TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.7, July 201
4, pp
. 5678 ~ 56
8
4
DOI: 10.115
9
1
/telkomni
ka.
v
12i7.527
6
5678
Re
cei
v
ed
De
cem
ber 2, 20
13; Re
vised
Ma
rch 10, 20
14; Accepted
April 2, 2014
Framework of Software Testing Based on Cloud
Computing
Bens
heng Y
a
ng
1
，
Xiang
m
eng Yuan*
2
，
Xiaoguan
g
Huang
3
1
Colle
ge of Re
courses, He
bei
Universit
y
of E
ngi
neer
in
g,
Heb
e
i Ha
nd
an
056
03
8, Chin
a
2,3
Colleg
e
of Informatio
n
an
d Electrical E
ngi
ne
erin
g, Hebe
i U
n
iversit
y
of En
gin
eeri
ng,
Heb
e
i Ha
nd
an
056
03
8, Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
addr
ess:
y
u
an
.xia
ngm
eng
@1
63.com
A
b
st
r
a
ct
F
o
r the pro
b
l
e
m th
at efficie
n
cy is low
an
d
cost
hig
h
ex
i
s
ts in the tra
d
i
tion
al softw
are testing
meth
od, the p
aper tested s
o
ftw
are using
clou
d test
ing techn
o
lo
gy. It i
n
troduc
ed rel
a
ted techno
lo
gi
es
including cloud testing,
an
d
described the
design of
over
all archit
ecture of the system in details,
designed
and i
m
pl
e
m
e
n
ted
th
e
sch
edu
l
i
ng mod
u
le usi
ng
a
hi
gh prior
i
ty
first
sched
ul
ing bas
ed on dyna
mic
pri
o
rit
y
.
T
he res
u
lts of
the Matl
ab s
i
mulati
on
exp
e
ri
me
nt sh
ow
tha
t
this sch
ed
uli
n
g a
l
gor
ith
m
c
a
n re
duce
test c
o
st
and r
eal
i
z
e
th
e
auto
m
ati
on
of softw
are
testing un
der th
e co
nditi
on of si
gn
if
icantly
improv
i
ng test efficie
n
c
y
and res
ourc
e
u
t
ili
z
a
tio
n
.
Ke
y
w
ords
:
cloud
comp
u
t
ing
,
Ta
aS,
H
adoop
, clo
u
d
te
sting
,
coa
l
mine
web
system
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. Introducti
on
Software te
sting is the pro
c
e
ss of executing
a pro
g
ram with the intent of finding errors.
Software
testi
ng a
s
a
n
im
p
o
rtant m
ean
s to gu
ara
n
tee
the q
uality of the
softwa
r
e
gets m
o
re an
d
more
pe
ople'
s attentio
n i
n
the fiel
d
of softw
are
engin
eeri
ng.
The tradition
al ap
pro
a
ch
of
manually cre
a
ting in-h
ou
se testing en
vironme
n
ts that fully mirror the
s
e co
mplexities a
n
d
multiplicitie
s
con
s
um
es
hu
ge capital an
d re
sou
r
ces,
i
t
is se
riou
sly rest
ricte
d
the
developm
ent
of
testing te
chn
o
logy [1]. Clo
ud comp
uting
is a
n
eme
r
gi
ng pa
ra
digm
whi
c
h o
pen
s
a ne
w do
or f
o
r
softwa
r
e te
sting. Clo
ud te
sting are
allo
cated dynam
i
c
ally to cre
a
te
a highly flexib
le and
scalab
le
comp
uting e
n
vironm
ent throu
gh u
s
in
g
virtualizat
io
n
t
e
chn
o
logy
.
Jame
s Whit
t
a
ke
r
p
r
o
s
pe
c
t
ed
the
future
of
softwa
r
e
testi
ng in the book
Expl
orato
r
y Software T
e
sting
, whic
h referred to the
s
o
ftw
ar
e te
s
t
in
g
as a
s
e
r
v
ic
es
ba
se
d o
n
c
l
ou
d co
mpu
t
ing [2]. Di
stin
guished
by th
e way they
are
utilized for testing, four
different
types
role can be pl
ayed by cloud
in the testi
ng [3]. They are
clou
d a
s
syst
em un
de
r te
st, cloud
a
s
te
stwa
re
utility, clo
ud
as test environ
men
t, and
clou
d
as
test logi
stics.
In co
ncrete
contexts, whe
n
t
he clo
ud pl
ays
o
n
seve
ral
rol
e
s simul
t
aneou
sly
the
s
e
types may overlap.
Cloud t
e
sting
can b
e
classifi
ed int
o
three follo
wing types a
c
counting to th
e
role
s clo
ud pl
ayed in the testing
process. (1) Te
st for the clou
d,
it involved the testing p
r
obl
e
m
s
about interna
l
stru
cture, reso
urce conf
igurat
io
n and
function extensi
on an
d so on in
clo
u
d
comp
uting [
4
]. (2) The
m
i
gration
of th
e testin
g, th
e tra
d
itional
test meth
od
s, mana
geme
n
t,
pro
c
e
ss a
nd frame
w
o
r
k a
r
e
migrated to the clo
ud [5]. (3) Te
st other
softwa
r
e sy
stem usin
g clo
u
d
comp
uting. T
he third
kin
d
i
s
mainly int
r
o
duced in
thi
s
pape
r. Te
st b
a
se
d on
clou
d platform
as a
servi
c
e
provided to
custo
m
ers by
Clo
u
d
testin
g se
rvice
p
r
ovide
r
s
a
c
ro
ss
the Internet ca
n well
solve the
pro
b
lems existin
g
in the tradit
i
onal te
st met
hod. In recen
t
years,
clou
d
testing
bega
n
attracting
the
attention
of
aca
demia,
a
nd m
any
re
search
re
sult
s are o
b
taine
d
. D-Cl
oud
i
s
a
large
-
scale
software
clou
d comp
uting
testing
e
n
vironm
ent m
o
del
with fau
l
t injectio
n f
o
r
depe
ndabl
e d
i
stribute
d
system [6]. Cloud9 is a pa
rall
el
symboli
c
e
x
ecution of compute
r
clu
s
t
e
rs
on publi
c
clo
ud infra
s
tru
c
t
u
re
s such as
Amazo
n
EC2
[7]. YETI is
an autom
ate
d
ran
dom
clo
ud –
based testing tool for Java with the ability to
test program
s
wri
tten in different programmi
ng
langu
age
s [8].
The re
st of this pa
per i
s
stru
ctured a
s
follows. Section 2 introd
uce
s
ba
si
c concepts
about
clou
d t
e
sting
and
its
scope, fe
ature
s
, be
nef
its a
nd the
fu
nction
mod
u
l
e
s i
n
cl
uded
i
n
a
nature clo
ud testing syste
m
.
Section
3
pre
s
ent
s
cl
o
ud testin
g sol
u
tion aim at
softwa
r
e te
sting
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fram
ewo
r
k of Software Te
sting Based
o
n
Clou
d Com
puting (Be
n
sheng YANG)
5679
and d
e
si
gns t
he overall st
ructure
of the
system
and
system archite
c
ture. Se
ction
4 analy
s
e
s
a
nd
impleme
n
ts the sche
dulin
g modul
e of system. Se
ction 5 Verify the pe
rform
a
n
c
e of sch
edul
in
g
sub
s
ystem th
roug
h the ex
perim
ent, sh
ows the
con
c
lu
sion
rema
rks an
d scrat
che
s
the futu
re
wor
k
.
2. Cloud Tes
t
ing Ov
er
v
i
ew
With the
exp
ansi
on
of the
scale
of
softwar
e a
nd th
e
co
mplexity o
f
hardwa
r
e,
software
testing en
co
u
n
tered u
n
p
r
e
c
ed
ented
cha
llenge
s. The
cha
r
a
c
teri
stics of clou
d co
mputing in
clu
des
the followin
g
points, allo
ca
tion of
resource
s is dyna
m
i
c, dema
nd fo
r se
rv
ice
s
i
s
cu
st
omiz
ed,
t
h
e
servi
c
e
s
can
be qua
ntified, re
sou
r
ce
s are
po
ole
d
and tran
spare
n
t, whi
c
h ca
n solve
the
probl
em
s. Cl
oud testin
g i
s
ba
sed
on
the clou
d and the
clou
d infra
s
tru
c
tu
re, usi
ng cl
o
u
d
tec
h
nology and s
o
lutions
for s
o
ft
ware tes
t
ing [9].
The enti
r
e te
sting envi
r
on
ments
can
be
config
u
r
e
d
from the cl
oud
on-d
e
man
d
at a co
st
that is pra
c
ti
cal an
d re
asonabl
e due t
o
the pay
-to-use n
a
ture o
f
cloud comp
uting and
with a
lead-tim
e
that
is
nea
r im
po
ssi
ble
within
a compa
n
y’s
own
data
cen
t
er [10].
Clou
d testin
g
serv
ice
provide
r
s pro
v
ide 24/7/36
5 on-dem
an
d automate
d
testing
se
rvice
s
,
it reali
z
es the
re
so
u
r
ce
sha
r
ing; Mo
re importa
ntly, the time to market
of the produ
ct is shorte
n on
the premi
s
e
of
guarantee th
e quality of the Web
syste
m
in
cloud te
sting environ
ment [11, 12]
Not all a
ppli
c
ations
are sui
t
able for te
sti
ng in
the
clo
u
d
. For
som
e
, the co
st of mi
gration
may outwei
g
h the amo
r
tized be
nefits.
Cha
r
a
c
teri
stic
s of an a
ppli
c
ation that can
make it fea
s
i
b
le
for its testing
process to migrate to th
e clou
d
inclu
de: (1) te
st tasks a
r
e ind
e
pend
ent from
one
anothe
r or
whose dep
end
enci
e
s a
r
e e
a
sily identifie
d, this is be
c
ause co
ncurrent executio
n
is
only po
ssible
in thi
s
situati
on, And
spee
dup
w
ill be came
true
th
ro
ugh co
nc
urre
nt execution;
(2)
a self-contai
n
ed and ea
sily
identifiable o
peratio
nal
en
vironme
n
t, in orde
r to kno
w
software a
n
d
hard
w
a
r
e e
n
v
ironme
n
t ne
eded in th
e testing p
r
o
c
e
s
s; and
(3)
a prog
ram
m
atically accessibl
e
interface suit
able for auto
m
ated testing
[13].
A mature cl
o
ud testing
sy
stem shoul
d
incl
ud
e the
following ei
ght types of
function
module
s
:
sca
l
able te
st env
ironm
ent service, mu
lti-te
nant
test
m
o
deling and a
dequ
acy serv
ice,
digital test manag
ement service, on
-de
m
and auto
m
at
ed test and
control
servi
c
e, test sol
u
tion
integratio
n a
nd com
p
o
s
ition se
rvice,
test tr
acking
and monito
r se
rvice, la
rge
-
scal
e test
simulation service, testing
cont
racting and billing
serv
ice [14].
3. Sy
stem Ov
erall Design
3.1. Cloud T
esting Steps
In the process of te
sting o
t
her
softwa
r
e
in the
clou
d, first of all, t
he test
req
u
e
s
ts a
r
e
made
by u
s
e
r
and
sent th
rough
the Inte
rnet to
cl
oud
testing
sy
stem
, and th
en th
ey are a
c
cept
ed
by syste
m
, in
the n
e
xt pla
c
e, test ta
sks
are
sc
he
dule
d
an
d di
sp
atched, mid
d
le
ware
se
rvices
are
p
r
o
v
id
ed
to
tas
k
s
,
vir
t
ua
l res
o
ur
ce
s ar
e
ma
tc
h
e
d a
nd
test tasks a
r
e
executed
an
d supe
rvise
d
. In
the end, test results an
d analytics a
r
e collectin
g
and deliverin
g to use
r
usi
ng web interfa
c
e [15].
Clou
d testing
step
s is sho
w
n in Figure 1.
Figure 1. Clo
ud Testin
g Steps
3.2. Sy
stem
Architec
ture
Design
Thre
e se
rvice model
s are defined in
cloud co
mp
uting, Infrastructure as a
Service
(IaaS), Platfo
rm a
s
a Servi
c
e (P
aaS), a
nd Softw
a
r
e as
a
Se
rvice (SaaS). Clou
d
testing can
be
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: 5678 – 56
84
5680
rega
rd
ed a
s
SaaS se
rvice
model u
s
ing
in the fi
eld of softwa
r
e t
e
sting.
We d
e
sig
n
the cl
o
ud
softwa
r
e te
sting system
(CSTS) a
r
chitecture in
the
light of that three service model
s wit
h
combi
n
ing
with the step of clou
d testing.
Cloud te
st
ing
system archi
t
ecture i
s
sho
w
n in Figu
re
2.
Figure 2. System Archite
c
t
u
re Di
ag
ram
(1) T
he IaaS layer
Processo
r, virtual
stora
ge,
net
wo
rk and
other i
n
fra
s
tructure resource
s a
r
e p
r
o
c
essed to
logical re
so
urce p
ool by u
s
i
ng vi
rtuali
z
ati
on technol
og
y to provide t
o
user in
the f
o
rm of
se
rvices,
whi
c
h i
s
in
a
unified, centralize
d
patte
rn
. Users
ma
ke
the reque
st
s to the
CSCT
for all
kin
d
s
of
resou
r
ces a
c
cording
to th
e
i
r o
w
n
re
qui
re
ment, an
d d
o
not ne
ed t
o
care
ab
out h
o
w
the
resou
r
ces
a
r
e a
lloc
a
te
d a
n
d
a sc
he
du
le
d
.
In th
is
w
a
y, th
e
utili
zation
of
hardwa
r
e
and
software
reso
urces
are imp
r
oved
and testin
g proce
s
s be
com
e
s mo
re intell
igent and a
u
tomated.
(2) T
he PaaS
layer
PaaS is a busine
s
s infra
s
tructu
re platfo
rm
of softwa
r
e developm
e
n
t with the purpo
se of
providin
g cu
stomers with
unified an
d cu
stomiz
ed d
e
velopme
n
t of middle
w
are platform, a
n
d
inclu
d
e
s
the
manag
eme
n
t of infrastruct
u
re reso
u
r
ce
s and te
st ta
sks
submitte
d by use
r
at
th
e
same time. T
h
is laye
r is
compri
se
d of test task ma
n
ageme
n
t mo
dule, middl
e
w
are man
age
ment
module
and t
e
st re
so
urce
manag
eme
n
t module. It impleme
n
ts th
e sched
uling
and allo
catio
n
of
vir
t
u
a
liz
a
t
ion in
fr
as
tr
uc
tur
e
in
th
e
vir
t
u
a
l
reso
urce m
ana
g
e
ment m
o
d
u
le. Mid
d
le
ware
manag
eme
n
t modul
e a
c
hi
eves the
follo
wing fu
nctio
n
s
, safety man
ageme
n
t, SLA (Service L
e
vel
Agreem
ent) p
e
rform
a
n
c
e
monitori
ng an
d re
sult coll
e
c
ting an
d an
a
l
yzing. Te
st task man
age
ment
module i
s
divided into u
s
er
manag
eme
n
t and test tasks sche
dulin
g manag
eme
n
t.
In the virtual
reso
urce m
a
n
ageme
n
t mod
u
le, all
kind
s
of virtual reso
urces may b
e
in two
states: i
d
le
state and
runni
ng
stat
e. All
kind
s of
re
so
urces a
r
e
st
o
r
ed i
n
to two li
sts
acco
rdin
g
to
their state
s
. The virtual re
sou
r
ces
co
nform to t
he re
quire
ment
s a
r
e allo
cated t
o
the test tasks
compl
e
ting th
e sched
uling
pro
c
e
ss b
a
se
d on re
sou
r
ce
s state.
(3)The Ta
aS layer
Acco
rdi
ng to
the test req
u
irem
ent su
b
m
itt
ed throug
h acce
ss int
e
rface, free
matchin
g
softwa
r
e
an
d
ha
rd
ware a
r
e cho
s
en
an
d in
stalled
to
build
the
target test
environment,
whi
c
h is
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fram
ewo
r
k of Software Te
sting Based
o
n
Clou
d Com
puting (Be
n
sheng YANG)
5681
maintaine
d
a
nd up
dated
by CSTS. Users
enjoy
th
e rig
h
t to u
s
e the
softwa
r
e, and
ca
n a
l
so
contin
uou
sly upgrade.
In addition,
use
r
inte
ra
ct with CST
C
throug
h the
web
bro
w
se
r, thin client
s explore
client
s o
r
p
r
ogra
mming
a
nd othe
r
wa
ys, sub
m
it
the test requ
e
s
t, co
nsu
m
e
the test
servi
c
e
provide
d
by CSCT, an
d finally get the test re
sult
s.
4. Scheduling Model Des
i
gn and Implement
Duri
ng the
cou
r
se of re
sou
r
ces
man
ageme
n
t in
clou
d testing
,
parallel te
st task
sched
uling a
nd dispatchin
g algo
rithms
is co
mp
li
cate
d becau
se of
the dynamic nature of th
e
infrast
r
u
c
ture
used
in
clo
ud e
n
viron
m
ent, whi
c
h
d
o
not
o
c
cur i
n
othe
r te
sti
ng m
e
thod
s
[16].
Various
scheduling
strategy will lead to com
p
le
x technical, securi
ty
issues and
determine the
executio
n se
quen
ce
of test tasks
and
whi
c
h vi
rtua
l machi
ne
wi
ll be dispatched to the t
a
sk
sorte
d
. The sche
duling m
o
dule was d
e
si
gned a
nd imp
l
emented of
CSTS in this
pape
r.
4.1. Scheduling Module Struc
t
ure
After user su
bmits te
st ta
sks,
it is an
in
di
sp
en
sable
part to
p
r
edi
ct the n
u
mbe
r
of virtual
machi
ne
nee
ded
acco
rdin
g to the
num
ber
of tasks
and th
e a
nal
ysis
of the te
st p
r
oje
c
t. In
this
way, the wa
ste in requ
esti
ng exce
ssive
virtual
mach
ines a
nd a
p
p
l
ying to the cloud resource
manag
er be
cause virtual
machi
ne i
s
not en
oug
h i
n
the
process of
te
st executio
n can
be
redu
ce
d and
avoid.
Sched
uling sub
s
yst
e
m modul
e st
ructu
r
e i
s
sh
o
w
n in Figu
re
3.
Figure 3. Sch
edulin
g Modu
le Structu
r
e
4.2. The Alg
o
rithm Proc
ess
Test ta
sk
sui
t
ed to
clo
ud
testing
are in
depe
ndent
from o
ne
anot
her,
so
we t
a
ke
no
accou
n
t of th
e dep
end
en
cy relation
ship
betwe
en ta
sks
du
ring th
e
cou
r
se of te
st task sch
eduli
ng
in the clou
d testing pl
atform, this is a si
mplification to
the sch
eduli
ng algo
rithm.
The
sched
uli
ng alg
o
rithm
we
use
d
in
th
is p
ape
r i
s
th
at the ta
sk
wi
th high
pri
o
rit
y
is firs
t
execute
d
, wh
ich i
s
b
a
sed
on dyn
a
mic p
r
iority, we
cal
l
ed it HDPF
algorith
m
, an
d takes the t
a
sk
waiting
time,
task exe
c
utio
n time a
nd ta
sk weight i
n
to
co
nsi
deratio
n. That i
s
to
say the p
r
io
rity of
tasks
will
be
cha
nge
d fo
r t
he ta
sk exe
c
utive co
ndi
tion
or
th
e inc
r
ea
s
e
o
f
w
a
iting
time
, an
d th
e
high pri
o
rity task is first execute
d
.
(1) T
he n
u
m
ber of ta
sks
and the
num
ber of
virtu
a
l machi
ne
a
r
e assume
d
to be
finite.
There is a
list of tasks
T
,
12
n
{
,
,
...,
}
TT
T
T
performed by a
set of machi
n
e
s
M
,
12
m
{
,
,
...,
}
M
MM
M
.
A
ny
t
a
sks
can be ex
e
c
ut
ed in an
y
v
i
rt
ual machi
ne.
Ho
w
e
v
e
r,
interruption i
s
unwarranta
b
l
e
befor
e the t
e
st p
r
o
c
e
ss i
s
complete
d.
Any task
ca
n
not be
split in
to
smalle
r
subta
s
ks. Every time after the f
i
rst sch
eduli
n
g pro
c
e
s
s, the numb
e
r of
waiting exe
c
u
t
ion
tasks a
r
e a
ssi
gned to n, an
d the numbe
r of virtual machin
e are a
ssigned to m.
Given a defi
n
ition of two
queue
s: WaitList
point
s the test task waiting qu
eue an
d
FreeVM
L
ist d
enote
s
free virtual ma
chin
e
queue.
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: 5678 – 56
84
5682
(2) Te
st tasks a
r
e
su
bmitted by u
s
e
r
th
roug
h Inte
rne
t
interface, a
nd the
wei
ght
value i
s
assign
ed to t
a
sk by u
s
e
r
according to
the prom
pt at
the sa
me ti
me. Then th
e
tasks a
r
e
ad
ded
into task
waiti
ng que
ue Wa
itList.
The value of
a task i
s
de
cide
d by use
r
s
whe
n
they submit test j
ob. Accordi
n
g to the
waiting time,
the executio
n time and t
he wei
ght of
task, its p
r
io
rity can b
e
calcul
ated by
the
following formula.
jj
j
j
j
qd
p
w
d
The te
rm j
de
notes the jth
task an
d i i
s
t
he ith virtu
a
l
machi
ne.
j
q
de
notes the
wai
t
ing
time of the jth tas
k
and the s
u
bs
cript j refers
to a tas
k
,
j
d
denote
s
the
execution ti
me of the jth
t
a
sk,
j
w
de
note
s
the
weight
of the jth ta
sk a
nd
j
p
den
otes th
e jth ta
sk p
r
io
rity.
de
notes a
para
m
eter
wh
ich i
s
impo
rte
d
to adju
s
t the weig
ht between
jj
j
qd
d
and
j
w
t
o
make t
h
em in
t
he
same o
r
d
e
r o
f
magnitude.
(3) T
he nu
mb
er an
d co
nfig
uration
of virtual
ma
chine i
s
fore
ca
sted
based on the
analysi
s
of proj
ect n
e
eded to
test
requi
rem
ent
and the
num
ber
of test ta
sk u
s
er sub
m
it. CSCT
send
s
messag
e to the clo
ud resource ma
nag
er to apply
fo
r the virtual
machi
ne. Th
e con
s
tructio
n
of
test environ
ment is com
p
leted thro
ug
h deploy
ing t
he ope
rating
system, software syste
m
and
hardware sy
stem and
so
on in light
of environm
ent configuration
t
e
mpla
te
and this will reduce
the req
uest o
f
the new re
source
s in
the
pro
c
e
ss
of test executio
n.
Of cou
r
se, there a
r
e e
rro
rs in
this pre
d
ictio
n
on the ba
si
s of experie
n
c
e an
d templ
a
te.
(4) Th
e n
u
mb
er
of tasks i
n
WaitLi
st qu
eu
e in
assig
ned
to the
n, a
n
d
the
numb
e
r
of tasks
in FreeVM
L
ist is assi
gne
d to the m.
(5) If
nm
, CS
CT
rele
ases the
re
dund
ant p
r
ocesso
rs, th
e virtual
ma
chine
s
can
be
assign
ed to the tasks, an
d
be relea
s
e
d
imm
ediately a
fter the test tasks pe
rforme
d.
(6) If
nm
, CTCS calculate
s
prio
rity of all ta
sks in
a
c
cord
an
ce
wi
th the prio
rity
formula,
so
rts them
a
c
cording to th
eir
prio
rity
, sele
cts the to
p m
tasks
and
assign
s th
em t
o
m
virtual ma
chi
nes. T
a
sks
which
we
re
ch
ose
n
a
r
e
re
moved from t
he qu
eue
WaitList a
nd vi
rtual
machi
n
e
s
whi
c
h are executing are
remov
ed from the q
ueue F
r
eeVM
L
ist.
(7) The
ne
w
tasks
are
ad
ded to th
e q
ueue
WaitLi
st, which a
r
e
submitted
by
user
and
tasks the
r
e
are any fault o
c
curre
d
in thei
r execut
io
n proce
s
s. Fre
e
virtual ma
chi
n
es a
r
e
adde
d
to
the queu
e FreeVMLi
s
t wh
en they turn u
p
. Execute
st
ep 4 until all tasks a
r
e
carri
ed out pro
p
e
r
ly.
4.3. Algorith
m
Flo
w
Char
t
The a
d
vanta
ges
of the
HDPF alg
o
rith
m: (1) If
nm
, there i
s
no
nee
d
to con
s
id
er t
h
e
prio
rity of each task, the total executio
n time is
the
runnin
g
time
of the task with the long
est
runni
ng tim
e
, in oth
e
r
wo
rds, it n
eed
s
(1)
O
time;
If
nm
,
t
he
sort
in
g t
a
ke
s
(l
o
g
)
On
n
time
and the loo
p
need
s
(l
o
g
m
)
On
, as a con
s
eq
ue
nce, the tota
l of the amount of time is
(l
o
g
l
o
g
)
(l
o
g
)
On
n
n
m
O
n
n
. (2) Virtual m
a
chi
n
e
s
are
relea
s
ed a
s
soon a
s
ea
ch task are
compl
e
ted resulting in the resour
ce
utilization of
cloud platform
are improved.
The tasks with
long
waitin
g
time, sh
ort
e
x
ecution
time
and
h
eavy
weig
ht take
pre
c
ed
en
ce
on exe
c
utio
n. Of
cou
r
se, the p
r
iority of all ta
sks mu
st be
reco
unt
ed b
e
f
o
re eve
r
y sch
edulin
g proce
ss, a
nd thi
s
will
increa
se the
co
st of the system. The al
g
o
rithm flow
chart is
sho
w
n
in Figure 4.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Fram
ewo
r
k of Software Te
sting Based
o
n
Clou
d Com
puting (Be
n
sheng YANG)
5683
Figure 4. Algorithm Flo
w
Cha
r
t
5. Experimental Verifica
tion
The
algo
rith
m HDPF
pro
posed i
n
thi
s
pap
er
and
F
C
FS al
gorith
m
are
cod
e
d
in M
a
tlab
and an
alyze
d
operatio
n efficien
cy from the persp
ec
tiv
e
of mathem
atical thro
ugh
the Simulation
Experiment.
Assu
ming
tha
t
the num
ber
of virtual ma
chine
s
is 25
0,
the num
be
r o
f
test tasks n
is
valued re
sp
e
c
tively {50, 150, 250, 350
, 450}, the
execution time
of each test
task is n
o
t the
same, th
e time of the ta
sk with
the lo
nge
st ex
ecution time i
s
12
ms, As
ca
n b
e
se
en from
the
Figure 5:
(1
) I
f
the n
u
mbe
r
of the virtu
a
l
machi
n
e
s
i
s
greate
r
th
an t
he n
u
mbe
r
of tasks,
the tot
a
l
turna
r
ou
nd ti
me of the
HDPF
sche
d
u
ling al
gorith
m
are
co
nst
ant, and F
C
FS algorith
m
is
increa
sed
wh
en ru
nning th
e sam
e
amo
unt of functi
o
nal test and
perfo
rman
ce
testing. (2
) If the
numbe
r of virtual ma
chin
e
s
le
ss than t
he nu
mbe
r
of
tasks, the
to
tal turna
r
ou
n
d
time of
HDPF
sched
uling
al
gorithm
in
cre
a
se
s gently and FCFS
al
gorithm
qui
ckly. (3
)
With
the in
crea
se
of
amount
of te
sting ta
sks, t
he exe
c
utive
efficien
cy of t
he
HDPF
alg
o
rithm i
s
b
e
tter tha
n
F
C
F
S
of
almost 20%,
having a di
stinct advanta
g
e
.
Figure 5. Total Turn
over T
i
me Com
pare
d
0
50
10
0
15
0
20
0
250
300
350
400
45
0
50
0
5
10
15
20
25
T
a
sks n
u
m
b
e
r
T
o
ta
l
tu
r
n
a
r
o
u
n
d
ti
m
e
(
s
)
FC
FS
HDP
F
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: 5678 – 56
84
5684
Con
c
lu
sion
remarks ca
n
be given from the anal
ysis results:
comp
are
d
with the
traditional te
st methods, th
ere i
s
no n
e
e
d
to make
an
investment i
n
software te
sting, such a
s
expen
sive te
sting tool
s, th
e buildi
ng, m
a
intena
nce a
nd up
gra
de o
f
the test envi
r
onm
ent an
d
so
on, whe
n
ru
n
the same
a
m
ount of fun
c
tionality te
sti
ng tasks
and
performan
ce
testing task.
In
addition, the t
e
st result rep
o
rt can b
e
go
t soon
a
fter t
he test ta
sks
are
sub
m
itted
.
Test efficie
n
cy
are imp
r
oved
signifi
cantly.
6. Conclusio
n
Combi
ned
wi
th cloud
com
puting techno
logy, this
pap
er intro
d
u
c
e
s
the stru
cture
of the
clou
d
testin
g system ba
se
d
on clou
d computing
mo
del fram
ework in
detail, an
d implem
ent
the
sched
uling m
odule
s
of the
system by u
s
ing a hi
gh p
r
iority tasks fi
st sched
uling
base
d
on the
dynamic priority. Compared with the c
l
ass
i
c
F
C
FS sched
uling al
gorithm, HDP
F
algorithm can
signifi
cantly improve the t
e
st effi
cien
cy of the softwa
r
e system, an
d redu
ce te
st co
st throug
h the
test experim
e
n
ts in Matlab.
In CSTS, the implem
entati
on of other m
odule
s
an
d the safety of the
system i
s
the focu
s of t
he next step wo
rk, further re
se
arch sh
ould b
e
deal with.
Ackn
o
w
l
e
dg
ements
This
pap
er
suppo
rted
by the
con
s
tru
c
t
prog
ram
of th
e key discipli
ne in
He
bei
provin
ce
and the natu
r
al sci
en
ce fou
ndation of He
bei provi
c
e
(
G
r
ant No. E20
1140
2046
)
Referen
ces
[1]
Vina
ya
Ku
mar M
y
l
a
v
a
rap
u
. T
a
king testi
n
g
to the cloud.
Cogn
iza
n
t.
http://
w
w
w
.
co
g
n
iza
n
t.com/insi
ghts/persp
ectiv
e
s/taking-testi
n
g-to- the-clo
ud.
[2]
James A, Whittaker. Explor
ato
r
y
Soft
w
a
r
e
T
e
sting. Addis
on-
W
e
sle
y
Pr
ofes
sion
al. 20
09.
[3]
Victou Cz
ente
r
, Performanc
e testing
me
ets
the cl
ou
d op
portu
nitie
s
and c
h
a
lle
nges. SQS
,
http://
w
w
w
.
sqs.
com/ en-gro
up/
_do
w
n
l
o
a
d
/W
hite_Pa
per
_Perf
o
rmanc
e_T
esting_
Clo
ud
_EN.
pdf
[4]
Cha
n
W
K
, Lij
un Mei, Z
h
e
n
y
u Z
h
an
g. Mode
lin
g an
d testing
of clou
d ap
plic
ations.
In Service
s
Comp
uting C
o
nferenc
e.
IEEE Asia-Pacific
; 2
009; 11
1-1
18.
[5]
T
auhida Parv
een, Scott T
ille
y
.
W
h
e
n
to Migrate Softw
are T
e
sting t
o
the Clo
ud.
In the T
h
ird
Internatio
na
l C
onfere
n
ce
on
Soft
w
a
re T
e
sting, Ve
rific
a
tio
n
,
and V
a
li
datio
n W
o
rksho
p
s (ICST
W
). 2010
;
424-
427.
[6]
T
a
kayuki Ba
n
z
ai, Hitosh
i
K
o
izumi, R
y
o
Kanb
a
y
as
hi.
D- Clo
ud: De
sign of a S
o
ftw
are T
e
stin
g
Enviro
n
m
ent f
o
r Rel
i
a
b
le
Di
stributed S
yste
m
s Us
ing
Clo
ud C
o
mputi
n
g
T
e
chno
logy.
Procee
din
g
s
o
f
IEEE/ACM Internatio
nal C
onfe
r
ence o
n
Clust
er,
Clou
d
an
d Grid Comp
utin
g, 2010, p
p
. 63
1-63
6.
[7]
Liviu
Ci
ortea,
Cristian Z
a
mfir
, Stefan Bu
cu
r. Clou
d9: A
Soft
w
a
re T
e
sting Serv
ice.
A
C
M SIGOPS
Operatin
g Systems R
e
vi
ew
.
2010; 43(
4): 5-1
0
.
[8]
Manu
el Ori
o
l,
F
aheem
Ul
la
h. YET
I on the
clo
ud.
Pr
oce
edi
ngs of
th
e Internatio
na
l W
o
rkshop
o
n
Soft
w
a
re T
e
sting in the C
l
o
u
d
.
2010: 43
4-43
7.
[9]
W
i
kiped
ia. Cl
o
ud testing. http
://en.
w
i
ki
pe
dia
o
rg/
w
iki/C
l
o
ud
_testing. 2
013;
8.
[10]
A Vanith
a Kath
erin
e, K Alag
ar
sam
y
. So
ft
w
a
r
e
testing
in cl
o
ud p
l
atform: a
surve
y
. Int
e
rna
t
iona
l Jour
na
l
of Computer A
pplic
atio
ns
. 20
12; 46(6): 2
1
-2
4.
[11]
Pri
y
ank
a, Inde
rveer C
han
a,
Aja
y
R
a
n
a
. Empiri
ca
l Eva
l
u
a
tion
of Cl
ou
d-
base
d
T
e
sting
T
e
chniques:
A
S
y
stematic R
e
vie
w
.
ACM SIGSOF
T
Softw
a
re Engi
ne
erin
g Notes.
201
2; 3
7
(3): 1-4.
[12] Jerr
y
Gao,
Xi
ao
yi
ng
B
a
i, W
e
r-T
ek
T
s
ai. Clou
d
T
e
sting- I
ssues, Ch
al
len
ges, Ne
eds
a
nd Practic
e
.
Softw
are Engin
eeri
ng: An Inte
rnatio
nal Jo
urn
a
l (SEIJ).
2011
; 1(1): 9-23.
[13]
Scott T
ille
y
,
T
auh
ida
Parv
ee
n. Soft
w
a
re
Testin
g
in
the
Clou
d
: Mi
grati
on
and
E
x
ec
u
t
ion. Spr
i
ng
er,
201
2.
[14] Jerr
y
Gao,
Xi
ao
yi
ng B
a
i, W
e
r-T
ek
T
s
ai.
T
e
sting
as
a S
e
rvice (T
a
a
S)
on C
l
o
uds.
Pr
ocee
din
g
s of
Service- Orie
nted S
y
stem En
g
i
ne
erin
g. 201
3: 212-2
23.
[15]
LVD Aa
lst, “Soft
w
ar
e testin
g
as a serv
ice
(ST
aaS). http://
w
ww
.
t
ma
p.net
/Images/Pap
er%20ST
aaS
_
tcm8-479
10.p
d
f
.
[16]
Micha
e
l L P
i
n
e
do. Sche
du
lin
g
:
T
heor
y
,
A
l
gor
ithms an
d S
y
stems. T
h
ird Edi
t
ion.
Spri
ng
er.
200
8: 11
8-
130.
Evaluation Warning : The document was created with Spire.PDF for Python.