1q LOS 928
AssiuNMENT-
Mapping the data warehou.sS0 to a MuldpveessoY archi tetture.
tRolaHonat database technotogy jor a data warohouse;
The dea 0 o1ganizatior that dovotops
data ttadehouse tr to deat With the increasin9 amounts 0 data
The goat ts to have
*Better pexjoimante
*Scatabitity
Speed up
The abltGty to exetute the Same euest on tho
same amount of data in less Hme
Scate up:
Tte abitity to Obtain Ssame Peafomane on the
Some seonuest even the database Stze Increases.
ineay speed up
cate up
*Jine08 speed up Doubung the number o4
PxDCeSS Tedutes the Yesponm ttme to halg
LnearStaleup-provide the Same peformance
muth daba.
bn twice
TYPes oy Parattettsm
The ven doxs staated taking advantages
Poxattet haa dwat AYchitetture by tmplementing Muttlservex
and Mutttthyoaded system to handle a targe number 04 eoyuests.
Tnte -ouey
*Tntrd auery
tntor quory Pasaltetism
Dijjerent Seave threade O prDess handle
the Same time Suecesul on smp system.
Mutiple Teuests at
Intor Oyuery aatteusm MOA Umited be cau se Mul dple ame nar
Were protesed Contuyrenty but eath aery l a prweed
Sernatly by a singte thsead.
ntra uery Paralhel/m
Thts om of ParatteUsm detompo.103 tho sena
Sat quevy into
into Dw levol opeatioru (eg) scan. Jjoin,otc. then
theye low level opesation ave egetuted tonu1uentty in paratlel
Hortzontat Parattelicm:
Thte database i patittoned aaoss muttiple
dtsHs and paatet pocessin9 occurs With n a Spe ufie tosH.
1) vorticat paralioüsm
OCturs dHtjerent tashs. Ati Guery
among
o1ecuted tn parattel, oi in pipelned.
Component s
Rouponae
Serna
RD8MS
SOte
Por zontal
join Vemitol
paral le
Cone 4
TPes o beMC parale ugm
Data povtlHoning:
0ato poYtitiontn a a koy Yenulremon t jor
e1ottve pavaliet erecuton O database opeaton. Ths 4ptoadu
data o a table across Mutiplo dtus so that [Link] ead,
white ton be pjovmod in Paraltet,
Ways o data parihoning9
Random
Inteltlgent
RondoM paatitioning
This fs done by yandon stnping of data
a
Stng le seJver. Round obin 18
atYOS Muttple dtuks on a
Method Df Tandom parHtioning. n thu Method eath now
the next dsK 0utgned ho the
re tord u Platod on
databaa
Advantage
Redutos 1/0 tontontron
Disadvantog e
Al the pajntloru May have b be jung tanned.
Tntetigent paaHtoning
A n thir Method the DBMJ knows the lotahon
04 a spetugic etotd hente Seanthing Hme h avoidod fulty
extoita araltet architetture.
Advantager
Atlows auOTY Paifoning, even ig a diaK parion 1
unavai table Method 0 intettgent aatiHontn9
Hash pavtitoning
A hash di9othm L4ued v (alcutate the
Paatiton humber bOa Dn the vatue 0 parHboning koy toa
eath row.
1) Key vange parhtioning
Row ae plated and toca ted in tho payin0n
auDrding th vatue 0 arhton Key
A to h in paatiion 4
to Paannbn a et
) sthema aation
The enthe table A plated bn one duk anothes
table on pthe dIA etc. nethod ta utetul ony fo
Smatt 7ejeven ce s
v [Link] deziod paundNontng
The table 19 aa titioned baund upon w
dej ined eipToaion).
Databa te arthitettute oa paaite1 pavesstrg:
There a7e three main DBMS S0jtware
oxchitettu3e.
Ahared everything
Shared dix
Shaved nothing
shaTod architettre tb1) ShaTed everything
Thta Method t the tradinonal apprwach
to im plement and RDBMS DA SMP havdwaTe
*Jt ts stmple to implement and utesstul but
imitaton ia Atalabiuty.
TnterLonnec tion Ne tworK
Yoee to Proce s1or
PYOte soa roeso
Vnt Pe) vnit (p) untt Pu)
hnit PU)
af lbbal shared Memory
Deuign point a44ett Stalabiiby 0 the
axthitetture Among them is
ProRis based imptementarin
Thread bod tmplementa ianrst
thread RDeMS Implements Tts
own thread. Thweod a3 too tght toupie d ajgect
*SMP $ystem ane not joltousing Unaar roth
SMp throughput doea not tneroole When tha nurbes
enteeda a Saturaion number
PDtessorS
shaed dia anchitectuue
shared diun anchitettuTe ho the lontep
the entHe databoa be toren RDBMS
D4 AhaTed DwneTs hip 0j
Mhtth ta unntng on a mode off dbuiburd
Apvvea euch D
perfome d by he DUM
emory Sysrem, The operarton
Mer and opplcaian deve Lopers. Thus
A hudden rom
Provi ding a ingie 9 etem mae
Thi architetture Louryaint on stalabiuty.
a nodos ase yeading
The wOYAE se
Atenao A that 1
data, RDBMS and DuM has to
and updating the &dmo
Yepouce3 o synchonlxarlon. This problem
Shend tot
tatted as pirging. thia pinging haa nogarie impatt on
Atalabiury
Ahared DaK Axchitecture
bottte hect duo to data GKOW
Reduces
Incre aues the system avatlability
Eliminate The bottle heek o sMP $gstom hente
dependeny on data panh onina
educes
Tnterconne ction Notwor
pote ssoy
proteMo procemo
wntt po) unit pu) Unit (pu)
Unit Pu)
wtal Local Local Loat
Membry Memory MeMOry Mamory
qlobal shared Memory
Shaved Nothing Archtoture
Each motos sor has its Dwn memory, dirK
and Cormmunftate
ith dthea ProteADA byechongirg Meuaqe
and doA ovey the hetwoak,
ThA anchttet hure njjess non uno &tatabiü'ty. This ts
the most diggiuit Impemeht due to.
Need a now pogramning Pazadigrn
New 0.S. New Compile1
* New 01 enhanted languo go.
)Suppodt jor JuncHon shipping
have to be Touted
Detomposed 86 statements
0eelod data ,
and executed on the Proe ssor haa tha
9) Paroatie Jotn strategtea
A fotn betwen tuo tablkA lu called a tolocated
bwA to be Jolned 7ulde on Aama pos trmon,
jotn When tuwo
InterLonne ction Ne tworic
pu PU PO
P
Local Lo cal Lotal Local
Memory Memory Memony Memory
a) Redirection Jon
sont to the hode
Rous O one table a3
table destdo os the
Whoro the toreAponding yows 0 anotho
O} both tables d1e sent to thtsd node
hode.
matchtng 30
b) Repayttoned Jotn er
Tte ows 0 both tablo. 08e bepaaritlonod
and thon edirotted to enable Cotactod
on the toln tslumn
jolna
C)Broadtast Jol0
Tte entive table is sent v alt nodoA
3) suppont jof data Teporttlonlng
When ono prDcesco o lotal dian jatls the
DBMS Must be able epnmitlon the darabae acossS
Temaining RMD nede Qurpmaicaly
4) ueny Compltation
thi's s the proeSS 0t Selethng an attoss
ath through out the databaje.
s) suppoit O database tarnacttohd
Payallet diutmbuted Mernory DBMS Jhoutd
ules o4 diz ibuted DBMS
LOmply wlth Many
