Вы находитесь на странице: 1из 10

JuIy 2011

Master of Computer AppIication (MCA) - Semester 4


MC0077 - Advanced Database Systems- 4 Credits
(Book ID: B0882)
Ass|gnment Set 1 (60 Marks)

1. . Describe the foIIowing:


4DimensionaI ModeI
4Object Database ModeIs
4Post - ReIationaI Database ModeIs

Ans: 1he dlmenslonal model ls a speclallzed adapLaLlon of Lhe relaLlonal model used Lo represenL daLa
ln
daLa warehouses ln a way LhaL daLa can be easlly summarlzed uslng CLA querles ln Lhe dlmenslonal
model a daLabase conslsLs of a slngle large Lable of facLs LhaL are descrlbed uslng dlmenslons and
measures A dlmenslon provldes Lhe conLexL of a facL (such as who parLlclpaLed when and where lL
happened and lLs Lype) and ls used ln querles Lo group relaLed facLs LogeLher ulmenslons Lend Lo be
dlscreLe and are ofLen hlerarchlcal for example Lhe locaLlon mlghL lnclude Lhe bulldlng sLaLe and
counLry A measure ls a quanLlLy descrlblng Lhe facL such as revenue lL's lmporLanL LhaL measures
can be meanlngfully aggregaLed for example Lhe revenue from dlfferenL locaLlons can be added
LogeLher
ln an CLA query dlmenslons are chosen and Lhe facLs are grouped and added LogeLher Lo creaLe a
summary
1he dlmenslonal model ls ofLen lmplemenLed on Lop of Lhe relaLlonal model uslng a sLar schema
conslsLlng of one Lable conLalnlng Lhe facLs and surroundlng Lables conLalnlng Lhe dlmenslons
arLlcularly compllcaLed dlmenslons mlghL be represenLed uslng mulLlple Lables resulLlng ln a
snowflake schema
A daLa warehouse can conLaln mulLlple sLar schemas LhaL share dlmenslon Lables allowlng Lhem Lo be
used LogeLher Comlng up wlLh a sLandard seL of dlmenslons ls an lmporLanL parL of dlmenslonal
modellng
5Object Database Models
Ans: ln recenL years Lhe ob[ecLorlenLed paradlgm has been applled Lo daLabase Lechnology creaLlng a
new programmlng model known as ob[ecL daLabases 1hese daLabases aLLempL Lo brlng Lhe daLabase
world and Lhe appllcaLlon programmlng world closer LogeLher ln parLlcular by ensurlng LhaL Lhe
daLabase uses Lhe same Lype sysLem as Lhe appllcaLlon program 1hls alms Lo avold Lhe overhead
(someLlmes referred Lo as Lhe lmpedance mlsmaLch) of converLlng lnformaLlon beLween lLs
represenLaLlon ln Lhe daLabase (for example as rows ln Lables) and lLs represenLaLlon ln Lhe
appllcaLlon program (Lyplcally as ob[ecLs) AL Lhe same Llme ob[ecL daLabases aLLempL Lo lnLroduce
Lhe key ldeas of ob[ecL programmlng such as encapsulaLlon and polymorphlsm lnLo Lhe world of
daLabases
A varleLy of Lhese ways have been Lrled for sLorlng ob[ecLs ln a daLabase Some producLs have
approached Lhe problem from Lhe appllcaLlon programmlng end by maklng Lhe ob[ecLs manlpulaLed
by Lhe program perslsLenL 1hls also Lyplcally requlres Lhe addlLlon of some klnd of query language
slnce convenLlonal programmlng languages do noL have Lhe ablllLy Lo flnd ob[ecLs based on Lhelr
lnformaLlon conLenL CLhers have aLLacked Lhe problem from Lhe daLabase end by deflnlng an
ob[ecLorlenLed
daLa model for Lhe daLabase and deflnlng a daLabase programmlng language LhaL allows full
programmlng capablllLles as well as LradlLlonal query faclllLles
Cb[ecL daLabases suffered because of a lack of sLandardlzaLlon alLhough sLandards were deflned by
CuMC Lhey were never lmplemenLed well enough Lo ensure lnLeroperablllLy beLween producLs
neverLheless ob[ecL daLabases have been used successfully ln many appllcaLlons usually speclallzed
appllcaLlons such as englneerlng daLabases or molecular blology daLabases raLher Lhan malnsLream
commerclal daLa processlng Powever ob[ecL daLabase ldeas were plcked up by Lhe relaLlonal
vendors and lnfluenced exLenslons made Lo Lhese producLs and lndeed Lo Lhe SCL language
5!ost - Relational Database Models
Ans: Several producLs have been ldenLlfled as posLrelaLlonal because Lhe daLa model lncorporaLes
relaLlons buL ls noL consLralned by Lhe lnformaLlon rlnclple requlrlng LhaL all lnformaLlon ls
represenLed by daLa values ln relaLlons roducLs uslng a posLrelaLlonal daLa model Lyplcally employ a
model LhaL acLually predaLes Lhe relaLlonal model 1hese mlghL be ldenLlfled as a dlrecLed graph wlLh
Lrees on Lhe nodes
osLrelaLlonal daLabases could be consldered a subseL of ob[ecL daLabases as Lhere ls no need for
ob[ecLrelaLlonal mapplng when uslng a posLrelaLlonal daLa model ln splLe of many aLLacks on Lhls
class of daLa models wlLh deslgnaLlons of belng hlerarchlcal or legacy Lhe posLrelaLlonal daLabase
lndusLry conLlnues Lo grow as a mulLlbllllon dollar lndusLry even lf Lhe growLh sLays below Lhe
relaLlonal daLabase radar
Lxamples of models LhaL could be classlfled as posLrelaLlonal are lCk aka MulLlvalue and MuMS
aka M

Lxp|a|n the concept of a uery? now a uery Cpt|m|zer works


Ans: Cuerles are essenLlally powerful lllLers Cuerles allow you Lo declde whaL flelds or expresslons are
Lo
be shown and whaL lnformaLlon Lo be soughL Cuerles are usually based on 1ables buL can also be
based on an exlsLlng Cuery Cuerles allow you seek from very baslc lnformaLlon Lhrough Lo much
more compllcaLed speclflcaLlons 1hey also allow you Lo llsL lnformaLlon ln a parLlcular order such as
llsLlng all Lhe resulLlng records ln Surname order for example
Cuerles can selecL records LhaL flL cerLaln crlLerla lf you had a llsL of people and had a gender fleld
you could use a query Lo selecL [usL Lhe males or females ln Lhe daLabase 1he gender fleld would have
a crlLerla seL as male whlch means LhaL when Lhe query ls run only records wlLh male ln Lhe Cender
fleld would be llsLed lor each record LhaL meeLs Lhe crlLerla Lhe you could choose Lo llsL oLherflelds LhaL
may be ln Lhe Lable llke flrsL name surname phone number daLe of blrLh or whaLever youmay have ln
Lhe daLabase
Cuerles can do much more Lhan [usL llsLlng ouL records lL ls also posslble Lo llsL ouL LoLals averages
eLc from Lhe daLa and do varlous oLher calculaLlons Cuerles can also be used Lo do oLher Lasks such
as deleLlng records updaLlng records addlng new records creaLlng new Lables and creaLlng LabulaLed
reporLs
1he 6:ery optimizer ls Lhe componenL of a daLabase managemenL sysLem LhaL aLLempLs Lo deLermlne
Lhe mosL efflclenL way Lo execuLe a query 1he opLlmlzer conslders Lhe posslble query plans for a
glven lnpuL query and aLLempLs Lo deLermlne whlch of Lhose plans wlll be Lhe mosL efflclenL CosLbased
query opLlmlzers asslgn an esLlmaLed cosL Lo each posslble query plan and choose Lhe plan
wlLh Lhe smallesL cosL CosLs are used Lo esLlmaLe Lhe runLlme cosL of evaluaLlng Lhe query ln Lerms
of Lhe number of l/C operaLlons requlred Lhe Cu requlremenLs and oLher facLors deLermlned from
Lhe daLa dlcLlonary 1he seL of query plans examlned ls formed by examlnlng Lhe posslble access paLhs
(eg lndex scan sequenLlal scan) and [oln algorlLhms (eg sorLmerge [oln hash [oln nesLed loop
[oln) 1he search space can become qulLe large dependlng on Lhe complexlLy of Lhe SCL query
Cenerally Lhe query opLlmlzer cannoL be accessed dlrecLly by users once querles are submlLLed Lo
daLabase server and parsed by Lhe parser Lhey are Lhen passed Lo Lhe query opLlmlzer where
opLlmlzaLlon occurs Powever some daLabase englnes allow guldlng Lhe query opLlmlzer wlLh hlnLs
MosL query opLlmlzers represenL query plans as a Lree of plan nodes A plan node encapsulaLes a
slngle operaLlon LhaL ls requlred Lo execuLe Lhe query 1he nodes are arranged as a Lree ln whlch
lnLermedlaLe resulLs flow from Lhe boLLom of Lhe Lree Lo Lhe Lop Lach node has zero or more chlld
nodesLhose are nodes whose ouLpuL ls fed as lnpuL Lo Lhe parenL node lor example a [oln node wlll
have Lwo chlld nodes whlch represenL Lhe Lwo [oln operands whereas a sorL node would have a
slngle chlld node (Lhe lnpuL Lo be sorLed) 1he leaves of Lhe Lree are nodes whlch produce resulLs by
scannlng Lhe dlsk for example by performlng an lndex scan or a sequenLlal scan
1oin ordering
1he performance of a query plan ls deLermlned largely by Lhe order ln whlch Lhe Lables are [olned lor
example when [olnlng 3 Lables A 8 C of slze 10 rows 10000 rows and 1000000 rows respecLlvely
a query plan LhaL [olns 8 and C flrsL can Lake several ordersofmagnlLude more Llme Lo execuLe Lhan
one LhaL [olns A and C flrsL MosL query opLlmlzers deLermlne [oln order vla a dynamlc programmlng
algorlLhm ploneered by l8Ms SysLem 8 daLabase pro[ecL 1hls algorlLhm works ln Lwo sLages
1 llrsL all ways Lo access each relaLlon ln Lhe query are compuLed Lvery relaLlon ln Lhe query can
be accessed vla a sequenLlal scan lf Lhere ls an lndex on a relaLlon LhaL can be used Lo answer a
predlcaLe ln Lhe query an lndex scan can also be used lor each relaLlon Lhe opLlmlzer recordsLhe
cheapesL way Lo scan Lhe relaLlon as well as Lhe cheapesL way Lo scan Lhe relaLlon LhaL
produces records ln a parLlcular sorLed order2
2 1he opLlmlzer Lhen conslders comblnlng each palr of relaLlons for whlch a [oln condlLlon exlsLs
lor each palr Lhe opLlmlzer wlll conslder Lhe avallable [oln algorlLhms lmplemenLed by Lhe u8MS
lL wlll preserve Lhe cheapesL way Lo [oln each palr of relaLlons ln addlLlon Lo Lhe cheapesL way Lo
[oln each palr of relaLlons LhaL produces lLs ouLpuL accordlng Lo a parLlcular sorL order
3 1hen all LhreerelaLlon query plans are compuLed by [olnlng each LworelaLlon plan produced by
Lhe prevlous phase wlLh Lhe remalnlng relaLlons ln Lhe query
ln Lhls manner a query plan ls evenLually produced LhaL [olns all Lhe querles ln Lhe relaLlon noLe LhaL
Lhe algorlLhm keeps Lrack of Lhe sorL order of Lhe resulL seL produced by a query plan also called an
lnLeresLlng order uurlng dynamlc programmlng one query plan ls consldered Lo beaL anoLher query
plan LhaL produces Lhe same resulL only lf Lhey produce Lhe same sorL order 1hls ls done for Lwo
reasons llrsL a parLlcular sorL order can avold a redundanL sorL operaLlon laLer on ln processlng Lhe
query Second a parLlcular sorL order can speed up a subsequenL [oln because lL clusLers Lhe daLa ln a
parLlcular way
PlsLorlcally SysLem8 derlved query opLlmlzers would ofLen only conslder lefLdeep query plans
whlch flrsL [oln Lwo base Lables LogeLher Lhen [oln Lhe lnLermedlaLe resulL wlLh anoLher base Lable
and so on 1hls heurlsLlc reduces Lhe number of plans LhaL need Lo be consldered (n!lnsLead of 4n)
buL may resulL ln noL conslderlng Lhe opLlmal query plan 1hls heurlsLlc ls drawn from Lhe observaLlon
LhaL [oln algorlLhms such as nesLed loops only requlre a slngle Luple (aka row) of Lhe ouLer relaLlon aL
a Llme 1herefore a lefLdeep query plan means LhaL fewer Luples need Lo be held ln memory aL any
Llme Lhe ouLer relaLlons [oln plan need only be execuLed unLll a slngle Luple ls produced and Lhen
Lhe lnner base relaLlon can be scanned (Lhls Lechnlque ls called plpellnlng)
SubsequenL query opLlmlzers have expanded Lhls plan space Lo conslder bushy query plans where
boLh operands Lo a [oln operaLor could be lnLermedlaLe resulLs from oLher [olns Such bushy plans are
especlally lmporLanL ln parallel compuLers because Lhey allow dlfferenL porLlons of Lhe plan Lo be
evaluaLed lndependenLly

. ExpIain the foIIowing with respect to Heuristics of Query Optimizations:


A) EquivaIence of Expressions B) SeIection Operation
) ro[ect|on Cperat|on D) Natura| Io|n Cperat|on
Ans: E6:ivalence of Expressions: 1he flrsL sLep ln selecLlng a queryprocesslng sLraLegy ls Lo flnd a
relaLlonal algebra expresslon LhaL ls equlvalenL Lo Lhe glven query and ls efflclenL Lo execuLe
We'll use Lhe followlng relaLlons as examples
CusLomerscheme (cname sLreeL cclLy)
ueposlLscheme (bname accounL# name balance)
8ranchscheme (bname asseLs bclLy)
Selection Operation:
1 Conslder Lhe query Lo flnd Lhe asseLs and branchnames of all banks who have deposlLors llvlng ln
orL ChesLer ln relaLlonal algebra Lhls ls
H bname asseLs(o cclLy"orL ChesLer"
(cusLomer deposlL branch))
1hls expresslon consLrucLs a huge relaLlon
cusLomer deposlL branch of whlch we are only lnLeresLed ln a few Luples
We also are only lnLeresLed ln Lwo aLLrlbuLes of Lhls relaLlon
We can see LhaL we only wanL Luples for whlch cclLy orL ChesLer"
1hus we can rewrlLe our query as
bname asseLs(occlLy"orL ChesLer"(cusLomer))
cusLomer deposlL branch)
1hls should conslderably reduce Lhe slze of Lhe lnLermedlaLe relaLlon
S:ggested R:le for Optimization:
erform selecL operaLlons as early as posslble
lf our orlglnal query was resLrlcLed furLher Lo cusLomers wlLh a balance over $1000 Lhe selecLlon
cannoL be done dlrecLly Lo Lhe cusLomer relaLlon above
1he new relaLlonal algebra query ls

1he selecLlon cannoL be applled Lo cusLomer as balance ls an aLLrlbuLe of deposlL

We can still rewrite as
lf we look furLher aL Lhe subquery (mlddle Lwo llnes above) we can spllL Lhe selecLlon predlcaLe ln
Lwo
1hls rewrlLlng glves us a chance Lo use our perform selecLlons early" rule agaln
We can now rewrlLe our subquery as
Second Transformational R:le:
8eplace expresslons of Lhe form o 12(C) by o 1(o 2( C)) where 1 and 2 predlcaLes and e ls
a relaLlonal algebra expresslon
Cenerally
P1( P2( C)) P2( P1( C)) P1`P2(C)
5!rojection Operation
Ans: !rojection Operation
1 Llke selecLlon pro[ecLlon reduces Lhe slze of relaLlons
lL ls advanLageous Lo apply pro[ecLlons early Conslder Lhls form of our example query
2 When we compuLe Lhe subexpresslon
We obLaln a relaLlon whose scheme ls (cname cclLy bname accounL# balance)
3 We can ellmlnaLe several aLLrlbuLes from Lhls scheme 1he only ones we need Lo reLaln are Lhose
LhaL
*appear ln Lhe resulL of Lhe query or
*are needed Lo process subsequenL operaLlons
4 8y ellmlnaLlng unneeded aLLrlbuLes we reduce Lhe number of columns of Lhe lnLermedlaLe resulL
and Lhus lLs slze
3 ln our example Lhe only aLLrlbuLe we need ls bname (Lo [oln wlLh branch) So we can rewrlLe our
expresslon as
noLe LhaL Lhere ls no advanLage ln dolng an early pro[ecL on a relaLlon before lL ls needed for some
oLher operaLlon*We would access every block for Lhe relaLlon Lo remove aLLrlbuLes
*1hen we access every block of Lhe reducedslze relaLlon when lL ls acLually needed
*We do more work ln LoLal raLher Lhan less!
at:ral 1oin Operation
Ans: AnoLher way Lo reduce Lhe slze of Lemporary resulLs ls Lo choose an opLlmal orderlng of Lhe [oln
operaLlons
naLural [oln ls assoclaLlve
AlLhough Lhese expresslons are equlvalenL Lhe cosLs of compuLlng Lhem may dlffer
Look agaln aL our expresslon
We see LhaL we can compuLe deposlL branch flrsL and Lhen [oln wlLh Lhe flrsL parL
Powever deposlL branch ls llkely Lo be a large relaLlon as lL conLalns one Luple for every accounL
1he oLher parL ls probably a small relaLlon (comparaLlvely)
So lf we compuLe flrsL we geL a reasonably small relaLlon
lL has one Luple for each accounL held by a resldenL of orL ChesLer
1hls Lemporary relaLlon ls much smaller Lhan deposlL branch
naLural [oln ls commuLaLlve
1hus we could rewrlLe our relaLlonal algebra expresslon as
8uL Lhere are no common aLLrlbuLes beLween cusLomer and branch so Lhls ls a artesian prod:ct
LoLs of Luples!
lf a user enLered Lhls expresslon we would wanL Lo use Lhe assoclaLlvely and commuLaLlvely of
naLural [oln Lo Lransform Lhls lnLo Lhe more efflclenL expresslon we have derlved earller (join with
deposit first, then with branch)

4 1here are a number of h|stor|ca| organ|zat|ona| and techno|og|ca| reasons exp|a|n the |ack
of an a||encompass|ng data management system D|scuss few of them w|th appropr|ate
examp|es

Ans: MosL currenL daLa managemenL sysLems uMS have been bullL on Lhe assumpLlon LhaL Lhe daLa
collecLlon or daLabase Lo be admlnlsLered conslsLs of a slngle medla Lype sLrucLured Lables of facL
daLa or unsLrucLured sLrlngs of blLs represenLlng such medla ob[ecLs as LexL documenLs lmages or
vldeo 1he resulL ls LhaL mosL uMS sLore and lndex a speclflc Lype of medla daLa and provlde a query
(daLa access) language LhaL ls speclallzed for efflclenL access Lo and reLrleval of Lhls daLa Lype
A furLher assumpLlon LhaL has frequenLly been made ls LhaL Lhe lnformaLlon requlremenLs of Lhe
sysLem users are known and can be used for sLrucLurlng Lhe daLa collecLlon and Lunlng Lhe daLa
managemenL sysLem lL has also been assumed LhaL Lhe users would only lnfrequenLly requlre
lnformaLlon/daLa from some oLher Lype of daLa managemenL sysLem
1hese assumpLlons have been crlLlclzed slnce Lhe early 1980s by researchers who have polnLed ouL
LhaL almosL from Lhe polnL of creaLlon a daLabase would noL (nor could) conLaln all of Lhe daLa
requlred by Lhe user communlLy (Cllgor Luckenbaugh 1984 Landers 8osenberg1982 LlLwln eL
al 1982 among many oLhers) A number of hlsLorlcal organlzaLlonal and Lechnologlcal reasons
explaln Lhe lack of an allencompasslng daLa managemenL sysLem Among Lhese are
1he senslble advlce Lo bulld small sysLems wlLh Lhe plan Lo exLend Lhelr scope ln laLer
lmplemenLaLlon phases allows a core sysLem Lo be lmplemenLed relaLlvely qulckly buL has lead
Lo a prollferaLlon of relaLlvely small sysLems
ueparLmenL auLonomy has led Lo consLrucLlon of deparLmenL speclflc raLher Lhan organlzaLlon
wlde sysLems agaln leadlng Lo many small overlapplng and ofLen lncompaLlble sysLems wlLhln
an organlzaLlon
1he conLlnual evoluLlon of Lhe organlzaLlon and lLs lnLeracLlons boLh wlLhln and Lo lLs exLernal
envlronmenL prohlblLs compleLe undersLandlng of fuLure lnformaLlon requlremenLs
arallel developmenL of daLa managemenL sysLems for parLlcular appllcaLlons has lead Lo
dlfferenL and lncompaLlble sysLems for managemenL of Labular/admlnlsLraLlve daLa
LexL/documenL daLa hlsLorlcal/sLaLlsLlcal daLa spaLlal/geographlc daLa and sLreamed/audlo and
vlsual daLa
1he resulL ls LhaL only a porLlon of an organlzaLlons daLa ls admlnlsLered by any one daLa
managemenL sysLem and mosL organlzaLlons have a mulLlLude of speclal purpose daLabases managed
by dlfferenL and ofLen lncompaLlble daLa managemenL sysLem Lypes 1he growlng need LoreLrleve
daLa from mulLlple daLabases wlLhln an organlzaLlon as well as Lhe rapld dlssemlnaLlon of daLa
Lhrough Lhe lnLerneL has glven rlse Lo Lhe requlremenL of provldlng lnLegraLed access Lo boLh lnLernal
and exLernal daLa of mulLlple Lypes
A ma[or challenge and crlLlcal pracLlcal and research problem for Lhe lnformaLlon compuLer and
communlcaLlon Lechnology communlLles ls Lo develop daLa managemenL sysLems LhaL can
provldeefflclenL access Lo Lhe daLa sLored ln mulLlple prlvaLe and publlc daLabases (8rodle1993 Purson

8rlghL 1996 nordboLLen 1988a 1988b and nordboLLen 1994a)
roblems Lo be resolved lnclude
1 lnLeroperablllLy among sysLems (lox Sornll 1999 LlwLln AbdellaLlf 1986)
2 lncorporaLlon of legacy sysLems (8rodle 1993) and
3 lnLegraLlon of managemenL Lechnlques for sLrucLured and unsLrucLured daLa (SLonebraker
8rown 1999)
Lach of Lhe above problems enLalls an lnLegraLlon of concepLs meLhods Lechnlques and Lools from
separaLe research and developmenL communlLles LhaL have exlsLed ln parallel buL lndependenLly and
have had raLher mlnlmal lnLeracLlon Cne consequence of whlch ls LhaL Lhere exlsLs an overlapplng
and confllcLlng Lermlnology beLween Lhese communlLles
ln Lhe prevlous chapLer a daLabase was deflned as a CCLLLC1lCn Cl 8LLA1Lu uA1A 8L8LSLn1lnC
SCML LCClCALL? CCPL8Ln1 ASLC1 Cl1PL 8LAL WC8Lu WlLh Lhls deflnlLlon nC llmlLaLlons are
glven as Lo Lhe Lype of
uaLa ln Lhe collecLlon
Model used Lo sLrucLure Lhe collecLlon or
ArchlLecLure and geographlc locaLlon of Lhe daLabase
1he focus of Lhls LexL ls on onllne elecLronlc and web accesslble daLabases conLalnlng mulLlple
medla daLa Lhus resLrlcLlng our lnLeresL/focus Lo mulLlmedla daLabases sLored on one or more
compuLers (u8 servers) and accesslble from Lhe lnLerneL Lxamples of such daLabases lnclude Lhe
lmage collecLlons of Lhe PermlLage Museum Lhe caLalog and full LexL maLerlals of Lhe ACM dlglLal
llbrary and Lhe cusLomer records for Lhe 7 slLes of Amazoncom
LlecLronlc daLabases are lmporLanL slnce Lhey conLaln daLa recordlng Lhe producLs and servlces as
well as Lhe economlc hlsLory and currenL sLaLus of Lhe owner organlzaLlon 1hey are also a source of
lnformaLlon for Lhe organlzaLlons employees and cusLomers/users Powever daLabases cannoL be
used effecLlvely unless Lhere exlsL efflclenL and secure daLa managemenL sysLems uMS for Lhe daLa ln
Lhe daLabases

S Descr|be the Structura| Semant|c Data Mode| (SSM) w|th re|evant examp|es
Ans: Modellng Complex and MulLlmedla uaLa
uaLa modellng addresses a need ln lnformaLlon sysLem analysls and deslgn Lo develop a model of Lhe
lnformaLlon requlremenLs as well as asseL of vlable daLabase sLrucLure proposals 1he daLa modellng
process conslsLs of
1 ldenLlfylng and descrlblng Lhe lnformaLlon requlremenLs for an lnformaLlon sysLem
2 Speclfylng Lhe daLa Lo be malnLalned by Lhe daLa managemenL sysLem and
3 Speclfylng Lhe daLa sLrucLures Lo be used for daLa sLorage LhaL besL supporL Lhe lnformaLlon
requlremenLs
A fundamenLal Lool used ln Lhls process ls Lhe daLa model whlch ls used boLh for speclflcaLlon of Lhe
lnformaLlon requlremenLs aL Lhe user level and for speclflcaLlon of Lhe daLa sLrucLure for Lhe
daLabase uurlng lmplemenLaLlon of a daLabase Lhe daLa model guldes consLrucLlon of Lhe schema or
daLa caLalog whlch conLalns Lhe meLa daLa LhaL descrlbe Lhe u8 sLrucLure and daLa semanLlcs LhaL are
used Lo supporL daLabase lmplemenLaLlon and daLa reLrleval
uaLa modellng uslng a speclflc daLa model Lype and as a unlque acLlvlLy durlng lnformaLlon sysLem
deslgn ls commonly aLLrlbuLed Lo Charles 8achman (1969) who presenLed Lhe uaLa SLrucLure ulagram
as one of Lhe flrsL wldely used daLa models for neLwork daLabase deslgn Several alLernaLlve daLa
model Lypes were proposed shorLly LhereafLer Lhe besL known of whlch are Lhe
8elaLlonal model (Codd 1970) and Lhe
LnLlLyrelaLlonshlp L8 model (Chen 1976)
1he relaLlonal model was qulckly crlLlclzed for belng flaL ln Lhe sense LhaL all lnformaLlon ls
represenLed as a seL of Lables wlLh aLomlc cell values 1he deflnlLlon of wellformed relaLlonal models
requlres LhaL complex aLLrlbuLe Lypes (hlerarchlc composlLe mulLlvalued and derlved)be converLed
Lo aLomlc aLLrlbuLes and LhaL relaLlons be normallzed lnLerenLlLy (lnLerrelaLlon) relaLlonshlps are
dlfflculL Lo vlsuallze ln Lhe resulLlng seL of relaLlons maklng conLrol of Lhe compleLeness and
correcLness of Lhe model dlfflculL 1he relaLlonal model maps easlly Lo Lhe physlcal characLerlsLlcs of
elecLronlc sLorage medla and as such ls a good Lool for deslgn of Lhe physlcal daLabase
1he enLlLyrelaLlonshlp approach Lo modellng proposed by Chen (1976) had Lwo prlmary ob[ecLlves
flrsL Lo vlsuallze lnLerenLlLy relaLlonshlps and second Lo separaLe Lhe u8 deslgn process lnLo Lwo
phases
1 8ecord ln an L8 model Lhe enLlLles and lnLerenLlLy relaLlonshlps requlred by Lhe enLerprlse
le by Lhe owner/user of Lhe lnformaLlon sysLem or appllcaLlon 1hls phase and lLs resulLlng model
should be lndependenL of Lhe u8MS Lool LhaL ls Lo be used for reallzlng Lhe u8
2 1ranslaLe Lhe L8 model Lo Lhe daLa model supporLed by Lhe u8MS Lo be used for lmplemenLaLlon
1hls Lwophase deslgn supporLs modlflcaLlon aL Lhe physlcal level wlLhouL requlrlng changes Lo Lhe
enLerprlse or user vlew of Lhe u8 conLenL Also Chens L8 model qulckly came under crlLlclsm
parLlcularly for lLs lack of ablllLy Lo model classlflcaLlon sLrucLures ln 1977 (SmlLh SmlLh) presenLed
a meLhod for modellng generallzaLlon and aggregaLlon hlerarchles LhaL underlle Lhe many
exLended/enhanced enLlLyrelaLlonshlp LL8 model Lypes proposed and ln use Loday

6. Describe the foIIowing with respect to Fuzzy querying to reIationaI databases:


4Proposed ModeI
4Meta knowIedge
4ImpIementation
Ans: 1he easlesL way of lnLroduclng fuzzlness ln Lhe daLabase model ls Lo use classlcal relaLlonal
daLabases
and formulaLe a fronL end Lo lL LhaL shall allow fuzzy querylng Lo Lhe daLabase A llmlLaLlon lmposed
on Lhe sysLem ls LhaL because we are noL exLendlng Lhe daLabase model nor are we deflnlng a new
model ln any way Lhe underlylng daLabase model ls crlsp and hence Lhe fuzzlness can only be
lncorporaLed ln Lhe query
1o lncorporaLe fuzzlness we lnLroduce fuzzy seLs / llngulsLlc Lerms on Lhe aLLrlbuLe domalns / llngulsLlc
varlables eg on Lhe aLLrlbuLe domaln ACL we may deflne fuzzy seLs as ?CunC MluuLL and CLu
1hese are deflned as Lhe followlng

lor Lhls we Lake Lhe example of a sLudenL daLabase whlch has a Lable S1uuLn1S wlLh Lhe followlng
aLLrlbuLes


Meta knowledge
Ans: AL Lhe level of meLa knowledge we need Lo add only a slngle Lable LA8LLS wlLh Lhe followlng
sLrucLure
ig


1hls Lable ls used Lo sLore Lhe lnformaLlon of all Lhe fuzzy seLs deflned on all Lhe aLLrlbuLe domalns A
descrlpLlon of each column ln Lhls Lable ls as follows
abel: 1hls ls Lhe prlmary key of Lhls Lable and sLores Lhe llngulsLlc Lerm assoclaLed wlLh Lhe fuzzy
seL
ol:mn_ame: SLores Lhe llngulsLlc varlable assoclaLed wlLh Lhe glven llngulsLlc Lerm
Alpha,Beta, Gamma, Delta: SLores Lhe range of Lhe fuzzy seL
5mplementation
Ans: 1he maln lssue ln Lhe lmplemenLaLlon of Lhls sysLem ls Lhe parslng of Lhe lnpuL fuzzy query As Lhe
underlylng daLabase ls crlsp le no fuzzy daLa ls sLored ln Lhe daLabase Lhe lnSL81 query wlll noL
change and need noL be parsed Lherefore lL can be presenLed Lo Lhe daLabase as lL ls uurlng parslng
Lhe query ls parsed and dlvlded lnLo Lhe followlng
1 Cuery 1ype WheLher Lhe query ls a SLLLC1 uLLL1L or uuA1L
2 8esulL ALLrlbuLes 1he aLLrlbuLes LhaL are Lo be dlsplayed used only ln Lhe case of Lhe SLLLC1
query
3 Source 1ables 1he Lables on whlch Lhe query ls Lo be applled
4 CondlLlons 1he condlLlons LhaL have Lo be speclfled before Lhe operaLlon ls performed lL ls
furLher subdlvlded lnLo Cuery ALLrlbuLes (le Lhe aLLrlbuLes on whlch Lhe query ls Lo be applled)
and Lhe llngulsLlc Lerm lf Lhe condlLlon ls noL fuzzy le lL does noL conLaln a llngulsLlc Lerm Lhen lL
need noL be subdlvlded

Вам также может понравиться