TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.7, July 201
4, pp
. 5693 ~ 56
9
8
DOI: 10.115
9
1
/telkomni
ka.
v
12i7.505
0
5693
Re
cei
v
ed
No
vem
ber 4, 20
13; Re
vised
Ma
rch 25, 20
14; Accepted
April 15, 201
4
Partitioning of Oracle Application in Structured
Electronic Med
i
cal Records
Chua
ndi Pan, Jiandong Hu*, Jian Shi
Dep
a
rtment of Comp
uter and
Information Ma
nag
ement, t
he
First Affiliated Hospit
al of W
enzho
u Medic
a
l
Univers
i
t
y
,
W
enzho
u, 325
005, Z
hej
ia
ng
Provinc
e
, T
he
Peop
le’s R
e
p
u
b
lic of Ch
ina
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: hjd1
05
6@1
6
3
.com
A
b
st
r
a
ct
T
h
is pap
er de
scribe
d
how
to use Oracle p
a
rtitio
n
i
ng to a
c
hiev
e classifi
ed storag
e of data fro
m
structured elec
tronic me
dic
a
l records.
B
e
cau
s
e structure
d
e
l
ectron
ic
me
dic
a
l rec
o
rds
stor
ed
massive
dat
a
,
the p
e
rfor
manc
e, security
an
d
stabil
i
ty
of d
a
t
abas
e h
ad
be
e
n
p
a
id
more
an
d
more
attenti
o
ns. In this
pa
p
e
r,
table w
i
th partit
i
on in structure
d
electro
n
ic me
dical
rec
o
rds s
ystem w
a
s pre
s
ented to solv
e
these prob
le
ms.
In add
itio
n, testing of co
mpar
ing
th
e p
e
rfor
ma
nce
of tabl
e w
i
th an
d
w
i
thout p
a
rtitio
n
had
be
en tak
e
n to
prove it w
a
s much faster to retrieve partiti
on
ed t
abl
es w
i
th mass
ive d
a
ta than n
o
t partitio
ned tab
l
es.
Ke
y
w
ords
: ora
c
le, partitio
n
tables, structure
d
electro
n
ic
me
dical r
e
cords,
VLDB
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Electro
n
ic
m
edical record
s is focused
on o
u
tpatie
nt and
ho
sp
italized
patie
nts (or
obje
c
ts), it is
the integration
system of
clini
c
al dia
g
n
o
si
s and
trea
tment of information data [
1
].
Structu
r
ed el
ectro
n
ic me
d
i
cal re
co
rd
s system
ap
pli
e
s structu
r
e
d
method to analyze medi
cal
document
s which a
r
e in
pu
t in natural l
angu
age fro
m
medi
cal in
formation p
o
i
n
t of view, and
save
s those
sema
ntic
stru
cture
s
in the
way of
relati
onal st
ru
cture to databa
se [2]. Structured
electroni
c m
edical record
s
system i
s
rega
rd
ed a
s
the future
of ele
c
troni
c medi
cal
re
cord
s
system, b
e
ca
use
it ha
s m
any me
rits.
Cho
o
si
ng a
p
p
rop
r
iate
dat
a sto
r
ag
e i
s
vital to functi
on,
perfo
rman
ce
and se
cu
rity of structu
r
e
d
electroni
c me
dical record
s
system, be
ca
use of too ma
n
y
data elem
ent
s, com
p
licate
d
stru
cture a
nd hug
e data
size. Th
ro
ug
h partitionin
g
related tabl
e
s
,
retrievin
g
efficien
cy will
be
improve
d
an
d dat
ab
ase will be mo
re
stable [3]. This
pape
r de
scri
b
e
s
how to u
s
e
Ora
c
le pa
rtitioning to a
c
hi
eve cla
ssi
fie
d
stora
ge of d
a
ta from stru
cture
d
ele
c
tronic
medic
a
l re
co
r
d
s.
2. Oracle Par
t
itioning
Partitioning i
s
provided by
Ora
c
le to sup
port VLDB (V
ery Large Da
tabase).a pa
rtitioned
table in
Ora
c
le
co
ntain
s
many pa
rtitio
ns
whi
c
h
ca
n re
sid
e
in
different tabl
e spa
c
e
s
. Each
partition i
s
an individu
al
segm
ent [4]. Partit
ioning is entirely transparent
to applications.
Retrievin
g
da
ta not only
can be i
m
ple
m
ented
by q
uerying t
able
with all
partit
i
ons, b
u
t also
can
be a
c
hieved
by queryin
g
certai
n pa
rtitions
of
table. There are two
sugg
est
s
about ‘Whe
n to
partition a tab
l
e’ from Oracl
e
:
(1) T
able
s
greater than 2
G
B should al
ways be con
s
id
ered fo
r partit
i
oning.
(2) T
able
s
containin
g
historical
data,
in whi
c
h n
e
w
data i
s
a
dded into th
e newest
partition. A typical exa
m
ple
is a histo
r
i
c
al
tabl
e wh
ere
only the cu
rre
n
t month'
s da
ta is upd
atabl
e
and the othe
r 11 months a
r
e read o
n
ly.
While
cre
a
tin
g
partition, p
a
rtition key shoul
d be
ch
ose. Oracl
e
aut
omatically direct
s
inse
rt, upd
ate
,
and d
e
lete
operation
s
to
map
s
dat
a
to the a
pprop
riate pa
rtition
throug
h the
u
s
e
of the partition colum
n
s. O
r
acl
e
provid
e
s
the followi
n
g
partitionin
g
method
s:
(1)
Ran
ge P
a
rtitioning: Range p
a
rtitio
ning map
s
d
a
ta to partitio
n
s ba
se
d on
rang
es
of
partition key values that you esta
blish for ea
ch pa
rtition.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5693 – 56
98
5694
(2) Li
st Partiti
oning:
List
p
a
rtitioning
en
able
s
you to
explicitly cont
rol how rows map to
partition
s by
spe
c
ifying a li
st of discrete
values
fo
r the
partitionin
g
key in the de
scriptio
n for e
a
c
h
partition.
(3) Hash Part
itioning: Hash Partitionin
g
map
s
d
a
ta b
y
hash value
of pa
rtition
key. You
can imp
r
ove
I/O perform
ance by Hash Partiti
oning
, becau
se
Hash Pa
rtitioni
ng map
s
dat
a
equally to ce
rtain partition
s.
(4)
Com
p
o
s
ite Partitioning: Com
p
o
s
ite
Partitioning, for ex
ample
Ran
g
e
-Hash
Partitioning o
r
Range
-Li
s
t Partitioning, allows
you
sub pa
rtition a
table by
cre
a
te pa
rtition i
n
an
exist pa
rtition
,
when
the ta
ble is con
s
id
ered
to
hu
ge
after pa
rtitio
ning o
r
whe
n
there
are ot
her
requi
rem
ents.
3. Conce
p
tu
al Data M
o
d
e
l of Struc
t
u
r
ed Electro
nic Medical Re
cords Sy
stem
Structu
r
ed
el
ectro
n
ic me
di
cal
re
co
rd
s
system
in
ou
r
hospital u
s
e
s
dynami
c
an
alysis of
data structu
r
e
.
The con
c
ept
data mode is sho
w
n in Fig
u
re 1.
Table na
med
medical
_
re
cord_i
ndex st
ore
s
i
ndexe
s
of medical reco
rd
s. Tabl
e named
medical_
re
co
rd_el
eme
n
t_d
e
finition store
s
eleme
n
t de
finitions of medical re
cord.
Table nam
ed
medical_
re
co
rd_d
ata
store
s
el
ement
da
ta of ea
ch
m
edical record. Tabl
e n
a
me
d mo
dify_trace
store
s
tra
c
e
of modificatio
n
of data of each
medi
cal record. Ta
ble name
d
speci
a
lty_form
at
store
s
me
dical record formats of ea
ch sp
e
c
ialty. Table nam
ed format_
d
e
finition stores
definition
s
of
every fo
rma
t
s. Tabl
e n
a
m
ed fo
rmat
_s
tyle
s
t
or
es
s
t
yle
s
o
f
ea
ch
fo
r
m
a
t. T
a
b
l
e
named
style_
definition sto
r
es def
initio
n of related styl
es.
Figure 1. Con
c
ept Data Mo
de of Struct
u
r
ed Electroni
c Medical Re
co
rds Syste
m
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Partitioning of
Ora
c
le Appli
c
ation in Stru
ct
ure
d
Electronic Me
dical Re
cords
(Ch
uandi Pan
)
5695
4. Ac
complishment of
Using Par
t
itio
n to
Stor
e
Data
From S
t
r
u
ctur
ed Elec
tronic
Medic
a
l
Reco
rds Sy
s
t
em
We u
s
e
parti
tion table to
store
data of
medical_record_
data tabl
e and m
odif
y
_trace
table, be
cau
s
e the
s
e tabl
e
s
a
r
e
extrem
ely huge
in
st
ructu
r
e
d
ele
c
t
r
oni
c me
dical
re
co
rd
s sy
stem
based on dyn
a
mic an
alysi
s
of data stru
cture
[5]. Reali
z
ation of this
stora
ge meth
od:
Firstly, we
create several
table spa
c
e
s
, lik
e t
s
_em
r1
, ts_em
r2, ts_emr3, ts_e
mr4, et
c,
whi
c
h
of dat
a files are
stored
in
different phy
si
cal disks. (Certai
n
ly,
these
da
ta
files ca
n be
store
d
in th
e
same
physi
cal disk, if I/O perfo
rm
an
ce is
not con
s
id
ered.
) Statem
ents of
creati
ng
table sp
ace a
r
e as b
e
lo
w:
cre
a
te ta
ble
s
pa
ce
ts_
e
m
r1
datafile
'D
:\oradata
\ts_emr1\data
01.dbf' si
ze
100M
autoextend o
n
next 100M maxsize 300
00M;
Rema
rk: nam
e of table
sp
a
c
e: ts_em
r1.
Data
file:
D:\oradata\ts
_e
mr1\data
01.d
b
f. Store
value of initial: 100M. Auto extension: al
low.
Next extensi
on: 100M
. Store value of maximum:
3000
0M.
Secon
d
ly, we
create
anthe
r table
sp
ace for stori
ng dat
a of index.
cre
a
te table
s
pa
ce ts_in
dex datafile
'D:\orad
a
ta
\ts_index\dat
a01.dbf' size
100M
autoextend o
n
next 100M maxsize 300
00M;
There are
statements of
cre
a
ting pa
rti
t
ion table of medi
cal_
re
co
rd_d
ata by range of
medical_
re
co
rd_id.
cre
a
te table
medical_
re
co
rd_d
ata (
medical_
re
co
rd_id n
u
mb
er(10
)
not null,
element_
defi
n
ition_id nu
m
ber(8) n
o
t nul
l,
seri
al_n
umbe
r numb
e
r(8) n
o
t null,
e_re
s
ult VARCHA
R2
(20
0
),
e_modifie
r
VARCHAR2(2
0
),
modify_date DATE,
prima
r
y key(medical_
re
co
rd_id
)
usin
g index table
s
pa
ce TS
_INDEX )
pctfree
10 pct
u
se
d 90
partition by ra
nge(medi
cal_
record_id
)
(pa
r
tition med
i
cal_
re
co
rd_d
ata_p1 valu
e
s
less than
(5
0000
) table
s
p
a
ce ts_em
r1,
partition me
di
cal_
re
co
rd_d
ata_p2 valu
e
s
less than
(1
0000
0) table
s
pace ts_em
r2
,
partition me
di
cal_
re
co
rd_d
ata_p3 valu
e
s
less than
(1
5000
0) table
s
pace ts_em
r3
,
partition me
di
cal_
re
co
rd_d
ata_p4 valu
e
s
less than
(2
0000
0) table
s
pace ts_em
r4
);
PCTF
REE: Specify a
whol
e num
be
r rep
r
esentin
g
the percenta
ge o
f
spa
c
e
in ea
ch data
block of th
e
databa
se
obj
ect reserved
for future up
dates to rows of the
obj
e
c
t [6]. While
the
percenta
ge
of spa
c
e
in d
a
ta blo
c
k re
ach
e
s the
value
of PCTF
REE, data bl
ock
which i
s
rem
o
ved
from FREE LI
ST is not allo
wed to in
se
rt. If the va
lue of PCTFREE is too large, d
a
ta block is
n
o
t
taken full adv
antage of. Otherwise, if th
e value of
PCTF
REE is so little, new reco
rd
s are fo
rce
d
to move to new data blo
c
k, and pe
rform
a
nce
will be de
grad
ed.
PCTUSE
D: Specify
a wh
ole
n
u
mbe
r
repr
e
s
e
n
ting t
he mini
mum
perce
ntage
of used
spa
c
e that O
r
acl
e
maintai
n
s for ea
ch
data bl
o
ck of
the database obje
c
t [6].
Only whe
n
used
space of
dat
a block fall
s
below P
C
TUSED, data bl
oc
k
will
be
re-linked to available li
st and will
be all
o
wed to
insert
data.
If the value
o
f
PCTUSE
D i
s
too
la
rge, f
r
equ
ent
re-using d
a
ta bl
ock
lead
s to in
creasi
ng I/O
consumption.
If the value
of PCTUSE
D is
s
o
little, this
works
to
the
disa
dvantag
e
of re-u
sing
data block.
However,
little PCTUSE
D value hel
p
s
to red
u
ce I/O
con
s
um
ption and
imp
r
ove perfo
rman
ce.
Cho
o
si
ng p
r
o
per P
C
TF
REE value and
PCTUSE
D value sum of
whi
c
h mu
st e
qual to p
r
less
than 100 is vitally important. In our
s
y
s
t
e
m
, w
e
se
t PCT
F
R
EE va
lue
as
10
a
nd s
e
t
PCTUSE
D value as 9
0
.
Table medi
ca
l_re
co
rd_
data
_p1 stores
da
ta
wh
i
c
h
med
i
cal_
re
co
rd_i
d is le
ss t
han
500
00
and
whi
c
h i
s
store
d
in
tabl
e spa
c
e of t
s
_emr1. Tabl
e
medi
cal_
re
cord_
data_
p2
store
s
data
which
medical_
re
co
rd_id i
s
equal
to or larger than 500
00 a
nd less than 1000
00 and
whi
c
h is sto
r
e
d
in
table sp
ace o
f
ts_emr2, an
d so on.
Enlargi
ng tabl
e sp
ace or
achieving cro
s
s
disk
sto
r
ag
e
can
be reali
z
ed by add
dat
a file to
table sp
ace, for exampl
e ts_emr1, ts_em
r2, ts_em
r3, ts_em
r4.
alter table
s
p
a
ce ts_em
r1
add datafil
e
'E:\oradat
a\ts_em
r1\dat
a02.dbf' size
100M
autoextend o
n
next 100M maxsize 200
00M;
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5693 – 56
98
5696
After creating table space which
will
be used
to
store new partition, we
can add a
partition to pa
rtitioned tabl
e
.
For an in
sta
n
ce,
we ad
d
a partition
wh
ich will
be u
s
ed to sto
r
e d
a
ta
whi
c
h m
edica
l_re
co
rd_id
is equal to
or l
a
rge
r
tha
n
20
0000
and l
e
ss than
25
000
0 and
whi
c
h
i
s
store
d
i
n
tabl
e spa
c
e
of t
s
_em
r5
to ta
ble m
edical_
record_d
ata. The stateme
n
t
is
sho
w
n as
belo
w
:
alter tabl
e
medical_
re
co
rd_d
ata a
d
d
partition
m
edical_record
_data_
p5 val
ues l
e
ss
than(2
500
00
) tablesp
a
ce ts_em
r5;
There
are statement
s of
cre
a
ting partiti
on tabl
e of modify_trace by range of
archive_d
ate.
cre
a
te
table modify_tra
ce (
medical
_
record_i
d num
be
r(1
0) not null,
element_d
efinition_id nu
m
ber(8) n
o
t nul
l,
e_re
s
ult VARCHAR2(200
),
modifier VARCHAR2(10),
modify_date DATE,
archive
_date
DATE not nu
ll,
primary
key(medical_
re
co
rd_id,
ele
m
en
t_definition_i
d, modify_date)
using in
dex table
s
pa
ce TS
_INDEX )
pctfree 0 p
c
t
u
se
d 99
partition by range
(a
rchive_date)
(partition
m
odify_trace_p
1 values less than(to_date
('2014-01-01',
'yyyy-mm-dd'))
tablespa
ce ts_emr101,
partition m
odify_trace_p2 values l
e
ss
than(to_date('2015-01
-01', 'yyyy-
mm-dd'))
tablespa
ce ts_emr102,
partition m
odify_trace_p3 values l
e
ss
than(to_date('2016-01
-01', 'yyyy-
mm-dd'))
tablespa
ce ts_emr103,
partition mod
i
fy_trace_
p0
values le
ss than (maxvalu
e
)
tablespa
ce ts_em
r1
99);
S
t
ruct
u
r
ed el
ect
r
o
n
ic m
e
d
i
cal re
co
rd
s
sy
stem ba
se
d on SaaS [7] can be
sh
ared to
several ho
spi
t
als throu
gh
by storing m
edical re
co
rd
indexes of d
i
fferent hospitals into different
list pa
rtition
s
.
There a
r
e
sta
t
ements
of
creating li
st
partition table
of
medical_
re
co
rd_in
dex by
li
st
of hospital
_
id
.
cre
a
te table
medical_
re
co
rd_in
dex (
medical
_
record_i
d num
be
r(1
0) not null,
patient_id nu
mber(10
)
not
null,
hospital
_
id n
u
mbe
r(5
),
format_id nu
mber(5
),
primary
key(medical_
re
co
rd_id
)
using in
dex table
s
pa
ce TS
_INDEX )
pctfree 2
0
pctused 8
0
partition by list(ho
spital
_id
)
(partition me
dical
_
re
co
rd_
i
ndex_p
1 values(1
) table
s
pace ts_em
r2
01,
partition med
i
cal_
re
co
rd_i
ndex_p
2 valu
es(2) tabl
esp
a
ce ts_em
r20
2
,
partition med
i
cal_
re
co
rd_i
ndex_p
3 valu
es(3) tabl
esp
a
ce ts_em
r20
3
,
partition med
i
cal_
re
co
rd_i
ndex_p
0 valu
es(default
)
tablespa
ce ts_
e
mr2
99);
5, Results a
nd Discu
ssi
on
We ta
ke tab
l
e modify_tra
ce fo
r an
e
x
ample to compa
r
e the
perfo
rman
ce
of not
partitione
d ta
ble and
partiti
oned tabl
e to prove that ta
ble with p
a
rtition will imp
r
o
v
e performan
ce.
Firstly,
two tables nam
e
d
modify_tra
ce_
non
with
out pa
rtition
and m
odify_trace
whi
c
h
is
partitione
d b
y
rang of archive
_date
are cre
a
t
ed.
There a
r
e
the statem
ents of crea
ting
modify_tra
ce
by range of a
r
chive
_date.
cre
a
te
table modify_tra
ce (
medical_
re
co
rd_id n
u
mb
er(10
)
not null,
element_
defi
n
ition_id nu
m
ber(8) n
o
t nul
l,
e_re
s
ult VARCHA
R2
(20
0
),
modifier VARCHA
R2
(10
)
,
modify_date DATE,
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Partitioning of
Ora
c
le Appli
c
ation in Stru
ct
ure
d
Electronic Me
dical Re
cords
(Ch
uandi Pan
)
5697
archive_d
ate DATE not nul
l,
prima
r
y key(medical_
re
co
rd_id, el
e
m
en
t_definition_i
d, modify_date)
usin
g index table
s
pa
ce TS
_INDEX )
pctfree
0 pctu
sed 9
9
partition by ra
nge(archive_
date)
(partition m
odify_trace_p1 values l
e
ss
than(to_date(
'2009-01-01', 'yyyy-mm-dd'))
tablespa
ce ts_emr101,
partition modify_trace_p2
va
lues less than(to_date(
'2010-01-01', 'yyyy-mm-dd'))
tablespa
ce ts_emr102,
partition modify_trace_p3
va
lues less than(to_date(
'2011-01-01', 'yyyy-mm-dd'))
tablespa
ce ts_emr103,
partition modify_trace_p4
va
lues less than(to_date(
'2012-01-01', 'yyyy-mm-dd'))
tablespa
ce ts_emr104,
partition modify_trace_p5
va
lues less than(to_date(
'2013-01-01', 'yyyy-mm-dd'))
tablespa
ce ts_emr105,
partition mo
di
fy_trace_
p6 value
s
less tha
n
(maxvalue
)
tablespa
ce ts_emr106
);
Then, after resp
ectively inse
rting sam
e
massive d
a
ta which co
ntains 10
000
000 ro
ws
archive_d
ate
ra
ngin
g
fro
m
01/0
1
/200
8 to
31/
12/2
013 i
n
to
bot
h table
s
,
we
execute
qu
ery
statement
s fo
r these t
w
o ta
bles.
sele
ct *
from
modify_trace_n
on
wh
ere
archive_
date bet
wee
n
to_date('20100716','yyyymmdd')
and to_date('201007
17','yyyymmdd') order
by archive_date;
sele
ct *
from
modify
_trace
whe
r
e
a
r
chive
_date
betwee
n
to_date('20100716','yyyymm
dd')and to_date('201007
17','yyyymmdd') order
by archive_date;
We obtai
n sta
t
istics from ex
ecutin
g these
statement
s:
(1)
The
execution plan of
table modify
_trace_non is
“TABLE A
CCESS FULL”, Cos
t
numbe
r i
s
1
7
833
and Byte
s a
r
e
190K,
as
sh
own
in
Tabl
e
1. Mo
re visible
resu
lt is that it ta
ke
s
3.813 second
s to retrieve 5
539 ro
ws of data.
Table 1. Execution Plan of Table mo
dify_trace_n
on
ID Oper
ation
name
Ro
w
s
B
y
tes
Cost(%CP
U)
0 SELECT
STATE
M
ENT
4641
190K
17833(2
)
1 SORT
ORBE
R
B
Y
4641
190K
17833(2
)
2
TABLE ACCESS
FULL
MODI
F
Y
_TRA
C
E
_NON
4641
190K
17832(1
)
(2)
The
execution pla
n
of t
able mo
dify_tra
ce i
s
“PARTITION
RANGE SINGLE
”
and the
n
“TABLE ACCESS FULL”.
Bec
a
us
e part
ition s
t
art and
partition
s
t
op is
both 3,
it only need to
scan
partition
3 n
a
med
ts_emr103. A
s
a
re
sult, it
o
n
ly take
s
2.9
85
se
con
d
s to ret
r
ieve
55
39
rows
mu
ch fa
ster th
an th
e
result of exe
c
uting
the
pre
v
ious
statem
ent an
d the
Cost n
u
mbe
r
a
nd
Bytes are re
spectively red
u
ce
d to 2492
and 15
7K, as
sho
w
n in Ta
ble2. Each p
a
rtition is a small
part of the wh
ole data table
,
so retrievin
g
a par
tition is
boun
d to improve retrieval
perfo
rman
ce.
Table 2. Execution Plan of Table mo
dify_trace
ID Oper
ation
name
Ro
w
s
B
y
tes
Cost(%CP
U)
Pstart
Pstop
0 SELECT
STATE
M
ENT
3935
157K
2492(2
)
1
PARTIT
O
N RA
N
G
E SIN
G
LE
3935
157K
2492(2
)
3
3
2 SORT
ORBE
R
B
Y
3935
157K
2492(2
)
3
TABLE ACCESS
FULL
MODI
F
Y
_TRA
C
E
3935
157K
2491(2
)
3
3
As a re
sult, the Ora
c
le
ra
nge pa
rtitioni
ng tech
nolo
g
y
can optimize the huge a
m
ounts of
data table qu
ery perfo
rma
n
ce [8].
6, Conclusio
n
There are ma
ny advantage
s of u
s
ing p
a
rtitioning of Oracle to
store data
from structured
electroni
c medical re
cord
s system. (1
) Improvi
ng
perfo
rman
ce:
Operatio
ns
of insert, del
ete,
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5693 – 56
98
5698
update
an
d q
uery a
r
e
di
stri
buted to
de
sti
nation
parti
tio
n
, so
pe
rform
ance i
s
imp
r
o
v
ed effectivel
y.
This i
s
th
e
most im
porta
nt advantag
e
for
stru
ct
ure
d
ele
c
tro
n
ic
medical reco
rds sy
stem.
(2)
Conve
n
ient
maintena
nce: We
ca
n mai
n
tain certai
n partition,
for example,
d
a
ta
re
pair. (3
) I/O
equilib
rium: I/O equilib
riu
m
is achieve
d
by storin
g
different pa
rtitions to dif
f
erent di
sks.
(4)
Increasi
ng
usability: If cert
ain partition
of table fail
s,
other partitio
ns
of table st
ill work well.
(5)
Conve
n
ient b
a
ckup
an
d re
covery: ba
ckup
a
n
d
re
co
v
e
ry
can
be
d
one to
certai
n pa
rtition.
(6)
Partitioning
is tra
n
spa
r
ent
to u
s
ers. P
a
rtiti
oning
ha
s mo
re
obvi
ous adva
n
ta
ges to
stru
ctured
electroni
c me
dical records
system b
a
se
d on SaaS.
Referen
ces
[1]
Di PC. D
e
si
g
n
of El
ectroni
c Medic
a
l
Re
cord S
y
stem
Based
on
Cl
o
ud C
o
mp
uting
T
e
chnolo
g
y
.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2014; 1
2
(5): 4
010-
401
7.
[2]
Cai S
h
a
ohu
a.
Desig
n
and
im
plem
entatio
n
o
f
electro
n
ic me
dical
recor
d
s.
Moder
n C
o
mp
uter
. 20
06; 7:
110-
112
[3]
Yang Y, Ye H, F
e
i S.
Design of partitioned tabl
e for VLDB in the vehicle monit
o
ring system.
Multimed
ia T
e
chno
log
y
(ICMT
)
. Inter
nation
a
l Confer
ence on
. IEEE. 2011;5377-
537
9.
[4]
He Pi
ng, Ya
ng
Shuq
ia
ng, Ji
a
Yan. Back
up
and r
e
cov
e
r
y
of VLDB
base
d
on
Oracle
1
0g.
Co
mpute
r
eng
ine
e
ri
ng
. 2
006; 32(
19): 79
-81.
[5]
Pan C
h
u
a
n
d
i,
Z
hou
Xinch
ao,
Shi Ji
an. D
e
si
g
n
of
e
l
ectro
n
ic
medic
a
l rec
o
rd
s
y
stem
bas
ed
on
d
y
n
a
mi
c
ana
l
y
sis of dat
a structure
.
Chi
na Di
gital Me
di
cine.
20
07; 2(3
)
: 20.
[6]
F
i
orill
o C. Or
acle
Data
bas
e
11
gR2
Perfo
rmance T
uni
n
g
Co
okb
ook:
Over 80
Rec
i
pes to
He
lp
Begi
nners Ac
h
i
eve Better Pe
rformance fro
m
Oracle
Dat
abas
e App
licat
ions. Packt P
ublis
hi
ng Ltd.
201
2.
[7]
Buxm
an
n P, Hess T
,
Lehman
n S. Soft
w
a
re
as a Service.
Wirtschaftsinform
atik
. 20
08; 5
0
(6): 500-
50
3.
[8]
W
ang P, T
i
an
A, Guo C. T
abl
e Partit
i
oni
ng T
e
chno
lo
g
y
Base
d o
n
M
a
ssive
Data.
TEL
K
OMNIKA
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2014; 1
2
(3): 1
676-
168
6.
Evaluation Warning : The document was created with Spire.PDF for Python.