Indonesian J
ournal of Ele
c
trical Engin
eering and
Computer Sci
e
nce
Vol. 2, No. 3,
Jun
e
201
6, pp. 720 ~ 72
8
DOI: 10.115
9
1
/ijeecs.v2.i3.pp72
0-7
2
8
720
Re
cei
v
ed Ma
rch 4, 2
016;
Re
vised
Ma
y 13, 2016; Accepted Ma
y 28
, 2016
Hybrid Disk Drive Configuration on Database Server
Virtualization
Ferd
y
Nir
w
a
n
s
y
ah*
1
, Suharjito
2
Magister i
n
Informatio
n
T
e
chnol
og
y, Bin
u
s Graduat
e Pro
g
r
am, Bina Nus
antara U
n
ivers
i
t
y
, Jakarta,
Indon
esi
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: ferd
y
.
n
i
r
w
a
n
s
y
a
h
@
b
in
us.ac.id; suhar
jito@
b
i
nus.e
du
A
b
st
r
a
ct
Soli
d State Dr
i
v
e (SSD) is
a
revol
u
tion
ary n
e
w
storage te
chno
logy. E
n
te
rprise stor
age
system
usin
g ful
l
SS
D
is still
very
exp
ensiv
e, w
h
ile
h
a
rd
di
sk
drive
(
HDD)
is stil
l w
i
dely
use
d
. This
study
discuss
e
s
hybri
d
co
nfig
ur
ation
stora
g
e
i
n
virtu
a
li
z
e
d
s
e
rver
data
bas
e w
i
th
benc
h
m
ark a
gai
nst fo
ur hy
brid
stora
g
e
config
uratio
n for four datab
a
s
es, ORACLE, SQL
Server,
MySQL and PostgreSQL on
W
i
ndow
s Serve
r
virtuali
z
a
t
io
n. B
ench
m
ark usi
n
g T
P
C-C a
nd T
P
C-H to g
e
t th
e best
perfor
m
ance
of four c
o
nfigur
ations
w
e
re
tested. T
he results of this study
indic
a
te HD
D storage co
nfigur
ations
as vi
sual disk dr
ive
OS and SSD as
visual
disk
driv
es data
bas
e g
e
t better p
e
rfor
ma
nce
as
o
n
li
ne trans
actio
n
process
i
ng (O
LT
P) and
on
li
ne
ana
lytical pr
oc
essin
g
(OLAP) databas
e serv
er compar
e
d
w
i
th SSD as visual disk dr
ive
OS and HDD
as a
vi
su
a
l
di
sk d
r
ive
da
ta
ba
se
.
Ba
se
d on
th
e
d
a
t
a
re
se
arch T
P
C-C,
OLT
P
get best perf
o
rmanc
e
at H
D
D
storage c
onfig
uratio
ns as vis
ual d
i
sk driv
e OS and SSD
as a visu
al d
i
s
k
drives d
a
tab
a
se an
d te
mp
o
r
ary
files.
Ke
y
w
ords
: Da
tabase, Hi
gh a
v
aila
bi
lity, Server
virtuali
z
a
t
io
n
,
Hybrid storag
e
Copy
right
©
2016 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
Datab
a
se be
come
s an int
egral p
a
rt of our daily life [1]. In
the modern b
u
si
ne
ss wo
rld,
the datab
ase
became
op
eration
a
l
sup
port
comp
an
y called
on
line tra
n
sacti
on p
r
o
c
e
ssin
g
(OLTP
)
. The
databa
se
is also
u
s
ed t
o
assi
st com
panie
s
in
an
alyzing
and
makin
g
d
e
ci
sions,
kno
w
n a
s
th
e on line a
nalytical p
r
o
c
e
ssi
ng (OL
AP). Databa
se ha
s tra
n
s
a
c
tion data
and
sup
portin
g
d
a
t
a need
ed. M
o
re th
e nu
mb
er of tran
sa
ctions
on th
e a
pplication a
n
d
the d
a
taba
se,
more
comple
x the ap
plication a
nd
data
base inf
r
as
t
r
u
c
ture
that
mu
st be
o
w
ne
d
by the
com
p
a
n
y
[2]. Infrastru
c
ture a
ppli
c
ati
ons an
d d
a
ta
bases is
e
s
sential in
orde
r to im
prove
high
availabili
ty.
To improve the perfo
rma
n
ce
can b
e
with per
fo
rm
ance and tu
ning, namely
on: first, se
rver
environ
ment
su
ch a
s
mai
nboa
rd, p
r
o
c
essor,
RAM, Lan
card a
nd othe
rs. S
e
co
nd, sto
r
a
ge
environ
ment,
third, d
a
taba
se environm
en
t, fourth
, net
work environ
ment, fifth, desktop
co
mp
uter
environ
ment [3].
Along with t
he develo
p
m
ent of tech
n
o
logy
, high
availability can be im
pro
v
ed with
virtualizatio
n. Manag
emen
t of server h
a
r
dware
and
software
be
co
mes mo
re p
r
actical. Some
of
the feature
s
that can b
e
d
one a
r
e: man
ageme
n
t of hard
w
a
r
e re
so
urces to e
a
ch
OS in a VM,
for
example, pro
c
e
s
sor, RAM
and sto
r
ag
e, monitori
ng re
sou
r
ce allo
ca
tion to each
OS and othe
rs.
At the storag
e level, we re
cog
n
ize seve
ral ty
pe
s of di
sk,
namely: solid state
driv
e (SSD)
and m
agnetic drive (SAS
and SATA
)
,
sometim
e
s call
ed the hard di
sk drive (HDD). SS
D i
s
a
revolution
ary new storage
t
e
ch
nolo
g
ies and
a
p
o
siti
v
e
impa
ct on t
he pe
rform
a
n
c
e of
system
and
databa
se.
In
vestment
of enterp
r
i
s
e st
orag
e syste
m
that u
s
e
s
full SSD sto
r
age
system
i
s
still
very expen
si
ve, on the ot
her
han
d the
r
e a
r
e m
any enterp
r
i
s
e storag
e
syste
m
usi
ng
HDD a
s
lega
cy
sy
st
e
m
.
With the con
d
ition of existing storage, it infl
uence
d
the desig
n of the IT infrastructure in
enterpri
s
e
storage sy
stem
s. In or
der for
the utilization of all
stor
age resources and the resulti
n
g
optimal pe
rformance, t
here
sho
u
ld be
bet
wee
n
SSD an
d HDD
config
uration. T
h
is
also
applie
s t
o
the configuration of storage on
the dat
abase
server. Th
is
study
will do a hybrid
configurati
o
n
databa
se
storage on the vi
rtualizatio
n se
rver.
To get the b
e
s
t po
ssi
ble
confi
guration should be
pe
rf
orme
d
be
nch
m
ark. The
cri
t
eria for
a go
od b
e
n
c
hmark fo
r th
e pe
rform
a
n
c
e are a
s
fo
llows
: firs
t, repres
entative, sec
o
nd, relevant,
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 3, Jun
e
2016 : 720
– 728
721
third, porta
bl
e, fourth, sca
l
able, fifth, and sixt
h verifi
able, sim
p
le. There are so
me ben
chm
a
rks
that reach
active indust
r
y standards. T
he most
com
m
only used i
s
the TP
C and SPEC. For
this
study, usin
g a TPC ben
ch
mark perfo
rm
ance mea
s
ur
ement. Beca
use the focu
s of research
in
the Database OLTP
and OLAP will
use the TP
C-C and TPC-H.
On
the other
hand, to
compl
e
te
this re
se
arch,
perfo
rman
ce
measureme
n
t testi
ng of f
our d
a
taba
se
s are O
R
ACL
E
, SQL Server,
MySQL and
PostgreSQL.
2. Related
Works
Mao m
a
kes parity-ba
se
d fram
ework Hybri
d
Disk
Array Architec
ture
(HPDA) that
combi
ne gro
up and two
HDD SSDs t
o
improve
p
e
rform
a
n
c
e
and reli
ability of SSD-based
st
ora
ge
sy
st
e
m
s.
I
n
HP
DA
,
S
S
D
(
data di
sk) an
d pa
rt of the HDD (
parity disk)
write array RAI
D
4.
Reliability analysis show
ed that the
reliability HPDA, in term
s
of Mean Tim
e
to Data Loss
(MTT
DL
), hig
her th
an the
HD
D
or SS
D. HP
DA
prototype impl
ementat
ion
a
nd pe
rforma
nce
evaluation
sh
owe
d
that sig
n
ificant
ly exceed
s HPDA
SSD and HDD [4].
Bassil
p
r
e
s
en
ts a
compa
r
at
ive study
on
the
pe
r
f
or
manc
e
o
f
th
e s
y
s
t
e
m
o
v
er
the
D
B
MS.
Testing i
s
in MSSQL Server 200
8, Ora
c
le1
1g, IBMDB2, MySQL5.5 and MS Acce
ss 2
010. T
h
is
test aim
s
to e
x
ecute
a diffe
rent SQ
L qu
e
r
y with
di
fferent levels
of complexity five DBMS tes
t
ed.
This
will open the
way to make a
head-to-
head comparative evaluat
ion
shows the average
execution time, memory usage an
d CP
U utilization
of each DBM
S
after compl
e
tion of the t
e
st.
The t
e
st
re
su
lt
s sh
ow
ed
n
o
DB
MS
that
has the m
o
st excell
ent
p
e
rform
a
n
c
e. I
B
M DB2 i
s
t
h
e
fastest
DBM
S
, but MS Access
has
a lower CP
U
utilization than
any other DB
MS and IBM
DB2
is the mo
st wi
dely con
s
um
ed main mem
o
ry [5].
Kim
made
a system call
e
d
the Hybrid
Stor
e whe
r
e there are
2 d
i
provide: Hyb
r
id-Pla
n
and
Hybri
d
Dyn. Hybrid
-Pl
an
will imp
r
o
v
e ca
pa
city
p
l
annin
g
for a
d
minist
rators with th
e ove
r
all
goal of ope
rational cost
-budg
ets. Hyb
r
idDyn
im
pro
v
e the perfo
rmance / lifetime gua
rante
e
s
durin
g epi
so
des
of deviat
i
on of the
wo
rklo
ad.
Te
sti
ng an
d imple
m
entation n
e
ed to be
kn
o
w
n
advantag
es a
nd di
sadvant
age
s of th
is framework. To
evaluate in
te
rms
of perfo
rmance an
d to
tal
co
st. Analysi
s
speed
perfo
rman
ce hyb
r
i
d
store is
n
e
a
r
ing full SSD
stora
ge, but
on top of the
full
stora
ge HDD [6].
Bausch, Pet
r
ov, and
Buch
mann
ob
se
rved the
di
ffere
nt pe
rform
a
n
c
e
of the
join
algorith
m
available in P
o
stgreSQL o
n
SSD and HDD. Firs
t ob
serving, point
query sho
w
s
the perfo
rma
n
ce
improvem
ent
of up to fifty times. Se
con
d
,
the r
ang
e of
the que
ry to perfo
rm is
we
ll on the SSD.
Join al
go
rith
m behave
s
di
fferently depe
nding o
n
ho
w well they co
nform to the
nature
of SSD or
HDD [7].
Acco
rdi
ng to
prop
ose an
d
desi
gn a
syst
ematic
exploration of the
u
s
e of SS
D to
improve
perfo
rman
ce
buffer man
a
g
e
r of the DB
MS. They pr
o
pose thre
e al
ternatives th
a
t
differ prima
r
ily
in ho
w th
e b
u
ffer ma
nag
e
r
cope
s
with
dirty pag
es to be
re
move
d from
the
b
u
ffer p
ool. T
hey
impleme
n
t this alternative, as well as
other n
e
w al
gorithm p
r
o
p
o
se
d for this study (TA
C
), in
SQLServe
r, and perfo
rm te
sting u
s
ing variou
s
be
nch
m
arks (TPC-C and TP
C-H) on so
me scale
factors. Empi
rical
evalu
a
tion
sho
w
e
d
si
gnifica
nt
pe
rforma
nce imp
r
oveme
n
ts
of their
metho
d
s
of
improvem
ent
on the
co
nfiguratio
n of th
e HDD (u
p
t
o
9.4X), a
nd
up to 6.8X m
o
re
accel
e
rati
on
with TAC [8].
In anothe
r
study, pre
s
e
n
ting an
alytical
tool
to a
s
sess the
co
nfigu
r
ation fo
rme
d
by the
combi
nation
of all kind
s o
f
reso
urce
s. Usi
ng
the to
ol to analyze
the logical volume
colle
ct
ed
statistics
coll
ected
from
1
20 la
rge
p
r
o
ductio
n
sy
st
e
m
s.
Thi
s
st
u
d
y
sh
ow
ed
a
co
mbinat
io
n
of
SSD, SCSI,
and SATA co
nfiguratio
ns i
n
many ca
se
s be better th
an just usi
ng SCSI devices in
all key aspe
cts: price, perf
o
rma
n
ce and
powe
r
co
nsumption. Thi
s
co
ntra
sts
with other re
ce
nt
studie
s
on e
n
t
erpri
s
e
syste
m
smalle
r pe
ssimi
stic
a
b
o
u
t profits SDD in enterpri
s
e setting
s [9].
Jo create
s
HybridCopy-on
-
Write (CoW) di
sk
stora
g
e
that combin
es SSD and
HDD to
Con
s
oli
date
environ
ment.
The pro
p
o
s
ed schem
e p
u
ts rea
d
-only
disk ima
ge
template on
th
e
SSD, while
write operations to
the HDD. Creating an efficient
com
b
ination of S
S
D and
HDD in
con
s
oli
date e
n
vironm
ent. Hybrid
storag
e Co
W
is cl
e
a
rly benefi
c
ia
l, for perform
ance and
co
st
effective. The
dra
w
b
a
ck i
s
workin
g in
V
M
wa
re l
e
vel
so
as to m
e
a
s
ure the
pe
rf
orma
nce of
e
a
ch
VM hard. From the test result
s obta
i
ned CoW
Hybrid sto
r
ag
e
perform
an
ce over sto
r
a
ge,
enabli
ng full HDD but still
belo
w
the full SSD storage
[10].
Wu a
nd
Red
d
y make a framework by
makin
g
d
r
ivers on
Linux. M
anag
ement o
f
stora
g
e
c
a
pa
c
i
ty on
a
s
t
or
ag
e
sys
te
m us
ed a
h
y
b
r
id SSD
an
d
HDD
. T
h
is
F
r
ame
w
ork
do
es
a
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Hybrid Disk
Drive Config
uration
on Data
base Server
Virtualization
(Fe
r
dy Nirwa
n
syah
)
722
combi
nation of
SSD
and HDD config
u
r
ation.
Te
stin
g and im
ple
m
entation n
e
ed to be
kn
o
w
n
advantag
es
and di
sa
dvan
tages
of this frame
w
o
r
k. Hybrid HDD ben
chma
rk
a
nd
pe
rforman
c
e
stripin
g
and h
i
s claim
coul
d
rise to 50% i
n
some
ca
se
s [11].
Lee ma
ke
s
t
h
ree
differen
t
types of SSD model from Samsu
n
g
. Shows ho
w SSD
techn
o
logy
h
a
s
advan
ce
d
to reverse
the trend
of
wide
ning
the
pe
rform
a
n
c
e ga
p b
e
twe
e
n
pro
c
e
s
sor an
d sto
r
a
ge. T
h
is
study
also
sho
w
s that
e
v
en a
sin
g
le
SSD can
out
perfo
rm
RAID 0
with eig
h
t 15
K-RPM
enterprise-cl
a
s
s di
sk drive
s
on
transactio
n
th
roug
hput,
co
st effectiven
e
s
s
and po
we
r co
nsum
ption [1
2].
Park present
s the techniques
used to
increase the
reliab
ility and performance of the
new SSD
RAID system.
First, they an
alyzed
the S
S
D RAID me
cha
n
ism
and
then develo
p
a
methodol
ogy
adapte
d
to the ne
w RAID
SSD storage.
Via trace-d
r
i
v
en simul
a
tio
n
, they evaluate
perfo
rman
ce
optimize
d
SSDs u
s
e
RAID mechani
sm. The p
r
o
posed meth
o
d
improve
s
t
he
reliability of SSD is 2% hi
gher than the
existi
ng RAI
D
system
s and improv
e I / O performance o
f
SSDs 28% hi
gher tha
n
the
existing RAID syste
m
s [1
3].
3. Rese
arch
Metho
d
SSD techn
o
l
ogy revolutio
n
izin
g sto
r
ag
e has th
e po
tential to cha
nge the a
r
chi
t
ectural
prin
ciple
s
DB
MS [6]. But
the SSD itself
is st
ill quite
expen
sive. On the other h
and ente
r
p
r
ise
stora
ge syste
m
s still
use
HDD
as lega
cy
system. In th
is c
onn
ectio
n
, then the
r
e
wi
ll be q
u
e
s
tion
s
like:
1.
Ho
w to impro
v
e perfo
rman
ce of d
a
taba
se
serv
e
r
s wi
th hybrid
storage
configu
r
ation that is
optimal in the
serve
r
virtual
ization
?
2.
Ho
w to utilize
all the reso
u
r
ce
s that exist
in storage
without comp
ro
mising the p
e
r
forma
n
ce
of the databa
se serve
r
?
To take a
d
va
ntage of all the re
sou
r
ce
s exis
ting ha
rd drive we
u
s
e hyb
r
id techniqu
e
.
Hybrid
techni
que
co
nfigure
s
the
virtual
d
r
ive in
Win
d
o
w
s
Serve
r
as
datab
ase se
rver and
pa
rt of
VMWa
re
serv
er virtuali
z
ati
on. Thi
s
re
se
arch ma
ke
s
hybrid te
chni
que by p
e
rfo
r
ming virtual
d
i
sk
config
uratio
n
of ha
rd
di
sk d
r
ive
s
for the
OS an
d data
b
a
s
e
usin
g differe
nt sto
r
ag
e.
The
advantag
e wi
th doing
hybri
d
tech
niqu
e o
n
virtualiz
atio
n is in te
rm
s
of pra
c
tical
a
nd convenie
n
c
e
than do the h
y
brid at the st
orag
e level.
The step
s of
this study: study
of literature, install
a
tion
of the instru
ment, create a
databa
se a
n
d
then loadi
ng
data of fourth
databa
se
fo
r TPC-C an
d TPC-H testin
g as
well a
s
t
he
repla
c
e
m
ent scheme confi
guratio
n
of the virtual drive
,
collectin
g the test data, the perfo
rma
n
c
e
evaluation
system configu
r
ations a
nd
th
en the co
ncl
u
sion
s an
d su
gge
stion
s
.
First
step,
the
re
sea
r
ch
be
gins by d
e
termining th
e b
a
c
kgro
und
an
d
pu
rpo
s
e
of t
he
study
as well a
s
de
fining the sco
pe. The litera
t
ure
stu
d
y is
done to d
eep
en the un
derstandi
ng of the
hybrid techni
que of virtual
disks to virtual driv
e
s
on
Windo
ws Server virtuali
z
ation se
rver.
In
addition,
a lit
eratu
r
e
study
wa
s al
so
con
ducte
d to
fi
nd
out th
e
re
sult
s of
hybri
d
st
orag
e te
ch
niq
u
e
has eve
r
don
e.
Secon
d
step,
the
re
sea
r
ch in
strum
ent
is
in
stallatio
n
of VM
Wa
re, Storag
e,
Wind
ows
Server, and S
Q
L Serve
r
. Hard
wa
re for the study was:
1.
Intel Modul
ar Server
Chassi
s
MFSYS25V2:
14 dri
v
e carri
ers,
1
GbE
switch, two
power
sup
p
lie
s, two
powe
r
su
ppl
y fan.Node I: Intel MFS2600KI Comp
ute Module, 2
x Intel (R)
Xeon (R) E5-2660
CPU 0
@ 2.20
GH
z 8
Core
s, 24 G
B
DDR3 RA
M.
2.
2 pieces h
a
rd
drive Seagat
e
Savvio HDD 300 GB 10
K RPM 2.5 ".
3.
4 pieces h
a
rd
diskFou
r
ce CORSAIR GS
240GB SSD
2.5 ".
4.
8 pieces SAT
A hard drive
s
Seagate
Mo
mentus 5
00
GB 7.2K RPM 2.5 ".
5.
Switch hu
b Ci
sco SG500
-2
8 28 port
s
.
Firstly, in
stall
VMWa
re
on t
he
se
rver, th
en in
st
allatio
n
of
4 pi
eces
disk SSD u
s
i
ng
RAID
10 for the virtual disk drive
OS Windo
ws Server
in a VM and 8 pieces ha
rddi
sk HDD u
s
ing
RAID
10 a
s
a virtu
a
l disk d
r
ive f
o
r d
a
taba
se
OR
ACLE, SQL Serve
r
,
MySQL and
PostgreSQL.
Then
installatio
n
of Wind
ows Se
rver a
s
a Vi
rtual Ma
chin
e
(VM) in th
e
VMWa
re virt
ual di
sk d
r
ive
in
setting
s u
s
e
s
a co
nfiguration that ha
s b
een p
r
ov
ide
d
.
After that, do install
a
tion
Hamm
erDB
on
Wind
ows Se
rver. Installati
on of th
e fou
r
datab
ases,
n
a
mely: ORA
C
LE, SQL Se
rver, MySQL
and
PostgreSQL.
Configu
r
e th
e datab
ase
so that
each
databa
se i
s
dire
cted to S
S
D on d
r
ive
E.
Tempo
r
a
r
y O
S
and d
a
taba
se file
s a
r
e di
recte
d
to SS
D on
drive F.
This
pre
p
a
r
a
t
ion is d
one f
o
r
testing in the
first re
sea
r
ch.
First re
se
arch
infrast
r
u
c
ture scheme
s
can be seen in
Figure 1.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 3, Jun
e
2016 : 720
– 728
723
Figure 1. First resea
r
ch infrastru
c
tu
re scheme
s
Third
step,
create data
b
a
s
es a
nd th
en l
oadin
g
data
for te
sting, al
so b
a
cku
p
i
m
age
of
Wind
ows Se
rver. The fo
urth step i
s
te
sting t
he TP
C-C
and
TPC-H. After
com
p
leted,
chan
ge
virtual disk d
r
ive configu
r
at
ion by usi
ng
a re
st
ore ima
ge of Win
d
o
w
s Serve
r
. Ma
ke
configu
r
ati
o
n
of the hard di
sk
with SSD for OS and th
e HDD
for da
tabase. Tem
pora
r
y files O
S
and databa
se
are di
re
cted t
o
HDD in
driv
e F. Install a
VM fr
om a b
a
c
kup ima
ge b
e
fore. After th
at, TPC-C
an
d
TPC-H test similar to prev
ious research
. The
second
rese
arch infrastru
c
tu
re scheme
s
ca
n be
see
n
in Figu
re 2.
Figure 2. Second re
se
arch
infrast
r
u
c
ture
scheme
s
The third re
sea
r
ch, make config
urati
on of HDD for OS and
SSD for databa
se.
Tempo
r
a
r
y files OS an
d d
a
taba
se a
r
e
dire
cted to th
e SSD on dri
v
e F. After th
at, TPC-C a
n
d
TPC-H te
st si
milar to
previ
ous
re
se
arch.
Third
re
se
arch
inf
r
a
s
tru
c
t
u
re schem
es can be se
en in
Figure 3.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Hybrid Disk
Drive Config
uration
on Data
base Server
Virtualization
(Fe
r
dy Nirwa
n
syah
)
724
Figure 3. Third resea
r
ch infrast
r
u
c
ture
scheme
s
Then fou
r
th rese
arch, ma
ke config
urati
on of SSD for OS and HDD for datab
a
s
e. OS
and datab
ase
tempora
r
y files are dire
ct
ed to the HD
D in drive F. Do the testin
g of TPC-C a
n
d
TPC-H simila
r
to previou
s
re
se
ar
ch. F
ourth
re
se
arch infrastructu
re
schem
es
can
be
seen
in
Figure 4.
Figure 4. Fou
r
th re
sea
r
ch infrast
r
u
c
ture
scheme
s
Used metho
d
for this st
udy from creat
e databa
ses to output
generated i
s
usi
ng
Hamm
erDB. There are fo
ur data
b
a
s
e
s
that w
ill be
examined, i
e
ORA
C
LE
XE, SQLServer,
MySQL and PostgreSQL. There are fou
r
config
uratio
ns
of virtual disk d
r
ives to be tested, first, a
virtual disk drive for the OS
using SSDs and virtu
a
l disk drive d
a
taba
se u
s
ing t
he HDD b
u
t the
temporary file
s of Win
d
o
w
s and data
b
a
s
e usin
g the
HDD, se
cond, virtual
di
sk d
r
ive for the OS
usin
g the HDD and virtu
a
l disk d
r
ive d
a
taba
se u
s
in
g SSD but the temporary files of Win
d
o
w
s
and data
b
a
s
e
usin
g the HDD, third, virtual disk
d
r
ive
for the OS u
s
ing th
e HDD and virtual di
sk
drive d
a
taba
se u
s
in
g the
SSD but the
tempor
ary fil
e
s of
Win
d
o
w
s and
data
base u
s
ing
SSD,
fourth, virtual
disk d
r
ive fo
r the OS usi
n
g SSDs
an
d virtual disk drive databa
se
usin
g HDD b
u
t
Wind
ows te
mporary file
s and
datab
a
s
e
s
u
s
ing
SSD. The
r
e
are t
w
o
sch
e
mes u
s
ed
by
Hamm
erDB, TPC-C for OL
TP and TPC-H for OLAP.
Method
s of d
a
ta colle
ction
is to record t
he
re
sults of
tests on fou
r
databa
se
s wi
th four
virtual drive
config
uratio
n
of two different sc
hem
es. TPC-C i
s
calcul
ated by TPM. TPC-H is
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 3, Jun
e
2016 : 720
– 728
725
cal
c
ulate
d
ba
sed
QPh
H
o
f
22 que
rie
s
are
ex
ecute
d
. At each
T
P
C sch
e
me
con
d
u
c
ted two
experim
ents on
e
a
ch
di
sk config
uratio
n and datab
ase
.
So
with thi
s
scena
rio
ha
s b
een
carrie
d
out 64 experiments. TPC-C usin
g 5 wareh
o
u
s
e
for 10 and 50 virtual use
r
, ramp
up time
30
minutes an
d
10 min
u
tes ti
me du
ration.
TPC-H u
s
in
g
SF 1 for
10
and
100 virt
ual u
s
e
r
an
d
1
query
set. In
each test
pro
duce results i
n
log f
ile
s. O
n
Postg
r
eSQ
L
, TPC-H fo
r
query
17, 20,
21
did not incl
ud
e in test beca
u
se it req
u
ir
e
s
quite a lon
g
time for execute the query.
3. Results a
nd Analy
s
is
Four ha
rd
dri
v
e co
nfigu
r
ati
ons for OL
T
P
datab
ase u
s
ing
the
TPC-C sch
e
me,
o
b
tained
databa
se p
e
rforman
c
e
co
mpari
s
o
n
of each co
nfig
u
r
ation. On
O
R
ACLE, 10 v
i
rtual u
s
ers
with
configuration III in
the first rank
, then second configuration I and
configuration II and IV in third
and fourth. Configuration I
and III was using
HDD
as OS
and
SS
D as stor
age for the database.
The differe
nce is the tem
pora
r
y files o
n
the OS
an
d databa
se
configurat
ion I dire
cted to the
HDD but the SSD to configurat
ion III.
Configuration III c
an
ra
ise the performance becomes
18.87%
com
p
ared
to th
e
configuration I.
For 50
virtua
l users
config
uratio
n
condit
i
on of th
e o
r
d
e
r
of perform
a
nce
equal to
10 virt
ual
users. Configurati
on III succ
eeded i
n
raisi
n
g the performance
of 119.04% o
f
the configuration I.
On SQL Server performanc
e configuration III
on 10
virtual users
s
lightly outperformed
the co
nfiguration I, while th
e co
nfi
guratio
n of the IV an
d II in third a
nd fourth. F
o
r 50 u
s
er virtu
a
l
c
o
ndition equal to 10 virtual users
.
Configuration
III slightly outperformed t
he firs
t configuration,
and then
con
f
igure
s
the IV and II in third and four
th.
Not se
en si
g
n
ificant pe
rformance gain
s
of
c
onfiguration III
to I. For th
e 50 virtual us
ers
ac
tually decreas
e perf
ormance 12:14%.
MySQL occur on different conditi
ons, for
10 virtual user’s
c
onfigurati
on I III configuration
outperfo
rme
d
the p
e
rfo
r
m
ance differen
c
e
16:44%, f
o
llo
wed by th
e configu
r
atio
n of II an
d IV in
third an
d fou
r
th. To 50 virtual use
r
s, configurat
ion I
II is supe
rio
r
to 5:55% co
mpared with
the
config
uratio
n I. This wa
s followe
d by
the config
uratio
n of the IV and II.
On Pos
t
greS
QL to 10 virtual us
ers
,
conf
igur
ation III ranked firs
t with
18:43% increas
e
in
perfo
rman
ce
comp
ared
Co
nfiguratio
n I.
Config
uratio
n
IV and
II in t
h
ird
and
fou
r
th. To
50 vi
rtu
a
l
user
configuration III is
superior 21:36% of I. Further
config
urati
on: Configuration II and IV
in
third an
d fourth. Figu
re
5 sho
w
s the
performan
ce of the OL
TP databa
se
on ea
ch di
sk
config
uratio
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Hybrid Disk
Drive Config
uration
on Data
base Server
Virtualization
(Fe
r
dy Nirwa
n
syah
)
726
Figure 5. Gra
ph of TPC-C
per Databa
se
Four
hard d
r
ive config
ura
t
ions for
OL
AP
databa
se
schema
usi
ng TPC-H, o
b
tained
databa
se p
e
r
forma
n
ce compa
r
ison o
f
each
conf
i
guratio
n. On
ORACLE, 10 virtual u
s
ers
c
onfiguration
III rank
ed first with 5.64%
inc
r
eas
e
in the performance of
c
onfiguration I. IV and I
I
in third and f
ourth. Fo
r 10
0 virtual user’
s
co
nfigur
atio
n con
d
ition of
the orde
r of perfo
rman
ce
are
c
onfiguration III rank
ed firs
t
wi
th 14:02% performanc
e improvement
on the configuration I, IV
and II in third and fourth.
On SQLServer, c
o
nfiguration I outperf
o
rm
ed c
o
nfiguration III of performanc
e
on 10
virtual users with pe
rform
a
nce diffe
ren
c
es 2:59%
, while the co
nfiguratio
n II and IV in third and
fourth. For 100 virtual users configuration I out
performed config
uration III 4.87% difference in
perfo
rman
ce,
configu
r
ation
II and config
uration IV in third an
d fourt
h
.
On MySQL,
10 virtual users
c
o
nfiguration
III outperformed confi
guration I with the
perfo
rman
ce
differen
c
e i
s
very slig
htly b
y
1.73%,
follo
wed
by the
configuration
o
f
the IV an
d II in
third and fourth. For
100 virtual
users, configurati
on I is
superior to
configuration III, then
followe
d by the config
uratio
n of the IV and II.
On Pos
t
greS
QL to 10 virtual us
ers
,
c
onf
igurat
ion III rank
ed firs
t followed c
o
nfigurations
I
ran
k
ed
se
con
d
and co
nfigu
r
ation IV and II in third
and fourth. For 10
0 virtual use
r
s co
nfiguratio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 3, Jun
e
2016 : 720
– 728
727
I is
s
u
perior to 3:18%
of
configuration I
II; next,
the configuration I
V
and
II in third
and fourt
h
.
Figure 6 sh
o
w
s the p
e
rfo
r
mance of OL
AP database on ea
ch di
sk
config
uratio
n.
Figure 6. Gra
ph of TPC-H
per Databa
se
4. Conclusio
n
and Futu
r
e
Works
Ba
s
e
d on
th
e d
a
t
a T
P
C-
C
r
e
su
ltin
g fr
om
th
is research, OLTP get
best perfo
rm
ance in
the
third confi
guratio
n
in
which
the OS usin
g the
HDD, data
b
a
s
e
usin
g SSD,
OS an
d d
a
ta
base
temporary file
s u
s
in
g SSD.
On
ORA
C
LE
even i
n
cr
ea
sed 1
19.04
%
at 50 vi
rtual
use
r
s
comp
are
with Co
nfigu
r
ation I. Confi
guratio
n II and IV
is not
recomme
nde
d to use
as
OLTP data
b
a
s
e
serv
e
r
.
In OLAP, b
e
st hard d
r
ive
configuration
s
are
fa
ir
ly
b
a
l
an
c
e
d
be
tw
een
th
e co
n
f
ig
ur
a
t
io
n I
and III. Requi
r
e further test
ing to determ
ine the
better performance amon
g two
configurations.
Config
uratio
n
II and IV are
not re
comm
e
nded to
use
a
s
the O
L
AP d
a
taba
se
serv
er. SQL Se
rver
on OLAP get the highe
st perform
an
ce, this is be
ca
use standa
rd
co
nfiguratio
n pa
ramete
r of SQL
Server is a
b
l
e to make
utilization of
storag
e utilization, mem
o
ry and p
r
o
c
e
s
sor optim
ally.
PostgreSQL i
s
not re
com
m
ende
d for use a
s
an O
L
AP becau
se
there are li
mitations on
the
databa
se e
n
g
i
ne.
To get the
best pe
rform
ance in the
OL
TP data
base se
rver can u
s
e a
hybrid
configuration with the configuration III. In OL
AP, configuration I and
III of benchm
ark dat
a is
impartial, ha
s not seen a
significa
nt differen
c
e. N
eed
to do further testing to Scale Facto
r
(S
F)
10 or
more and num
ber of virtual us
ers.
But this test
will require ha
rdware that i
s
higher than in
t
h
is re
sea
r
ch.
The ratio of the hard drive
is us
ed in this study is 1: 2. W
hen u
s
in
g a hard d
r
ive
with 1: 1
ratio the
conf
iguration III will get a
better ratio
of performance to
the
configuration I compared
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Hybrid Disk
Drive Config
uration
on Data
base Server
Virtualization
(Fe
r
dy Nirwa
n
syah
)
728
this
re
sea
r
ch. But the
s
e
re
sults will
not
be
relev
ant
when
used fo
r
the ratio
wa
s
raised
to 1:3
or
1:4 and mo
re
. Due to the i
n
crea
se in th
e ratio of
u
s
e
of hard di
sk, the
sp
eed
of the HDD
stora
g
e
will be increasingly offset by t
he perform
ance of SSD storage.
Referen
ces
[1]
T
Conno
ll
y an
d
C Begg, “
Data
base Syste
m
s
”
.
Sixth Editi
on, Pearso
n. 201
5.
[2]
RK Lad
a
y
,
H Sukoco a
nd Y N
u
rha
d
r
y
an
i. “Di
s
tr
ibuted S
y
ste
m
and Multim
a
s
ter Repl
icatio
n Mode
l
o
n
Reli
ab
ilit
y Opti
mation
Data
ba
se”.
T
E
LKOMNIKA Indo
nesi
an J
ourn
a
l
of Electrical
En
gi
neer
ing
. 20
15
;
13(3): 52
9–
536
.
[3]
R Schiess
e
r. “IT
sy
stems ma
n
agem
ent”.
(2nd
ed.). Prentice Ha
ll. 2
010: 1
1
0
–12
6.
[4]
B Mao, H Ji
an
g, S W
u
, L T
i
an, D F
eng, J
Che
n
, an
d L Z
eng. “H
pd
a”.
ACM T
r
ans. Sto
r
age
. 20
12
;
8(1): 1–2
0.
[5]
Y Bassil. “A Comp
arative
Stud
y
on t
he
Performanc
e o
f
the
T
op DBMS S
y
stems”.
arXiv Prepr
.
arXiv1
205.
288
9
. 2012: 2
0–3
1
.
[6]
Y Kim, A Gupta, B Urgao
nka
r
, P Berman a
nd A
Sivas
ubr
aman
iam. “H
ybridStor
e
: A cost-efficient,
hig
h
-perform
an
ce storag
e s
y
s
t
em combin
in
g
SSDs and
HD
Ds”.
IEEE Int.
Work. Model.
Anal. Si
mu
l.
Com
p
ut. Telecommun. Syst. - Proc.
2011: 22
7–2
36.
[7]
D Bausch, I Petr
ov an
d A
Buc
h
man
n
. “
On th
e p
e
rfor
manc
e
of data
bas
e q
u
e
ry pr
ocessi
ng
al
gorith
m
s
on flash so
li
d state disks
”. Proc. - Int. W
o
rk.
Datab
a
se E
x
p
e
rt S
y
st. Appl. DEXA. 2
011: 1
39–
14
4.
[8]
J Do, D Zhang, JM Patel, DJ
DeWitt, JF Naughton
an
d A Halverson. “
T
u
rboc
harg
i
n
g
DBMS
buffer
poo
l usin
g SS
Ds
”. Proc. 201
1 Int. Conf. Manag. dat
a - SIGMOD ’11. 201
1
:
1113.
[9]
R Shau
ll, T
Ro
n and A Littma
n
. “Enterprise
Storage Prov
isi
oni
ng
w
i
th F
l
as
h Drive”. 20
10.
[10]
H Jo, Y K
w
on,
H Kim, E Se
o, J Lee and
S Maeng.
“SSD-HDD-
h
y
bri
d
virtual disk i
n
conso
lid
at
e
d
envir
onme
n
ts”.
Lect. Notes
Co
mp
ut. Sci. (inclu
di
ng Su
b
s
er. Lect. Not
e
s Artif. Intell. Lect. Notes
Bioi
nfor
matics)
. 2010; 20
09(6
043 L
NCS): 37
5–3
84.
[11]
X W
u
an
d A
L
N
Red
d
y
. “
Managing storage s
pace in a fl
ash and disk hy
brid storage syst
em
”. Pr
oc.
-
IEEE Comput.
Soc. Annu. Int. S
y
mp. Model. Anal
. S
i
mul.
Comput. T
e
lecommun. S
y
st. MASCOT
S
.
200
9: 610
–6
13
.
[12]
SW
Lee, B Mo
on a
nd C P
a
rk
. “
Advances i
n
flash
me
mory
SSD techn
o
lo
gy for enter
pri
s
e data
bas
e
app
licati
ons
”. Proc. 35th SIGMOD Int. Conf. Manag.
dat
a SIGMOD 09. 2009; 14(3): 8
63–
870.
[13]
K Park, DH L
ee, Y W
oo, G Lee, JH L
ee
and
D
H
Kim. “Reli
a
b
ilit
y a
nd
performa
nce e
nha
nceme
n
t
techni
qu
e for S
S
D arra
y stor
a
ge s
y
stem us
in
g RAID mec
h
a
n
ism”.
20
09 9t
h Int. Symp. C
o
mmun. Inf.
T
e
chno
l. Isc. 2
009
. 20
09: 14
0
–14
5.
Evaluation Warning : The document was created with Spire.PDF for Python.