DARWIN: On the Incremental Migration of Legacy Information Systems h. h~lel l . Bro~tie & .l~ie~lael .SIf)nehra~-er R~/7(~t ~ SE~(~[ olA 2000 Technical Report ~7 E~ F3er~;el~ C~ orlliLl 917~() ** No page found ** A bstract A.s is painlullv e-ident today, the deterioration of the transportation, education-. and other nati(;nal infra.structures ne~atively impacts manv asp~ct~s ol~ life, hu.sine.s.s. and our economy. This ha.s resulted, in part, when responses to short term crise.s discoura~e investin~ in infrastructure enhancement and when there are no eifechve mean.s t'nr inlra.structure evolution. This p aper responds to the detcrioration ol the inlormLation sy.stem (IS) intrLastructure that ha.s stron~ ne~ative impact.s on IS.s, on the or~ani~Lation.s they .supl-ort, and, ultimately, on the economy. This paper addres.ses the prohlem ot le~acy IS mi~ration hy methods that mcdiate hetween .short term cri.se.s and lonc term ~oals. It presents an etfective strat~y and a spectrum ot supportin~ methods lor mi~ratin~ le~acy ISs into a tar~et environment that includ~.s ri~ht.~ (l hardware and modern technolo~ies (i.e, inlra.structur~) such as a client-.server architecture, DB~l.S.s and CASE. We illustrate the method.s with two mi~ration ca.ie .studie.s ot multi-million dollar, mission cntical le~acy IS.s. The contrlhu~ion ol this paper i.s a hi~hly tl~xihle set ot mi~lration method.s Ihat i.s tailorahl~ lo mo.~t lc~acy IS.s and hu.sin~.s.s contcxts. Th~ ~oal i.s to .support continuou.i. itcrativc cvolution. Thc critic,ll succc.s.s tactor. and challcn,~ in d~plo~mcnt. i.s to idcntity apl-ropriat~ portion.s ot thc IS and thc associatcd plannin~ and mana~cmcnt to achi~vc an incrcmcntal mi~ration that i.s tca.sihlc vvith r~.sp~ut to th~ tu~chnic;ll and husinc~s r~uircmcnt.s. Thc p,apcr con-ludc.s with a list ol dc.sirahlc mi~ra~ion tool.~i tor which hasic rc.scarch i.s rc~luircd. Thc principl~.s dcscrihcd in Ihis papcr can hc u.scd to dc.si~ln luturc IS.s and an intrastructurc that will supr)ort c(!ntinuous IS cvolution tn avni(l tuturc lc~lac~ IS.s. l~he l~ on l~roject ol Whitten ('ommunication~. Inc.. i~ atteml-ting to e~reate a innovahve tor-l-rofit alternative to the ~ov(rnmcnt tun~le(l .~ichool .sy.~tem.s. It is ha~e(l on the ~rinci~le that Dar~,vinian evolution a~ lie~ not onlv to the .~ ie~. hut al.~o to our in~titutioni an(l or~ani~ation.~i. ** No page found ** 1. ~11GR.~TI~'G LEG.~CY Il~,FORMAT10~ S~STE~1S Most larg~ orcani~ations ar~ d~ ly mircd in thcir inlormation sy.stcm.s (IS) . in.s of ti1t /~ Tvl-icall~, thcir I.S.s Irc lar~T~ (c ~T,, 1()7 lincs ol codc). cacriatric (~ g ~ morc than l() y~ar.s old). writt~n in C()B()L. and usc a lccacy dataha.sc servicc (~ c IBM's IMS or no datahasc manaccm~nt .svst~m (DBMs) at all). Thcsc ISs ar~ mission critical (i.~.. cs.scntial to the or~Tani~ation .s husinc.s.s) and mu.st he opcrational at all times. Thes~ charact~ri.stic.s d~tine what wc ùall legacy information systems (legacy IS). Today, lc~acy IS.i pose onc ol thc mo.st scnous prohlcms lor lar~c or~ani~ations. Costs due to prohlems ol a .sin~le lc~acy IS (c.g.. lailure.s, maintcnancc, inappropriat~ functionality. lack ot documentation, poor perlormance) can oftcn excccd hundrcds of millions ot dollar.s per ~ear. They ar~ not only inordinately expensiv~ to maintain, but also Inflexi~l~ (i.c., dillicult ~o adapt to chancin~T husincs.s ne~ds), and brittl~ (i.e., ~a.sily broken whcn modilicd lor any purposc). Pcrhars worse is thc widcspread lcar that legacy ISs will. onc day, hr~ak ! ~vond rcrair. Such l'car.s comhincd with a lack ot t~chni(luc.s or t~-hnoloyy to tix l~yacy IS prohlcm.s r~sult in lS ap(lplexy. That i.s, lcy;lcy l.S.i con.sumc '~ k to ~5'-7, ol all IS rcsourc~.s. Thi.s prcvcn~.s oryani7ation.s trl~m moviny to ncw~r soltwarc. .such as clicn~- .scrv~r conti~uration.. currcnt ycncration DB~lS.s, and lourlh ycncration lanyua~c.s (~GLs). Consc~ucn~ly. or~ani/ation.s arc pr~vcn~c(l lrom right.si..ing that mvolvc.s movmy lrom larY~ maintramc compulcrs to .smallcr, lcs.s cxpcnsiv~ comrut~r.s that lully mc~ curr~nt IS r~luir~mcnt.~. Thi.s a~oplcxy, in turn, is a kcy contrihutor lo th~ soltwarc cris~ ~cw r~uircm~nls, ot~cn call~d lhc lS hackloy, canno~ hc mc~ .sincc lcYacy l.Ss canno~ h. cx~cndc(l and ncw IS.s canno~ hc dcv~lorcd wilh lh~ 5'~ lo 1()'-7f r~rnaininy rc.sourcc.~. Thc.~c prohlcm.s arc holh kcy molivation.s ol and major roadhlock.s ~o thc world-widc mo~cmcnt l(- rc-~n,Yinc~r coll-oralion.s an(l thcir major I.S.~. In lc~acv I.S miyration. an cxi.slim~ I.S i.s cv()lv~d inlo a ~ary~ I.S hy r~placiny lhc hardwarc and mu~h ot lh~ .sotlwar~ includiny thc in~.rtac~.s. appli-a~ion.s. and dalahas~.s. I~ndcr .somc ~ir~um.slan~.i, .som~ xi.~limy ~omron~n~.s ~am an(l .should. h~ in~or~orat~d inl~ th~ ~ar ~nvironm~n~. In lhi~ l~ar~r lh~ target envir(~nments ar~ int~nd~d to tak~ maximum ad~anlaY~ ot th~ h~n~tit.~ ol ri~,~hr.si~e~l ~omrul~rs. a ~ nl-.~r~r ar~hit~tur~. and mod~rn .~otlwar~ .su~h a.s r~lali()nal DBM~S.s, lGL.s, and CA.SE. For ~xamrl~, a mod~rn DBM~ (~.y., ha~kur, r~o~n-v, lran.~a~lion .sul~ort. in~r~a.~d data ind~r~nd~n~. r~rtorman~ iml-r~ m~nl.~) a.~.sl.~t.~ in in~r~a.~iny ~onlrol ov~r ~h.~ I.S ~.y., main~nan~n I~ al. r)rovi(l~s a ha.si.~; tol tulur~ ~wolulion and in~Yra~ion wi~h olh~r l.S.~i. For ~xam~ oBMS ~ould t;~ a lil~r~lti~J/~ . any a~rli~a~ion ~ould u.s~ th~ DB~S ~o a~ aluahl~ ~orr(-ra~ hat i.~ urr ina~ ihl~ du~ ~o lh~ ].~ ha .~r~ A lundam~nlal '_(!.1~ _a~y 1.S mi~ra~ion is ~ha~ ~h~ ~r_.~ I.S n(~ h~om~ a l~Ya~v IS. V~, h~n th~ mi~_r;l~ion i.~ ~omrl~ h~ ~ar~ IS i~ ~ullv or~ra~ional in ~h~ ~al Y~ nvironm~n~. .~11 aprli~a~i.)n ~d~ an(l U~'I in~ria~s ar~ ~omrl~ lv ~ri~n in a IGL an~l ~h~ da~aha.s~ r~.~id.~.~ on a ~urr~ m ~ n DB~IS. A ~is~ hoi~ ol ~ar~ nvironm~n~ will ia~ilita~ ~h~ ar~ til)n h~in~ mo~d ~o a wi(l~ vali~lv oi ~urr~m~ and ~utur~ d~sktor ma~hin~s and ~h~ dai;lha. ~ h~in_ d~rlo\,~d on a ~id~ v;lri~v (~1 ~omru~ine rlatlorm.s. H~n~. th~ tarY~l IS ~an r~.~alil~, h~ r(!rt~d ~o ~nvironm~ n~.s ar~rorliat~ to ~urr.~n~ and iutur~ r~uir~m~nt.s. Tha i.s. lh~ tare~l lS i.~ .si~n~d to h~ v~ry ~ ihl~ .e.. rortahl~) lor ~urr~nl and ~ulur~ ri.,~ht.~i.in~J an(l lo avoid h~()min~ a lulur~ _a~y IS In ~h~ r~sl ot th~ introdu~lion, w~; ~di.s~u.s.s miYration . Ira~el~.s ~n~ L onl~ L Illlel.lLlUII ohj~tiv~s, and introdu~ h~ alt~rna~iv~ mieYration ar~hit~tur~.s and m~thod.s that ar~,~ pr~.s~nt~d in th~ par~r. ** No page found ** ** No page found ** ** No page found ** ** No page found ** ** No page found ** ù L.~ acy IS.s .an he too hi,Y lo cul-over Manv leYacv ISs mu.st he operational almost 1()()'7~ ot the time. Manv leYa~V dataha.ses or tile.s re(luire weeks to dump or download. Even if th~ rewritt~n IS ~4ere tully op~rational. ~here are no techni(lue.s ~o mi~rate the live data lrom the le,Yacv IS to lhe new IS within the lime that the husine.s.s can support heinY without iLs mission critical IS. Live data mu.st al.so he converted ~o til the new sy.stem, a,Yain increa.sin~ project time and comrlexily. Thi.s may not ju.st add com~lexity, il often prohihits Cold Turkey alto~ether. ù .~lanaYement ot lar~e projecLs i.s hard. The difficulty ot mo.st lar~e projects is seriously under-estimated. Hence, there i.s a tendencv tor them to ~row uncontrollahly in head count. Few or~ani~ations are capahle ot mana,Yin,Y the development of an IS wilh the .several hundrcd con~nhutors that are common tor ISs o f the si~.e and complexity we are considerin~. Mana~in~ more and more peorle inevitahly hrinYs on the tamou.s Brf~o~.~ effe~t [BRoo75l resultin~ in less and less usetul work ù Latene.s.s i.s ieklom lolerate~l. LarY~ prol~l.s ar~ in~vilahly lal~ du~ to lh~ rrohl~ms ~it~d ahov~ anaY~m~n~ ra~l~n~u w~ars ()u~ ~ui~klv. ~.sr)~ially in or~ani~ation.s who.s~ ha.si~ tun~ion is not .sot~war~ rrodu~ion. This tr~ u~nlly r~sull.s in ~h~ t~rmina~ion ot l-arllv or mos~ly ~om~ d pro ju~ ~.s. ù Lar,Y~ p ro j~ts ~nd ~o hloa~ Th~r~ is a ~md~n~y tor larY~ pl-o j~ .s ~o h~om~ hloal~d wilh non~.s.s~n~lal ,Yroup.s. ~or ~xampl~, t()r a proj~ a.s ~ri~i~al as a l~,Ya~v IS mi~rali~n. or~anization.s may wanl t~- ~crlor~ lh~ ~ntrodu~ n ot n~w mana~m~nl l~hni~u~.s and t~hnok-,Yi~.s ~.,Y.. R~- ~n~in~rin,Y. CA.SE). This i.s oll~n don~ hy addin,Y addilional ,Yrour.s lo th~ alr~ady lar,Y~ proj~t. Group.s lhal ar~ not ~ri~i~al ~o ~h~ miYration it.s~lt in~r~a.s~ ~h~ hudY~t an(J mana,y~m~nt ~(-m~ xity, thu.s makin~ th~ rroj~t mor~ ùuln~rahl~ to t~rmination. C~)ld Turke~ in-ol~es hi~h ri-k. It has heen arrlied and has lailed manv times m lar~e or~ani/a~inn.~ 'e n()w lurn our a~tention tn the all~rnalive. low-ri~k and nov~l .strate~. Chicken Little. ~he t~)-us and eontrihu~ion (lt ~his parer. Chicken Little invol~e.~ mi~r;l~in~ ~he leYa~y IS, in pla~e. hy ~mall inercmun~al ~p~ un~ h.~ de.iircd lon~ ~erm oh~ e i.s re;lehe(l. Eaeh ~p re~luire.~, a rela~ively .all re.~ouree ailo~a~ioll (L'.'~., a tew ~)er.Soll yealS~, a .~,hort ~ime. and rrodu~ a ~reeiti~ mall ru.sul~ ~owa~ hu de.~,il-e(l ~ al. Thi.s i.~ in sharp eon~ra.s~ ~o Ih.' ~ as~ re~ouree re~luiremen~. ,~t a e()mple~. re\hli~. ~e. n. hundned.s ot rur~on ye;lr.~), a mul~ ear de~elopmen~. and one ma.~.~ivc re.sul~. 11 a Chi~l;ell Li~le s~er tail.s, only ~hu t;lilcd ~er mu.i~ h~ rerea~e(l ra~her ~han Ih.~ en~ire plole~. .Sinee .~,~er.s are de.siYneJ lo he relatively in~xpen.slve, .~u-h in~remun~;ll ster.s do n()l nued ~o rromi.~e dramalie new lune~ion ~o ~lu~ tunded. Eaeh prohlem ei~ed in Seelion l. l ean he addre.s.sed in an ineremental tashion. In addition~ hilures in individual .sters may indicate lar~e or previously untor~seen prohlems. Du~ lo the incremental nature ot Chicken Little, .such prohlem.s can he addre.ssed incremen~ally. Hence, Chicl;en Liltle i.s much saler and more lea.sihle than Cold Turk~y. In this pap~r. we inve.sti~ale and aprly the Chicken Little mi~ration strate~y to le~acy ISs, trom well .structured to un.structured. The key to succe.sst'ul Chicken Little mi~ration, and ** No page found ** ** No page found ** ** No page found ** ** No page found ** Migration 2: Mi~rate OUL;jde Plant. We proposed OSP for the second mi~ration. .since it was the next easiest t~ .separate. However, the SAG ~nd OSP mi~rations could proceed in parallel since they are so similar. The OSP mi~ration proceeds as t'ollows. Separate the le~acy OSP appli~ation module and data from TPS, as illustrated in Fi~ure 5.3. Au~ment the TPS Application Gateway to deal with OSP calls. Mi~rate the OSP interfaces, application. and database. We recommended an it~rative cut-over a~ain focused on central offices. Once OSP is cut-over, retire the le~acy OSP application and data from le~acy TPS, as illustrated in Fi~ure 5.4. 20+ malor201 major 1,200 Mlnor lS~ lS~ lS~ TPS I Core Switch ~ TPS Intertace ! TPS ~Advanc-d~val~dation DOA I Servlces I Figure 5.~: TP.S ~ ration ~ Architecture ~ligration 3~ rate TP~S .Switch Intertace. For many ot the re;lson.s cite(l aho~ve, we recommen~e(l TP.S Switch Intertace tor the third mi~ration. It procee~t.i as does SAG and ()SP and has the mi~ration architecture illu.strate~t in Fi~rur~ 5.~. Migration l: ~licrateThe TPS Interlaces. The TPS interface mi~r~tion is a lar~e, complex proee.s.s. It i.s mis.sion cntical since all intertaces must remain operational ~hrou~hout the mi~ration. This includes intertaccs tor over 1,~()() minor ISs and 4() major ISs. The TPS intertace migration was complicated hy a second t;actor. Further an;lly.si.s revealed that the application.s yct to he migrated were not decomposahle, as ~A~ere .some ot their user and sy.stem interlaces. To .simplify the miyration, we proposed that the non-separahle intertaces he eliminated by altering the leyacv applications to use one ot the .separahle TPS intertace.s (i.e., TPS SOE, TPS MIA, TPS Access Interface). we viewed this as nece.s.sary even though it violated our ,Yoal ot altenn~Y legaey code as little a.s possible. As a result, there will be a large number of le,Yacy user and system inter~ces to he migrated. 4() ** No page found ** illu.strated in Fi~ ure 1.4. That is, after micration ~, we tound that th~ remainln~ applie~-ti()ns were non-decoml-o.sahle. exeept hy .sic7nil~icant ~h;3nf es to ~he le~ acv IS whi~h w~re deemed to he far too risky Hence~ micration 1 end~d the semi-d~comro.sahle micration The remaind~r of TPS re(luired a miC7ration method lor non-deeomro.sahle IS.s as descrihed in the next .s~cti(~n. 42 ** No page found ** da~;l or func~ all he identitied hut not proven to he independen~. thev may .sim~ have to he rel-lic;lteJ .md thc le~acy and tar~et copies coordinated until i~ i.s demon.~rahly safe lo re~lr. the lc~a~v ~er.sion.s. Step ~'3 l~era~i~vely de-omro.se the le~acy IS structure. (See Step S~) Peelin~ Ihe onion' hy takin~ otf separahle portions may produce a number ot decomposahle components and a non-decomposahle core. These can he treated separa~ely. Even if non-decomrosahle le~acy ISs cannot he d~compo.sed, some restructurin~ ~ould tacilit~te the mi~ration. Step ~ era~ively desi~n the tar~et interfaces. (See S~ep S4) Step I~'5 Itcratively desi~n the tar~et applications. (See Step S5) Step ~h Iterativelv desi~n the tar,~et dataha.s~. (See Step S6) Step ~'7 l~cr;lti~ely create and in.stall the IS ~ateway. (See Ster .S7) An IS catcway encap.sulate.s an entire l.S. a.s illu.strated in Fi~ures l.~ and ~. l. It i.s similar to an in~erlac~ ~a~ew;ly excel-t that it mu.~t l-rovide the coor(lination lunction v~ith the inlormati(!n availahle at the u.ser and .sy.stem intertace level.s. This m~ke.~j the IS catew;lv ver~ dillicul~ lo huikl. It could al.so include a ~(~rnrnllr7it (rti~ns ~lt~U~I~ r~luired to micratc ~rum one communication.s technolocy to another (e.c., IBM'.s CICS to TCP/IP), therehy a.s.~ tinC the mi~ra~ion Irom mainlrame to cli~n~-.server architec~ure.s. In previou.s me~ho(l.li, the ~ateway lunctionality decr~a.se.s a.~ le~acy IS component.s are n~tired. Due ~o th.' lack ol . truc~ur~ in tht non-decompo.s;3hl~ lecacy IS, it may not he po.~.sihle to n~tire any le~acy comronent.s until the mi~ration i.s comrl~te. Hence, the IS ~atu\h;lv v~ill continu~ to increa.se in comrl~xity until the non-decompoxahl~ le~acy I~S can he relir~d. Furth~r, it ma~ not he u.se~'ul to mi~rate .some lecacy I.S component.s at all ev~n thouch they are .~till re~luired. For example. they may he re(luir~d lor .seldom ac~es.sed. arehival inlol-ma~ion ()r lun~tion.s. Henee, the lS ~ate~hay may heeome an intecral i~ar Lhe u]~ima~e IS. Step ~ et.l~ively miC~ra~L~ the lecaey dat;lh;~se. (See .Step .S~) Th. dillieuily ~1 Ihi.~ ~tep derend.~ on the re.sult.s ol Ihe previou.s .step.s. It ma~ he verv dillieull or .~ lo ae~e.~.s data in le~a~y da~aha.~e servi~ in non-deeomro.s;lhle l~ca~y l.i.~. Thi.~ ma~ h. ~-iue lo ~he le~a~y da~;lha.s~ .~el-viee. ~he ar~lieation.i. or Lhe .slru~ture ol the d,.la .~11 ol Ih.~.~. Iael~ arrlied to Ihe TP~ mi~ration. whieh wa.~, a par~ieularlv ha(l ~a.se ol n()n-dueomr~)sahl. le~a~y d,ltah~ . mi~ration. Fir.~t. there wa.~ almo.st no knowledce ol the int~ln~ Iruelur~ Ih. Ie~a~y dataha.~e. Thi.s wa.s ~ompli~ated hy ~he la~t that th~ phy.~ al ~tru~ture ~a.~ elullul-ed wilh implemellta~ion artila~.s that were hard to di.stin~ui.sh Irom appli~;lli()n iallu A.~ a re.sult, we propo.sed that the exi.s~inS~ leca~y appli~atlon.s (e.~ d;ltaha.se ~luerv, r~i ort ~eneration, and ae-e.~. routine.s) he u.sed to ex~ra~t leca~y data. Thi.s ~i~nili~an~ly ien~lllened the diltahase mi~ralion time. Step ~ el-atively micrate the lecaey appli~;ltion.ci. (see st~r ~ Step ~l~) Iteratively micrate the le~7a~y inter~aee.s. iSee St~p ~Sl()) Step ~'ll Iteratively ~ut-over the tarcet IS. (See Step ~1 1) 41 ** No page found ** int~rlac~.s. TPS S()E. TPS MIA, and TPS ACCESS Int~rtace; and ~he re.st ot TP~. ~allud TPS-rest. Funhur analv.sl.s tound that pan oi TPS-rest ~a.s semi-decompo.sahle. illu.s~ra~ as TPS-r~st-~ in Fi~ure 7.~. The r~mainder ot TPS-r~.st, TPS-rest-~ was tound ~o hc non- d~composahle. IS~ End U~er~ ~ 1 _ ~ ~ Gateway C ~t t~' Leqacy Inter~ace Legacy Applicatlon Leclacv Database Service Legacy n:lta 44 Fi~lur~ 7. l: Hyhri(l Le~aey IS ~ ration Arehi~eture S tep TPS 1.3: ltel-a~i~elv d~ oml-o~u anLI mi ~rat~ -rPs ~rl~ ~he d~orn~o.~ahlu le~a~y IS mi~ration method to the deeomrosanle l~ ~omronen~ a.~ I.S rropo.s~d in Se~lon ~. A~F~ly ~he ini~rtaeu mi~ra~ion mu~ho(i to th~ de~oml-o.sahle l.S interta~e.s. al.so de.serih~d in Se~lon 5. Arl~ly ~he non-d~oml-o.sahle le~aey IS mi~ration method to TPS-REST~ The re.sultin~ TPS mi~rati()n ar~hit~uture i.s a.s illustrated in Fi~ur~ 7.' 4~ ** No page found ** ~. RESEARCH AGE!~'DA There are tcw conccpts and almost no ade4uate tools or techni(lue.s to .support le,nacy I.S mi,Yration. Those that exi.st do not scale up to meet the challenges po.sed in the design, development, mainten;lnce, or evolution of large scale ISs iBRoD~]. This iection comments hrietly on technolc)gie.s and tools that could dramatically helr the migration methods descnhed ahove. Many ot the.se suggestions re~luirc re.search. 8.1 Gateways Gateways are critical to all mi,Yration methods and architectures proposed in this paper. They also pose the greatest te-hnical challenges and can he extremely costly to huild and operate. Indeed, we recommendcd that they he designed and huilt hy datahase systems experts. It would he enormou.sly valuahle to dcvelop generic ~ateway technolo,~y to encapsulate datahascs, intertaces, communications~ or any other major system component. Gatcways are sometime.s called .surrolln~l technolo,Yy. For ex;lmrle. C~IS need.s to move lrom VSA~I liles to DBM~ technolo~Yy .~ ly (i.e., the old mu.st gracetully evolv~ ~o the new without a di.scontinuous hreak). Gatewav.s are much more hroadly applicahle than .simply lor migration architecture.s. They may he~ome major ~omponent.s ot luture IS architecture.s. Whenever two sYstem.s ar~ re~uired to interact. .some interta~e i.s re~luired. 11 the interactions involve anvthin~ mor~ than simple remote proce(iure c;lll.s. it i.s o~ten nec~s.sary t(! de~elop a soltware Interlace component. Su-h a comron~nt i.s a yat~way (i.~., hetw~en the two systems). Gatewavs have .strinyent r~(luirement.s. For e.Yamrle, those h~tween mi.~ ion critical ISs must h~ very reliahle, rohu.st. and etlicient. They c;lnnot impo.se a .siyniticant perlormance p~naltv on the old IS (.say more than l() percent). It i.s not ~le;lr how to achiev~ thi.s level ot ~rlorman~. R~(luir~menl.s tor l.S inI~ra~tion (ott~n mi.sleadinyly ~alled I.S inteyration! and di.strihutecl computiny (e.y.. client-serv~r) are ~rowiny dramatically. Gateways are a tre~luently proposed .s()iu~ . Indeed. .some proro.sal.s tor nex~ _ener:l~ion compu~iny ~)MG~la. ()~lG'~lh. '.~IA.~ '. BR()D'~', BR()D~I ar~ ha.sed on y at~way.s. Res~ar~h is re~uired to under~tan(l an(l de-~lo~ y ~nerie y;l~way te~hnoloyy. Th~re i.s an inereasin~ numh~r ot ad hoe (1.~., tor a .~mall numhcr ot .s~ .sy~m.~i ya~way ~roduct.s ~urr~ntlv with minlmal tun~ nal.~ n. no u~ eoor(iina~ion). Gen~ral l-ur~o~e ~ool.s are re~luired t()r ~on.~m~in~ ya~-~ ays, ju.st a.s DB~I.S.s are yencrie ~ool.~i tor eon.s~ru~in~ da~;l in~en.siv~ a~ a~i()n~. Ad h()~ _a~ ay.~ ar~ ~x~remely ~omr-lex and ~o.s~ly ~o ~on.s~rue~. Thi.s ~an he se~n In lh.~ .(?.~ an(l ~omplexity ol yat~ ay~ rrovided hy DB~1.S vendors [~) .sur?l~or~ hel~ro~ ene~u.~ di.~l-ihu[e(l DB~I.S.~. to~u~e(l on lheir r?rodu~t. X.~ Specirlcation E!-tractor .S~e~iti~a[ion.s tor l._a.~ IS.~ are ne;llly non-~xi.st~n~. For ho[h C~IS and TPS. th~ ~od~ i[.~ell i.~ Ilmo.st the only docum~ntatlon. In r1(lrro~t domam.s ~uch a.~ hankiny. it may h~ practical to write a ~ro~ram that e.Ytract.s .srucit~ica~ions trom ~hc old cod~. .Such a tool wouki h~ imm~n.sely in dccrv~ modul~.i to h~ rewn~en. ~.3 Application Analvzer Current codu analyz~r.s ;3re ade~t at determinin~ stati.s~ic.s lor a yiven arplication (~ numher ot' lin~s ol eode, call ~ rarh). Althou,~h this inlormation i.s valuabl~. n~xt ~cn~ration tools coul(l do much more. A code analy7.er could insp~ct th~ mo(iul~s in a lar~ sysl~m and .1~ ** No page found ** he mad. a.i l.lr~u a~ ros.~ihle. Allhou~lh lh.' eurrenl ti(1ure ot merll ~'() 1 Imrr~umun~ make.s re~n~ln~ rracli-al~ il woul(l he more altraclive w ilh hiyher le~cl ~ U h~l ~u~h : ~GL would ~(!nsi.~ ol i.s an open re.~earch ~lue.~ n. .~ Performance Tester ~i~in~ ot new har(l~hare ha.s heen a ba~l. oJ th~ en~ p~ exercise in Ihe ca.se sludie.s. Al~houyh i~ ea.sy ~o di.s~nhu~e da~a and pro~e.s.siny c ver multiple jell- bean sy.stem.s. i~ would he de.sirahle to have a perJorm~mc~ anal~ ~r that could predict resron.s~ time tor the C,7~1S and TPS tran.saction load.s usiny a new .schema on n~w hardwar~. Such si~iny studies are time con.sumin~7 and coukl he aided hy a powertul tool. The okl ada~7e ''load a larye datahase earlY- and h~nchmark your hi~h volume interactic)n.s'' .should he rerlaeed hy advice that i.s cheaper and ea.sier to l'ollow ~.~ Application Cut-over The C~I~S and TPS miyration re~luire several. eritieal cut-over.s. In ~ larye terminal (or PC) network. i~ i.s unreali.stie to move in.stantlY tlom ()nu I~ ~o Ino~her. A s7raeelul way ol movinL ~o a new environmen~ a tew u.ser.s a~ a ~ime ~ ould hc ~ery helrtul. Re.seareh i.s re~uir~d in~o me~hod.s and lool.s lo ~urrorl smoolh. ileralive eu~ er trom icyaclv to ~ar,Le I.S ~l~mronen~.s whilc Ihe I~ i.s under c(~nlinuou.s opcralion. X.11~ Distributed IS Development and ~ ,ration Environment Di.suihutc(l elient-~el~er eomru~in,L i.s vcry ro~ular t~or whieh an inereasin~ly lary e numh~r ~l eonee~t.s. I~ol.s. and te-hni~ue.s are heiny rrovided. 1~ woul i he hi~yhly de.sirahie to have I eon.si.stent cnvironmen~ ~ha~ .sui~ort.s ~he develol-men~ o~ di~uihu~ed I~.s. ~Sinee tuture di.s~lihu~ed [~.s will inelu(le eomronen~.s ot leyaey IS.~ an i mu.s~ he eontinuI)u.sly evolved. Ihe envil-onmunt .should al.~o sui~rort Ihe mi~rati(-n.s de.s-nh~d in thi.s i~arer The envil-onlnenl .~houl i .suri~ort Ihe inleyralion ol arhiuary tool.s. It .shouki l rovide a eollcelion ol lo(-l.s lo .su~r,or~ Ihe ie.iiyn. duvelo~menl. le.sliny. del~loymcnl. an i mana~7emcnl ,~t dalah;l.se.s. ainr~lieation.s. and u.ser inl~rtace.s Ihal will he ii.strihuled over a ~al "et en~ n-~nmenl. Be.~ide.s the t()ol.s de.senhed aho~ e. lh~ envirI)nmenl mi~hl al.so inelude o(?l.~ r ~)rllmal di~lrihulion ot eo(le an(l dala. dalah.l.~e (~I-Iimi/allon. and di.slrihule X. I I ' lihrati~n Plannin~ and ~1ana~ement Thu .~mr)~ IS ll-al e.~ l.s durin, mi~rallol1 h~ m~!ru .~)mrlex Ihan il.s ~)mronenl le~ae~, ~r lar~el l.S~ l-hi.~ eomrl.lxil~ re~luires ear-lul mana~T.menl due lo lh. mis.sion erilieal n;~ r~ ~t lh. li~ an~ 1~! e~l ol .n-rol-~. The 1~ - mi~1rali,nl eh.lllen~u.s invol~e rlannin~ and malla~im~ Ih. rmeu~ Th.~. melude: .~ele-lill~ Ih.~ inel-emenl.s 1() he micn-;lled. in~rleavin~ Ih.~ lar~ in~ n~ien~ ~ur~ Ind .se~luen.im~ lh.~ le~.s l~ dimini.~h ris~ ~)1 lailure. The.se are eun-unll~ .nlil-~l~- inlulllve deei.~ioll.~ Ih. man~ dill~renl inlertaee.s. al-rlieallon.s. and d.ltaha.~e mi~raliolls rroe.ed in ral-;lllel and are eut-over in .sla~e.~ v~ith h(-~h lecaey, and ~ar~1et ~.r~i~!n~. lherl, will he many ~er~ion.s lo manace. ~licratinn plannin~ tor TPS and C.I~IS. ar~ r~ ~omrlux ploC.~S~.S ~ha~ mu.st h~ mana,~(i. sinc~ Ih~v ~r~ holh tor mi.s.sion crllical. op~r.ltional l.S.s. C~i.S and TPS ar~ hoth ~o compl~x that no on~ p~r.son or .small croup ot p~ opl~ und~r.st;ln(l.s th~ ~ntir~ .sy.st~m. Du~ to th~ir cntical natur~, .som~ ;l.sp~ct.s ol mi~ration planninc and manaC~m~nt .should h~ automat~d. ~ h~ v~ that th~ inlra.structur~ lor tulur~ IS.s mu.st support con~inuou.s micration (i.~ volution). H~nc~, .support lor ~volution (call~d mlcration h~r~) planninc. manaC~m~nt. and d~v~lopm~nt .. S~ction ~.1()~ must h~com~ an int~cral part ~t tho.s~ inlra.structur~s. S() ** No page found ** ~. ~O~CL~SIO~'S A~'D EPILO(~I.E Future IS technoloay should support continuous, iterative evolution. 1~ mi~ration i.s not I process wilh a heainninc, middle, and end. Continuous chan~e is inevitahle. Current re~uirements rapidly hecome outdated, and luture re~luirement~s cannot he anticipated. The primary challen~e lacinc IS and related suprort technolocy is the ability to accommodate chancae (e ~cJ ~ in re~uirement.s. operation, content, tunction, intertaces, and encineenna). The incremental and iterative evolution ot IS. as well as all other .sottware. is heinc ~onsidered in many related area.s such a.s sottware systems evolutic)n [HUFF'~I specitications, prototypinC [G()LD'~], and oh ject-oriented de.sicn and development. I.S evolution and miaration mu.st he considered to he a dominate tactor in IS lite cycles. When ISs cost hundreds of mlllions ot dollar.s and are mission critical for business processes that have values that are orders of macnitude creater, it is sheer folly to do otherwise. An approrriately de.sicaned tarC~et IS architecture can facilitate tuture evolution and int~cration with oth~r IS.s. It the tar~et IS in nol aprrornal~ly d~.sicn~d. it will h~com~ a next ,aeneralion lecacy I.S lo he miaraled al addilion;ll cost. ()ne of lhe careale.~ onlrihulions ot dalahase le-hnoloay [SILB'~11 i.s data indeF-enden~e, the coal of in.sulatin~ any chance.s to the dataha.se or to the arplications trom each other (i.e.. ahllity to chance one wilhoul alle-tinca lhe olher). We mu.st now exlentl da[a ind~pendencc lo olher asrecls ot ISs. The miaralion melhod.s pre.senled in Ihi.s r aper ha~ve addl-e.ssed al~pli~ation.s and interlace.i in addition to dataha.ses. The ~aoals in ;,-.~se areas, which corre.srond to Jata inde~endence, could he called arrlication independence. u.s~r interface inderendence, and I.S interlace indel-endence. R.~sear~h i.s re~luire(l into ~ore technolo~ie.s (e c - DBMS.s. oreratin~ sy.stem.s. rrocramminC lancua~es. oh ject- ori~ntation. ~i~si~n an~'t duv~lorm~nt m~thods and unvironm.nts! t~ achieve the.s~ ~oals. Re.s~arch i.s al.s() r.~uir~d into .sy.st~m.s archit~tur~ that will in.sulate all a.srec[.s ol th~ s~.st~m.s irom ~ach oth~r .so that th~y c;ln hu moditi~d ind~p~nd~ntly to me~t ~v~r ~hanrin~ r~luir~m~nt.~ and t(1 tal~ advanta~1~ ot n~w t~hnol()~ .s. Thi.s ~oal un terli~s a~rlica~ion ar~hil~ctur~s ~ , modulari[Y. ohj~c~-ori~nt;ltion) and n~t (~n~rati()n c~mnutin~ archit~ctur~.s. suuh a.~ middl~w;~ BRoD')~1. In Ihi.~ I-ap~l. w~ al.~ att~mrtin~ to contrihut~ to th.~ ~uprort ~t c~ntinuou.s ~volution. ~'~ h;~ r(-ro.~-d th.' Cn;~ n Littl~ .slrat~ y and a ~p~.trum ~t .~urp(~rtin~ m~thod.~ with wtlich t() mi~ratu l~acy IS.~ into advanc~d tar~t ~n~ironm~nt.s. V~,~ illu.strat~d ~h.~ m~lho(i.s hv m.~ans ~1t ~w(~ mh~ra~ion ~as~ udi~s ol multi-milli(m dollar. mis.sion critica ac~ I.S.~,. B~)th miy~ thln~ al~ tollouin~y th~ rroros~ i rlan.~ hut ar~ rro~yr~.s.slny at a v~r~ ratu. .~t~ arrrnxn-1a~ul~ ,)n~ y.;lr. h()th mi~ration~ ar. .~till in ~h~ (Jutail~d rlanniny rha.~ ot th~ tlls~ ml~ralh~n Thi~ iilu~[rat~ thu ~xt~nt ~o uhi~h ~u.h sBynlti~;ln~ ~hany~ on~rmny ml.~ n .r~ al I.i~ al.~ ~nlv rartly t~hni~al Buin~ ru~ron~ihl~ )r th.~ mi.s.si(-n ~riti~l hoo~ t Ih. h~n~ ~r th.~ h~)ok.s ot th~ lulurhon~ mramvy m~ .s mana(1~m~nl v~ry .au~iou~. Thu ruali~ ha~ ~h~ miyralion mu.s~ h~ nlially ri.~k tr~ tl~liv~ in a~hi~in~ oal~. and rav ha~l~ in Ih~ hort lurm. Th.~r~ al~ al.~o r.sv~holo~ al and .si~ill ham~rs u-hun m~ my trom a ~v~li un~i~r.sl( o i lo a l~ un~ rsloo(l ~n~ironm~nl. T~ ~on~rihulion ot Ihi.~ rar~r i.i a hiyhly tl~xihl~ t ot mlyration m~lhod.s ~ha~ i.s ~ailorahl~ lo mo.sl l~a~v l.~.s an i hu.sin~ .s ~on~x~.s. Ea~h m~ho i ~on.~i~t~ ot tiv~ hasi~ .st~r.~: It~rativ~ly miyrat~ th~ ~omrutlny ~nvlronm~nt. It~raliv~ly miyral~ th~ l~ya~y arrli~ation.s. ** No page found ** -~C l~O~ LEDGE~lE~T~ Th~ au~h(?r.~i ackm)~ contrihu~ion.s to this work lrom a numh~r ~)~ p~orl~. m~ludin~ Bla~n~ ~1arin~. G~or~ Kocur, Laur~tt~ Bradl~v, and ~lark Horni~k: and th~ tollowin~ partlcip;ln~.i in ~h. C~1~S mi~ration .study: Rich D~lia. Jaim~ Carhon~ll. Carn~ M~llon U niv~r.sity; Paull Hawthorn, Miro Sy.st~m.s, Inc.: Lawr~nc~ Markosian, R~a.sonin ~vst~ms; ~Sai ~tolto. Columhia ~niv~r.sity; Roh~rt Wil~n.sky, Univ~r.sity ol Calitornia; an(l John Davis. And~r.son Con.sul~insn 5 A wor.sc archilcclurc tor micration purpo~cs i.s a semi-decomposable legac~ IS. illustrated m Ficurc l.'. In comparison wilh a dccomposahlc lcc~ac~ IS, onlv u.scr intcrlac~s, Ulj, and systcm intcrlaces, SIj, arc .scparat~ modules. Thc arplication.s and datahas~ serlvicc arc not scparahl~ sincc their structur~ is morc complex. not adc~uat~lv en~incercd in accordance with currcnl standards, or is poorly und~rstood. The lack ~;t ir.lhl~ structurc makc.s analYsis and mi~ration morc complcx and crror rronc. IS IS End End User User U1 J. . .( Un Legacy Applications and Database Service Legacy Data Fi 1urc 1.': .Scmi-Dccomro.sahlc l ccacy l.S Architeclurc Thc wor.sl archil.cturc tor mb~ration i.s a non-decomposable legacv IS. illustrat~d in Fi~urc l.~. Such l.S.~ arc, trom a .~yslcm l~oinl ot vicw, hlack hoxcs .sinc~ no lunctional comroncnt.s ar~ sc~;lrahlc. End l'.s~rs and l.Ss inlcract dircctly with onc, arparcntl~ n ~ v 1.5 ISs End Users Legacy Interfaces, Applications, and Database Service Legacy Data Fi~urc 1.~: ~on-Decomrosahlc ll~acv IS Archit~cture In ccneral. th~ architcctur~ ol mo.sl l~cacy I.S.s may he neither .strictly dccomposahle, semi- decomposahle, nor non-dccoml-osahle Durinc jLS dccadc lonc evolution. a le~caey IS may i ~ILB~J I .S~EEX' I I Vv'.~C~ ~,llh~r~.ha~/. A., M. Sl~n~hrak~r. ;3nd J. ~ilman ~E.l.~, I -L): ~Iha~ m. A.hi~ m~nl.~ and ()~r~ortunili~.s', C~)mmuni~a~n~n.~, ~,~ I() ( h ~ohur l'~ t t~ .SI;..n. D.. I~'on-Blo~kin~ Commit Proto~ ''. Prf~. IY$~ ACil~l SlG.ll()D t~-tn~ nM(ln(~ mfntofD~lt(~ rlando. Fla..Jun. i'~X~ Wa~ht~r, H.. and R~ul~r, A., 'Th~ Contra~t Mod~l". in Tr(ln.sll~tion 1~ ls t(~r A~ In( f~l D~lt(~ .s f~ Appli( ~Itif~n~ Mor~an-Kaulmann. l '~ ~h