (89%) for 21-23, rspctivly. Th assignmnts hlpd m undrstand th subjct matrials. Th class projct hlpd m acquir practical skills in Computr Ntworking. Th class projct hlpd m undrstand th subjct matrials. Th tsts hlpd m find out how much I hav undrstood (or hav not undrstood). Th rsponss to both statmnts rgarding th class projct wr quit similar for th small class in 22 in that around 8% of th rsponss from th small class (strongly) agrd that th class projct was hlpful in th two aspcts with slightly mor favorabl rsponss on th practical skill. Th rsponss rcivd from th small class in 23 wr vn mor ncouraging. For th first tim, th fdbacks on th Assignmnts Projct for practical skills Projcts for undrstanding Mid-trm tsts 6 5 3 2 1 Agr Agr N u t r a l Figur 4: Effctivnss of assssmnt componnts for a small class in 22. Assignmnts Projct for practical skills Projcts for und r standing M id-t r m t s t s 6 5 3 2 1 Agr Agr N u t r a l Figur 5: Effctivnss of assssmnt componnts for a small class in 23. projct surpassd that for th assignmnts mor than 9% of th rspondnts thought positivly of th projct on th practical skill aspct. Evn with a larg class, th projct was vry usful to most of th studnts. Around 65% of th rsponss (strongly) agrd that th class projct hlpd thm acquir practical skills in Computr Ntworking, whil 51% of rsponss (strongly) agrd that th class projct hlpd thm undrstand th subjct matrials. Th rsponss wr quit undrstandabl. Although undrstanding th ntworking principls was actually a major objctiv of th projct, studnts still prcivd that acquiring th practical skills throughout th procss was mor important. Basd on studnts writtn commnts, many of thm flt that acquiring th skills of stting up ntworks from ground up would hlp thm land on jobs upon graduation. Anothr rason is that most of thm had nvr st up a computr ntwork bfor; thrfor th practical work itslf smd to b mor rwarding and significant. 3.2 Studnts intrsts in th subjct Th most important studnt fdback, in my opinion, is whthr studnts ar mor intrstd in Computr Ntworking aftr taking th cours. Thrfor, studnts wr askd whthr thy ar intrstd in Computr Ntworking at two sparat tims: during th first class (indicatd by pr-taching in Figs. 6-7) and during th last class (indicatd by post-taching in Figs. 6-7). Th rason for doing this is to accuratly assss th impact of this cours on thir intrsts in th subjct. Prviously in th larg class, studnts wr askd whthr thy wr mor intrstd in th 211
8 7 6 5 3 2 1 Strongly Agr Pr-taching Post-taching Agr Nutral Disagr Strongly Disagr Figur 6: Studnts intrsts in Computr Ntworking bfor and aftr taking th cours in 22. 6 5 3 2 1 Strongly Agr Pr-taching Post-taching Agr Nutral Disagr Strongly Disagr Figur 7: Studnts intrsts in Computr Ntworking bfor and aftr taking th cours in 23. subjct aftr taking th cours, but th rsponss did not clarly indicat th absolut lvls of intrsts. Although th group projct is just on of th many activitis that would affct th studnts intrsts in th subjct, I bliv that th projct s contribution is much mor significant than othrs. It is quit asy to obsrv a positiv corrlation btwn a studnt s intrst in th subjct and his intrst in th projct. Som studnts wr vry xcitd about stting up a privat ntwork and obsrving how protocols work, and in turn thy wr willing to spnd mor tim on th projct. On th othr hand, othr studnts would prciv th projct just anothr burdn to bar, and this attitud in turn dcrasd thir intrsts in th subjct. Th fdbacks ar vry rvaling. For th 22 class, for xampl, around 9% of th rspondnts (strongly) agrd that thy ar intrstd in Computr Ntworking in th first class. This rsult is also prdictabl, bcaus of th importanc of th Intrnt. Morovr, som studnts wrongly blivd that Computr Ntworking is quivalnt to WWW, or othr ntwork applications. Thrfor, aftr larning th actual Computr Ntworking principls, many studnts prcptions about th subjct changd. Som bcam awar that Computr Ntworking was not for thm, bcaus th contnt may b a littl too difficult to comprhnd. Howvr, othrs found th subjct much mor intrsting than bfor, bcaus both th xtnt and dpth of th subjct far xcdd what thy knw bfor. As a rsult, w s a highr varianc in th post-taching rsponss. Th prcntag of strongly-agrd cass jumpd from 19% at th bginning of th cours to 32% at th nd of th cours. On th othr hand, th prcntag of agrd cass dcrasd significantly from 72% at th bginning of th cours to 43% at th nd of th cours. Th othr cass also incrasd at th nd of th cours. Similar rsults wr obsrvd for anothr small class in 23, as shown in Fig. 7. 4. CONCLUSIONS AND FUTURE WORK In this papr I hav prsntd a group projct basd on prsonal computr ntworks. Th main advantag of this larning approach is to lt studnts practis and vrify what thy hav larnt from lcturs and txtbooks as th cours advancs. By doing so, th abstract ntworking concpts can b mad mor concrt to th studnts. At th sam tim, thy ar taught practical skills in configuring and diagnosing an actual computr ntwork. On possibl improvmnt is to furthr tightn th intgration of th class projct with th instructional part. Currntly, I start giving out th projct approximatly a month aftr th bginning of th cours. In th nxt yar, I am planning to giv out th class projct in th first class, and studnts ar xpctd to xprinc a crtain ntworking topic on thir prsonal computr ntwork as soon as th topic is covrd in lcturs. This approach is somwhat similar to th xprintial larning approach discussd in th ducation sctor. In contrast to cognitiv larning, xprintial larning attmpts to provid incntivs to larn, and in my cas th incntiv is to build a prsonal computr ntwork aftr complting th cours. Acknowldgmnt This work was partially supportd by a Taching Dvlopmnt Grant from Th Hong Kong Polytchnic Univrsity. Th ntwork stup in Fig. 2 is part of th projct rport by Edmond Chan, Joy Chan, Kith L, and Paul Chu [9]. 5. REFERENCES [1] D. Brtskas and R. Gallagr, Data Ntworks, Scond Edition, Prntic-Hall, 1992. [2] A. Tannbaum, Computr Ntworks, Fourth Edition, Prntic Hall PTR, 23. [3] S. Kshav, An Enginring Approach to Computr Ntworking, Addison Wsly Longman, 1997. [4] L. Ptrson and B. Davi, Computr Ntworks: A Systms Approach, Scond Edition, Morgan Kaufmann, 2. [5] A. Lon-Garcia and I. Widjaja, Communication Ntworks: Fundamntal Concpts and Ky Architcturs, Scond Edition, McGraw-Hill, 24. [6] B. Forouzan, Data Communications and Ntworking, Scond Edition, McGraw-Hill, 2. [7] J. Kuros and K. Ross, Computr Ntworking: A Top-Down Approach Faturing th Intrnt, Scond Edition, Addison Wsly Longman, 23. [8] ACM Workshop on Ntworking Education, Aug. 22. Availabl from http://wwwnt.cs.umass.du/sigcomm/ducation/workshop1.html. [9] E. Chan, J. Chan, K. L, and P. Chu, A rport on Ntworking in an Unsaf World!, availabl from http://www4.comp.polyu.du.hk/ csrchang/edmond.pdf, May 23. 212