Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
5, N
o
. 1
,
Febr
u
a
r
y
201
5,
pp
. 15
8
~
16
5
I
S
SN
: 208
8-8
7
0
8
1
58
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
Service Level Agreement
s
in
Cloud Computing and Big Data
K. Radh
a
1
, B.
Thiruma
l
a Rao
2
, S
h
aik
Mas
t
han B
a
bu
3
,
K
.
Thirup
athi
Rao
4
,
V.
Kris
hna Re
dd
y
5
,
P. Sa
ikiran
6
1, 2,4,5,6
D
e
pt. o
f
C
S
E, K
L
U
n
ivers
i
t
y
, G
untur
, Ind
i
a
3
Sri Sai
Educational Societies, I
ndia
e
m
ail: rad
h
a
.saitej
@
g
m
ail.co
m
1
, th
ir
u
m
ail
@
yaho
o.co
m
2
, bo
bb
ycr
eativ
e.m
a
sth
a
n
@
g
m
ail.co
m
3
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Sep 15, 2014
Rev
i
sed
No
v
30
, 20
14
Accepted Dec 18, 2014
Now-a-day
s
Mo
st of the
industr
ies ar
e hav
i
ng large volumes
of
data. Data
has range of Ter
a
b
y
tes to Peta by
te
. Organ
i
zatio
ns are looking to
handle th
e
growth of data.
Enterpr
i
ses are
usi
ng cloud d
e
p
l
o
y
ments
to add
r
ess the big
data
and an
aly
tics with respect to th
e in
ter
action
between
cloud
and big data.
This pap
e
r pres
ents big d
a
ta issues
and research dir
ections towards th
e
ongoing work of
processing
of big data
in th
e d
i
stributed
environ
m
ents.
Keyword:
An
alytics
Big
Data
Clo
u
d
Di
st
ri
b
u
t
e
d En
vi
r
onm
ent
Service Level
Agreem
ents
WSL
A
Copyright ©
201
5 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
K R
a
dha
,
Depa
rtem
ent of C
o
m
puter Sc
ience and
Engi
neeri
n
g,
KL Uni
v
er
sity
,
Gree
n Fi
el
ds, Vad
d
es
waram
,
G
unt
ur
, 52
2
5
0
2
, I
ndi
a
e-m
a
i
l: rad
h
a
.saitej
@
g
m
ail.co
m
1.
SERV
ICE LE
VEL A
G
REE
M
ENTS
OF
CLOU
D
CO
MP
UTIN
G
Worldwid
e en
t
e
rprises are m
o
v
i
n
g
t
o
ward
s Big
d
a
ta to
pred
ict th
e in
sigh
ts fro
m
th
e rev
o
l
u
tio
n
of
inform
ation created from
different s
o
urce
s [1
].
Data is in
creasin
g rap
i
d
l
y
w
ith
t
h
e in
crease in
d
i
g
ital
wo
rl
d
’
s
Pet
a
by
t
e
s o
f
d
a
t
a
from
Soci
a
l
M
e
di
a. B
i
g D
a
t
a
has g
r
ow
n i
n
2
0
10
US
$
3
.
2
and
i
s
ex
pect
e
d
t
o
gr
o
w
i
n
2
0
1
5
i
s
US$
1
6
.
9
b
illio
n [2
]. Scale-o
u
t
u
s
es co
llectio
n
o
f
st
orag
e serv
ers t
o
d
e
liv
er m
o
re cap
acity wit
h
same
o
p
e
ration
a
l co
st. In
2
020
th
e
u
n
i
v
e
rse
will produ
ce 50
ti
m
e
s o
f
i
n
fo
rm
atio
n
,
7
5
tim
es o
f
d
a
ta sto
r
ag
e serv
ers
[3
].
Acco
rd
ing to
th
e
Gart
n
e
r
repo
rt, b
y
2
020
,
2
3
0
b
illio
n
dev
i
ces are conn
ected
t
o
th
e i
n
tern
et. Sm
art meters,
sens
ors
,
act
uat
o
rs c
ont
i
n
u
o
u
s
l
y
send m
a
ssi
v
e
am
ount
o
f
da
t
a
t
h
at
shoul
d be anal
y
zed an
d st
ore
d
. T
o
da
y
Dat
a
Sci
e
nt
i
s
t
s
are u
s
i
ng
Yot
t
a
by
t
e
t
o
desc
ri
be
ho
w m
u
ch g
ove
r
n
m
e
nt
dat
a
t
h
e NA
SA
have
o
n
pe
o
p
l
e
t
oget
h
er
. I
n
th
e n
e
ar fu
ture, Bron
tob
y
te
will b
e
th
e m
e
asu
r
em
en
t to
describ
e
t
h
e type o
f
sen
s
or d
a
ta will b
e
g
e
nerated
fr
om
t
h
e Int
e
r
n
et
of T
h
i
n
gs (
I
oT
).
10
24
Yo
tt
ab
yte will b
e
o
u
r
Di
g
ital u
n
i
verse to
d
a
y,
(250
trillio
n
o
f
DVDs)
and 1
0
27
Bron
to
b
y
te will b
e
o
u
r d
i
g
ital u
n
i
v
e
rse t
o
m
o
rro
w. Clou
d
co
mp
u
ting
is a busin
ess fram
ewo
r
k
on
whic
h re
sourc
e
s can s
h
are the resources
over the i
n
ternet
i
n
a
pay
-
pe
r p
a
t
t
e
rn. Due
t
o
rapi
d
G
r
owt
h
of dat
a
with
em
erg
i
ng app
licatio
n
s
th
ro
ugh
t
h
e
S
o
ci
al
m
e
di
a, sem
a
nt
i
c
web
anal
y
s
i
s
, B
i
oi
nf
orm
a
t
i
c
s net
w
o
r
k
anal
y
s
i
s
di
ve
r
s
i
t
y
of dat
a
i
s
pr
od
uce
d
. M
a
nagi
ng t
h
e d
a
t
a
and a
n
al
y
s
i
s
of m
a
ssi
ve dat
a
i
s
t
h
e b
i
ggest
chal
l
e
ng
e.
Gar
t
ner
def
n
e
d
B
i
g Dat
a
as
H
u
ge-
V
ol
um
e,
Huge
-
V
el
oci
t
y
and
H
uge
-
V
ari
e
t
y
of dat
a
se
t
s
are
neede
d
i
n
ne
w
fo
rm
s for p
r
o
cessi
ng e
ffi
ci
e
n
t
deci
si
o
n
-m
aki
n
g an
d
pr
oc
ess o
p
t
i
m
i
zat
ion
.
Fr
om
201
2-
2
0
2
0
,
Western
Eu
rope’s
Dig
ital Un
iv
erse Gro
w
t
h
rate is 53
8 Exabyes to
5
.
0
Zett
ab
yte [7
].
I
n
t
h
e co
m
p
u
tin
g
wo
r
l
d Clou
d co
m
p
u
ting
is n
e
w bu
sin
e
ss p
a
r
a
d
i
g
m
. A
ccord
ing
to
N
I
ST, Cloud
com
put
i
ng i
s
para
di
gm
for
al
l
o
wi
n
g
pe
rv
asi
v
e o
n
-
d
em
and
net
w
or
k ac
cess t
o
col
l
ect
i
on o
f
co
nfi
g
u
r
abl
e
com
putational resources
that can be providi
n
g
with
se
rv
ice pr
ov
id
er
i
n
ter
actio
n. Cloud
ser
v
ices pr
ov
i
d
e th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Service Level
Agree
m
ents
in
Cloud Computi
n
g and Bi
g
Dat
a
(
K
Ra
d
ha)
15
9
reliab
ility an
d
m
a
n
a
g
eab
ility in
a d
y
n
a
m
i
c
world
.
Clou
d co
m
p
u
tin
g
characteristics are as fo
llo
ws,
Rap
i
d
elasticity, Pool
of Res
o
urces,
Broa
d net
w
ork access, on-d
e
m
and self se
rvice and
deliberate service of [1]-[4]
.
Because of
ra
pid growt
h
of
C
l
oud C
o
m
puting Market, it is
providing t
h
e new
services
with
t
h
e
inte
ra
ction of
clo
u
d
serv
ice prov
id
ers and
serv
ices.
Serv
ice lev
e
l ag
r
eem
en
t is p
r
ov
id
ed
for th
e cu
sto
m
ers who
are
u
tilizin
g
th
e clo
u
d
serv
i
ce to
p
r
o
v
i
d
e
th
e qu
ality o
f
serv
ice.
In
th
e
Dyn
a
m
i
c d
i
g
i
t
a
l en
v
i
ro
n
m
en
t, serv
ices are prov
ided
st
ri
ct
l
y
t
o
an o
n
-
d
em
and. A
g
r
e
em
ent
i
s
prov
i
d
ed am
ong t
h
e cl
oud se
rvi
c
e
pro
v
i
d
ers,
br
o
k
ers
,
cust
om
ers. SL
A
i
s
a l
e
gal
bond
bet
w
een t
h
e s
e
rvi
ce p
r
o
v
i
d
e
r
and cl
i
e
nt
. S
e
rvi
ce l
e
vel
A
g
reem
ent
(SL
A
) ha
s go
al
s t
h
r
o
ug
h
Qu
ality of
Serv
ice (Qo
S
) attribu
t
es,
Qu
ality o
f
pro
t
ec
tio
n
attribu
t
es, descrip
tion
o
f
actio
n
s
t
o
p
r
ov
id
e t
h
e
service acc
ording to the
QoS attributes etc. SLA
is
designe
d
to c
r
eate the awa
r
enes
s on QoS and
Resp
on
si
b
ilitie
s. SLA en
ab
les th
e en
d-u
s
ers to
ag
ree on
wh
at k
i
nd
o
f
serv
ices are
o
ffered,
h
o
w t
h
ese services
will b
e
d
e
liv
ered
and
who
will b
e
respo
n
sib
l
e fo
r th
e
serv
ice ex
ecu
tion, serv
ice in
terru
p
tion
s
and
priv
acy
aspect
s. SL
A i
s
not
p
r
o
v
i
d
i
n
g
t
h
e assura
nce
abo
u
t
t
h
e ser
v
i
ce expi
ry
[
4
]
.
C
l
i
e
nt
orga
ni
z
a
t
i
on nee
d
t
o
con
f
i
r
m
th
at th
e au
thorized
d
a
ta to
ou
tsid
e resou
r
ces sho
u
l
d
k
e
p
t
to
th
e sam
e
h
i
gh
lev
e
l stan
d
a
rd
s if th
e
d
a
ta is
co
n
t
ro
lled
i
n
tern
ally. If th
ere is v
i
o
l
atio
n
o
f
security th
e v
e
ndo
r m
a
y
leg
a
lly resp
on
sib
l
e
b
u
t
th
e client
or
ga
ni
zat
i
on
m
a
y
affect
. C
l
i
e
nt
o
r
ga
ni
zat
i
o
n
sh
o
u
l
d
u
n
der
s
tand
th
e busin
ess
p
r
ac
tices with
org
a
n
i
zatio
n
a
l
standa
rds.
To
ho
st d
a
ta in
to
th
e clou
d, stan
d
a
rd
ag
reemen
ts su
ch
as co
nfid
en
tiality,
serv
ice lev
e
l ag
reem
en
t
s
need t
o
be ex
t
e
nde
d [
26]
.
C
l
ou
d com
put
i
ng g
o
v
er
na
nc
e el
em
ent
s
ar
e C
onfi
d
ent
i
a
l
i
t
y
agreem
ent
s
, No
n-
d
i
sclo
su
re, leg
a
l lo
catio
n,
restrictin
g
t
h
e so
ft
ware
licen
se.
Nond
isclo
s
ure con
t
ract is eq
u
a
l to th
e
co
nfid
en
tiality
ag
reem
en
t. Th
ere is a d
i
fferen
ce
b
e
tw
een th
e con
f
id
en
tiality ag
ree
m
en
t an
d
Non
d
i
scl
o
sure
ag
reem
en
t. Non
d
i
scl
o
sure agreem
en
t n
eed
th
e certifie
r no
t tru
l
y reveal th
e info
rm
ati
o
n. C
o
nfid
en
tiality
ag
reem
en
t sign
atory resp
on
sib
ility is to
secu
re t
h
e in
fo
rmatio
n
.
If a
dealer h
a
s
signed
o
n
a non
d
i
sclo
sure
ag
reem
en
t, if
it h
a
s security v
i
o
l
ation
,
t
h
ey will n
o
t
con
s
id
er as a d
e
fect fo
r an
y rev
e
latio
n
wh
ereas in
co
nfid
en
tiality
ag
reem
en
t, if
secu
rity
b
r
each
o
c
cu
rs wh
en a d
ealer h
a
s sig
n
e
d
on
con
f
i
d
en
tiality ag
ree
m
en
t,
the deale
r
s
h
ould
res
p
onsi
ble for
res
u
lts of t
h
e sec
u
r
ity v
i
olatio
n
.
Lo
cating
th
e serv
er on n
e
twork
with
so
m
e
speci
fi
ed s
p
ee
d o
f
ba
nd
wi
dt
h
i
s
not
an i
ssue
from
user si
de
but
l
o
cat
i
o
n m
a
t
t
e
rs l
e
gal
l
y
. Every
user s
p
eci
fi
es
wi
t
h
a ser
v
i
ce l
e
vel
agreem
ent
by
t
h
e servi
c
e pr
ovi
der
.
Dy
nam
i
c negot
i
a
t
i
on i
s
occ
u
rre
d
whe
n
t
h
e
user
need
s
to speci
fy the
exact condition for the
re
quests of us
ers.
Weighted
sum
m
odel (W
SM
) deci
des t
h
at
whic
h
p
r
ov
id
er
is o
f
fering
b
e
tter u
s
er.
SLA
p
a
ram
e
ters
are
scalab
ility, p
r
i
v
acy, se
curity, av
ailab
ility.
SLA
p
a
ram
e
ters
m
a
x
i
mize th
e reliab
ility, co
n
f
iden
ce lev
e
l of
clo
u
d
serv
ice
p
r
ov
id
er and
clo
u
d
u
s
er rel
a
tio
n
.
Mu
lti-criteria
Decision
Mak
i
n
g
(MCDM
)
parad
i
g
m
is u
s
ed
to g
e
t t
h
e
best serv
ice prov
id
er for cloud
u
s
e
requ
est. Th
is parad
i
g
m
is u
s
ed
to
d
e
crease th
e serv
ice
co
st
to
u
s
er. Th
is m
o
d
e
l co
n
t
ain
s
m
u
lti-lev
e
l SLAs fo
r
m
u
lt
i
p
l
e
users
and
dy
nam
i
c negot
i
a
t
i
o
n. C
o
nst
r
uct
a t
a
bl
e
fo
r ev
ery
cl
o
ud
ser
v
i
ce p
r
o
v
i
d
e
r
w
h
i
c
h c
ont
ai
n
s
service,
data and infra
struct
ure to co
nfirm
that whic
h p
r
o
v
i
der o
ffe
rs best
service f
o
r us
er’s re
q
u
est. S
e
rvice
pr
o
v
i
d
er
a
n
d In
frast
ruct
ure pr
o
v
i
d
er
a
n
d has
co
n
n
ect
ed in
tern
ally b
u
t th
ey h
a
v
e
th
eir serv
ice lev
e
l
agreem
ents. Se
rvice provi
der
would like t
o
r
un
hi
s ser
v
i
ce
i
n
i
n
f
r
ast
r
uct
u
r
e
t
o
st
ore a
nd
t
o
com
put
e. If
t
h
e
infra
struct
ure
provider se
rvices are
accepte
d by service provide
r
he
nee
d
s to
pay the m
oney to proc
ess hi
s
serv
ice. In
t
h
i
s
m
o
d
e
l th
ere is a
b
r
ok
er between cl
ou
d
u
s
er and serv
i
ce prov
id
er, users
will g
e
t co
st for
requested se
rvice from
all service
provide
r
s, t
h
e
n
c
o
m
p
are th
e co
st
o
f
all serv
ice
p
r
o
v
i
d
e
rs to select less
bet
t
e
r cost
t
o
m
eet
t
h
e cl
oud
user re
q
u
est
u
s
i
ng
wei
g
ht
ed
sum
m
odel
.
Wei
g
ht
ed s
u
m
m
odel
defi
nes
t
h
at
t
h
e
serv
ice
p
r
ov
ider who
m
eet t
h
e Qu
ality o
f
serv
ice elem
e
n
ts su
ch
serv
ice, d
a
ta an
d dead
lin
e for th
e u
s
er.
Dy
nam
i
c
m
odel
gi
ves
o
p
t
i
m
u
m
cost
for
t
h
e
servi
ce
re
quest
ed
by
t
h
e
use
r
.
Ev
ery cu
st
o
m
er is allo
cated
with
serv
ice lev
e
l agreem
en
ts th
en
i
n
v
e
stm
e
n
t
h
a
d
set to
g
i
v
e
th
e t
o
tal
cost
m
eet
t
h
e cl
ou
d
user
i
n
v
e
stm
e
nt
by
get
t
i
ng
fr
om
serv
i
ce pr
o
v
i
d
e
r
.
Whe
n
t
h
e
dat
a
i
s
m
i
grat
i
ng t
h
e
dat
a
fr
om
one cl
ou
d ser
v
i
ce pr
o
v
i
d
er t
o
t
h
e
ot
he
r cl
ou
d p
r
ovi
d
e
r secu
ri
t
y
i
s
an i
ssue. T
o
p
r
o
v
i
d
e t
h
e s
o
l
u
t
i
on
fo
r
th
is issu
e, Clou
d v
a
lid
ation
i
s
u
s
ed
as a secu
re storag
e syste
m
fo
r t
h
e clou
d. Cloud
v
a
lid
atio
n b
a
sed fl
ex
ib
le
d
i
stu
r
b
e
d
app
r
o
ach
fi
nd
s th
e secu
rity v
i
o
l
atio
n
s
su
ch
as in
teg
r
ity, con
f
id
en
tiality. Cryp
t
o
grap
h
i
c im
p
l
e
m
en
ts
in
vo
lv
e in loop
ing
th
e si
g
n
a
t
u
re and
sim
u
lc
ast en
cryp
ti
o
n
. Si
m
u
lcast en
cryp
tion
allo
ws
mig
r
atio
n
to
en
cry
p
t
m
e
ssage t
o
ra
nd
om
num
ber num
b o
f
use
r
s. Si
gnat
u
re
l
o
o
p
i
n
g i
s
a
pr
ocess
of
p
r
od
uci
n
g c
o
n
s
e
c
ut
i
v
e
sig
n
a
t
u
res
from in
itial s
i
g
n
a
t
u
re and
a secret h
ead
si
g
n
a
t
u
re. In
terch
a
ng
e o
f
cl
o
u
d
v
a
lidatio
n
s
secu
re t
h
e d
a
ta
m
i
grat
i
on p
r
oc
ess.
T
h
ese val
i
dat
i
o
n
s
are
ess
e
ntial that allows t
h
e
users
t
o
c
h
eck
t
h
e
cl
ou
d m
i
scon
du
ct
an
d
cl
ou
d p
r
ovi
de
rs t
o
pr
ot
ect
agai
nst
vi
ol
at
i
ons
. C
l
o
ud
se
rvi
ce
pr
ovi
der
s
are faci
ng t
h
e p
r
obl
em
due t
o
u
n
c
ertain
ty o
f
work
l
o
ad
to
reach
th
e serv
ice lev
e
l ag
reem
ent
s
. C
l
o
ud f
e
d
e
rat
i
on
para
di
g
m
pro
v
i
d
es se
r
v
i
c
e
fo
r t
h
e
cl
o
u
d
u
s
ers. C
l
ou
d
us
ers
have
a c
hoi
ce t
o
sel
ect
am
on
g cl
ou
d
pr
o
v
i
d
e
r
s. C
o
m
m
o
n
so
u
r
ces
of
s
e
rvi
c
e
l
e
vel
agre
em
ent
s
vi
ol
at
i
o
n
s
u
n
e
xpect
e
d
i
n
t
e
rr
upt
i
o
ns a
r
e a
f
f
ect
ed by
har
d
w
a
re, s
o
ft
ware a
nd
net
w
o
r
k
fai
l
u
res
.
Clo
u
d
serv
ice d
i
sru
p
tion
s
are still
o
ccu
rring. Reaso
n
for d
i
srup
tion
s
th
ere are
m
o
re un
kn
own
m
a
ssiv
e
-scale
failu
re situ
ation
s
un
d
e
r wh
ich
recov
e
ry wil
l
fail.
Failu
re as a serv
ice
(FaaS) is
propo
sed
to p
e
rm
it th
e
clo
ud
services
. T
h
is
Failure
as a
Service, cl
oud servi
ces c
o
mm
only perform
m
a
ssive scale failure
in
real
d
e
p
l
o
y
m
e
n
t
s.
Failu
re as a serv
ice is d
u
e
to
h
i
gh
un
-stab
ility
in
d
i
strib
u
t
ed
env
i
ro
nmen
t. A
s
th
e cloud
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 1
,
Febru
a
ry
2
015
:
15
8
–
16
5
16
0
com
puting
res
o
urces a
r
e growing in
re
search
and
indu
stry, th
e po
ssib
ili
ty o
f
failures affect th
e app
licatio
n
s
whi
c
h are r
u
n
n
i
n
g o
n
t
h
e cl
ou
d. Fai
l
u
re a
s
a servi
ce pa
radi
gm
i
s
pro
pos
ed t
o
u
s
e by
t
h
e cl
ou
d servi
c
e
p
r
ov
id
ers.
H
a
do
op
p
l
atfo
r
m
p
r
ov
id
es app
licatio
n
v
e
ndo
r
s
and
H
a
doo
p ser
v
ice t
o
test th
e so
f
t
w
a
r
e
tow
a
rd
s
in
crease of
num
b
e
r failu
res.
Failu
re as a Serv
ice is im
p
l
e
m
ented in Ha
doop
use
r’s
se
rvice to eval
uat
e
their
ap
p
lication
s
i
n
th
e
p
r
esen
ce
o
f
failures
[2
8]. On
ce the com
p
u
tin
g
was
do
n
e
lo
cally it is b
e
i
n
g don
e i
n
the
clo
u
d
.
Du
e t
o
m
o
re n
u
m
b
e
r o
f
serv
ice are th
ro
wn
u
n
t
o
the clo
u
d
,
cloud
in
frastru
ct
u
r
e size will in
crease then
th
ere is a p
o
ssib
ility o
f
failu
res to
b
e
o
c
cu
rred
in
ind
i
v
i
du
al
clo
u
d
s. App
licatio
n
s
shou
ld
n
o
t
prep
are for o
n
l
y
th
e infrequ
en
t
failures
o
f
their clou
d infrastru
c
ture
and th
ey sh
ou
ld
await th
e
failu
res as a portio
n
of
appl
i
cat
i
o
ns c
o
m
m
on o
p
erat
i
n
g
pr
oce
d
u
r
e.
At th
e ti
m
e
o
f
o
p
e
ration
,
numero
u
s
app
licatio
n
s
an
d
serv
ices are run
n
in
g
o
n
t
h
e clou
d
will g
e
t
dive
rse
failure
s. T
h
ese
failur
e
s are
ra
ngi
ng
fr
om
har
d
dis
k
e
r
r
o
rs
to
ent
i
re rac
k
s.
T
h
e
s
e failu
res i
n
fl
uenc
e
great
l
y
on t
h
e
appl
i
cat
i
o
n pe
r
f
o
r
m
a
nce and
som
e
t
i
m
e
s it
leads t
o
t
e
m
por
ari
l
y
out
o
f
ser
v
i
ce. Sm
al
l sand
b
ox
testin
g
is no
t en
oug
h
t
o
test th
e failure effects o
n
real applicatio
n
s
wh
ich are ru
nn
ing
on
m
a
ssiv
e
n
u
m
b
e
r
o
f
no
des
.
Fai
l
u
re
as a Servi
ce f
o
r Ha
do
o
p
cl
ust
e
rs i
s
desi
g
n
e
d
and i
m
pl
em
en
t
e
d wi
t
h
G
o
ogl
e’s M
a
p R
e
d
u
c
e
for
t
h
e fram
e
wor
k
of cl
o
ud c
o
m
put
i
n
g. F
a
i
l
u
re
as a servi
ce i
s
used
by
t
h
e cl
ou
d ser
v
i
ce p
r
ovi
der a
nd
use
r
s w
h
o
will ru
n
t
h
eir
serv
ices
o
n
cl
ou
d. Mo
st of the h
u
g
e
en
terprises lik
e Am
az
o
n
, un
in
terru
p
t
ed
failure p
a
rad
i
g
m
th
at con
tin
ually fails d
i
v
e
rse p
a
rts of th
ei
r
in
frastru
ct
u
r
e to
recogn
ize the d
e
fect in
app
licatio
n
s
wh
ich
are
running
on the
cloud. Ma
ny
devel
ope
rs
and
orga
nizations cannot acces
s th
ese
clouds. To test the
fa
ilures
,
sm
a
ll san
d
b
o
x
p
a
rad
i
g
m
is u
s
ed
. Failure as a Serv
ice in
t
o
p
u
b
lic H
a
doop
clou
d
w
ill p
e
rm
i
t
n
o
r
m
a
l
H
a
doo
p
users t
o
u
n
i
n
t
e
rr
upt
e
d
l
y
i
n
jec
t
di
verse c
o
m
b
i
n
at
i
on
of
fai
l
u
res i
n
t
o
t
h
ei
r cl
ou
d a
ppl
i
cat
i
o
ns an
d al
s
o
ev
al
uat
e
s
h
o
w t
h
ey are affected
b
y
fail
u
r
es. Th
is
will h
e
l
p
in id
en
tifyin
g
t
h
e
d
e
fect
s in
d
e
si
g
n
i
n
g
th
eir serv
ices
.
Som
e
tools are effectively injecting
the failu
res int
o
cloud softwa
re syste
m
s such
as HDFS.
Ha
doop contains
failure
recov
e
ry an
d
fau
lt to
leran
c
e. Pro
g
ramm
ab
le failu
re in
j
e
ctio
n
too
l
g
i
ves failu
re no
ti
o
n
s
t
o
testers write
p
r
o
c
ed
ures t
o
eli
m
in
ate m
a
n
y
failu
re co
m
b
in
atio
n
s
. Failu
re as a Serv
ice is used as
qu
ality co
n
t
ro
l t
o
o
l
for
or
ga
ni
zat
i
ons
and al
so
fo
r u
s
ers. E
v
al
uat
e
Had
o
op i
n
di
fferen
t failure situ
atio
n
s
b
y
in
jectin
g
failu
res at th
e
ti
m
e
o
f
job
run
tim
es o
f
Net
w
ork in
ten
s
iv
e, I/O In
te
nsi
v
e
an
d C
P
U i
n
t
e
nsi
v
e
.
I
n
vi
rt
u
a
l
servi
ce
net
w
o
r
k
,
Cloud fe
de
ration provi
d
es that users are
able to exec
ute t
h
e se
rvices
dy
nam
i
cally provide
d
by the
s
e
rvice
pr
o
v
i
d
er
s.
Al
l
o
cat
i
on
of
re
que
st
s fr
om
servi
c
e pr
o
v
i
d
e
r
s s
h
oul
d
not
be
red
u
ced
wi
t
h
o
u
t
r
e
l
e
vant
se
rvi
c
e
l
e
vel
agreem
ent
s
g
u
a
rant
ees a
n
d
w
i
t
hout
a
n
e
ffi
ci
ent
res
o
urce
m
a
nagem
e
nt
[
2
9
]
. Ser
v
i
ce Le
v
e
l
Ag
reem
ent
s
t
ool
s
an
d lang
u
a
g
e
s ex
ist eith
er sev
e
rely limited
o
r
too
d
i
fficu
lt to
u
s
e
b
y
wide set of serv
ice prov
id
ers. Serv
ice
lev
e
l ag
reem
e
n
ts can
b
e
imp
r
ov
ed
in
a less ti
m
e
with
su
itab
l
e to
o
l
s.
Reso
urce m
a
n
a
g
e
m
e
n
t
b
a
sed o
n
th
e
service level a
g
reem
ents is requi
red to a
g
ree a pool
of
vi
rt
ual
i
zed c
o
m
put
ers
inter-connected c
o
m
puters
am
ong
cl
o
u
d
u
s
ers a
n
d cl
o
u
d
pr
o
v
i
d
er
s.
C
l
o
u
d
ag
ency
reac
t
s
t
o
t
h
i
s
be
ne
f
i
t
adde
d
val
u
e
t
o
act
ual
c
o
m
put
i
n
g
services
.
As
pe
r the
available
offers
it produces
SLA
th
at presen
ts
th
e reso
urce n
e
g
o
tiatio
n
resu
lt.
Du
e t
o
in
cap
a
b
ility o
f
clo
ud serv
ice
p
r
ov
id
ers, th
ey are
u
n
a
b
l
e to
satisfy th
e Qu
ality o
f
Serv
ice
levels are
speci
fied i
n
se
rvice
level
ag
reem
ent
s
whi
c
h l
e
a
d
t
o
cl
o
u
d
fede
rat
i
on
vi
si
o
n
.
Thi
s
cl
o
u
d
fe
derat
i
on i
s
easi
l
y
act
i
n
re
spo
n
se
t
o
c
h
a
n
ges i
n
w
o
rkl
o
a
d
,
res
o
u
r
ce a
n
d
net
w
or
k c
o
n
d
i
t
i
ons
are a
g
g
r
essi
vel
y
c
o
o
r
d
i
nat
e
s
the num
erous
clouds in fe
de
ration. It is im
practicable
to cloud
provi
der to m
a
in
tain datace
nters
in each
co
un
try, clou
d fed
e
ration
prov
id
es th
e add
e
d
b
e
n
e
fit to
satisfy th
e requ
ire
m
en
ts of
g
e
o
l
o
g
i
cally d
i
stri
bu
ted
users
t
h
a
n
si
ng
l
e
cl
ou
d se
rvi
c
e pr
o
v
i
d
e
r
s
[2
7]
. Si
n
g
l
e
cl
o
u
d
pr
ovi
si
on
m
odel
c
o
nt
ai
ns
C
l
ou
d u
s
ers
ar
e w
h
i
c
h
cont
ai
n
s
fl
exi
b
l
e
cl
oud fe
de
r
a
t
i
on m
odel
.
C
l
ou
d fede
rat
i
on c
ont
ai
n
s
n
u
m
erous cl
o
u
d
servi
ce p
r
ovi
d
e
rs are
abl
e
t
o
i
n
t
e
ract
am
ong t
h
em
sel
v
es co
nsi
s
t
e
n
t
l
y
. C
l
oud co
m
put
i
ng devel
ops t
h
at
ne
xt
gene
rat
i
o
n cl
o
ud
hav
e
ab
ility
to
fo
rm
a fed
e
ration
wh
ere it will
l
e
v
e
rag
e
s co
m
p
u
t
atio
n
a
l reso
urces wh
ich
redu
ce th
e v
i
o
l
atio
n
of
ri
sks o
f
cl
o
ud
user
’s ser
v
i
ce l
e
vel
agreem
ent
s
by
m
ovi
n
g
t
h
e jo
bs bet
w
een t
h
e pr
o
v
i
d
ers fede
rat
i
o
n.
C
l
ou
d
fede
ration to s
a
tisfy the se
rvi
ce level a
g
ree
m
ent acceptan
ce, a m
i
ddleware layer is
ne
eded to coordi
nate the
clo
u
d
p
r
ov
id
er’s activ
ities and
cloud
u
s
ers activ
ities. Mi
d
d
l
eware layer sho
u
l
d
ad
ap
t t
o
ch
ang
e
t
h
e co
nditio
ns
in
d
i
stribu
ted
en
v
i
ron
m
en
t i
n
resp
on
se t
o
th
e ev
en
ts th
at may trig
g
e
r the serv
ice lev
e
l
ag
reem
en
t v
i
olatio
n
.
M
i
ddl
ewa
r
e ne
eds ag
g
r
essi
ve
scal
i
ng t
o
assi
s
t
heavy
w
o
r
k
l
o
ad.
Arc
h
i
t
ect
ur
e of C
l
o
ud
fed
e
rat
i
on i
s
p
r
op
os
e
d
whi
c
h i
n
cl
u
d
es
a new m
i
ddl
eware l
a
y
e
r i
s
d
e
si
gne
d ba
sed
up
o
n
t
h
e dy
na
m
i
c dat
a
dri
v
e
n
ap
pl
i
cat
i
on s
y
st
em
s
m
odel
pri
n
ci
pl
es. M
i
d
d
l
e
t
i
e
r c
hos
e t
h
e
cl
ou
d
ser
v
i
ce
pr
ovi
ders
t
o
m
oni
t
o
r
u
s
er
t
a
sk
s t
o
a
ssu
re t
h
ey
are
p
r
ov
id
ing
within
th
e bo
und
s of
serv
ice
lev
e
l ag
reem
en
ts
con
d
ition
s
.
2.
LIFE CYCLE
OF
SERVICE LEVEL AGREEME
N
T (SLA
)
SLA l
i
f
e
cy
cl
e m
a
nagem
e
nt
i
s
di
vi
de
d i
n
t
o
fi
ve st
a
g
es s
u
c
h
as
Ser
v
i
ce
D
e
vel
o
pm
ent
,
N
e
got
i
a
t
i
o
n
and M
a
r
k
et
i
n
g,
Devel
o
pm
ent
,
Im
pl
em
ent
a
t
i
on an
d eval
ua
t
i
o
n
.
SL
A m
a
nagem
e
nt
offer
s
t
w
o t
y
pes
of se
rvi
c
e
s
whi
c
h are
p
r
e-
ru
n-t
i
m
e and r
unt
i
m
e [5]
.
Pr
e-r
u
n
tim
e
refe
rs that,
be
fo
re
the service runtim
e is started, SL
A
R
e
gi
st
rat
i
o
n
,
S
e
rvi
ce I
n
qui
ry
and C
o
nt
ract
a
nd
ne
got
i
a
t
i
o
n
has t
o
be
do
ne
. Ser
v
i
ce Pr
o
v
i
d
ers s
h
oul
d re
gi
st
er
th
e SLA to
pro
v
i
d
e
th
e serv
i
ces to
th
e SLA m
a
n
a
g
e
m
e
n
t
syste
m
. Serv
i
ce clien
t
s will search th
e services in
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Service Level
Agree
m
ents
in
Cloud Computi
n
g and Bi
g
Dat
a
(
K
Ra
d
ha)
16
1
syste
m
service library base
d
on
QoS ne
eds
.
Service providers and Se
rvic
e
Clients get personalized
with each
ot
he
r t
o
ne
got
i
a
t
e
t
h
e SL
A c
o
nt
ract
t
o
assu
re
t
h
at
cl
i
e
nt
can
pay
as
per
nee
d
s s
u
c
h
as
SL
A m
e
t
r
i
c
s and
penal
t
y
ru
les. Clien
t
sh
ou
ld ob
ey the ru
les g
i
v
e
n
b
y
th
e serv
i
ce
Provi
d
ers
.
Run tim
e
stage
is also
called
as
serv
ice
o
p
e
ration
stag
e. Ru
n
tim
e s
t
ag
e task
is to
mo
n
itor and
co
ntro
l th
e SLA metrics an
d
mak
i
ng
th
e Vi
o
l
atio
n
of
ru
les. Mak
e
sure th
at SLA m
e
trics m
e
e
t
s th
e requ
irem
en
ts and
defin
e
th
e pu
n
i
sh
m
e
n
t
d
ecisio
n
s
wh
en th
e ru
l
e
is v
i
o
l
ated
. SLA o
f
fers th
e Serv
ice Prov
id
ers to
d
i
ffe
ren
tiate th
e
m
fro
m
th
e to
d
a
y’s co
m
p
etitiv
e en
v
i
ron
m
en
t.
SLA a
n
d QoS
are propose
d
Web Service
Offeri
ngs
la
nguage
(WS
O
L). Term
s and C
o
nditions am
ong t
h
e
o
r
g
a
n
i
zatio
n
s
an
d clou
d
prov
id
ers set th
e
respo
n
s
i
b
ility
a
t
en
terprise level. Clo
u
d
Co
m
p
u
ting
con
t
ract
s m
a
y
set SLA, License agreem
ent to activat
e
on non-perform
a
nce contract. SL
A is
good pract
ice to ass
u
re t
h
e legal
areas t
o
see
k
t
h
e ap
pr
o
p
ri
a
t
e l
e
gal
gui
da
nce. S
o
m
e
of
t
h
e areas a
r
e
–o
bl
i
g
at
i
o
n
wi
t
h
dat
a
pr
ot
ect
i
on
leg
i
slatio
n
,
free o
f
in
form
at
i
o
n
leg
i
slation
o
b
lig
ation
s
, con
f
i
d
en
tial in
form
at
io
n
,
m
o
n
ito
ri
n
g
th
e
u
s
ers an
d
security of dat
a
, etc. Service
Leve
l Ag
reemen
t Issu
es are Data p
r
o
t
ectio
n, Data Security, Lo
catio
n
of d
a
ta,
Licen
sing
, Con
f
i
d
en
tiality, Law and
ju
risd
i
c
tio
n
,
Reten
tio
n
of d
a
ta,
Direct d
a
m
a
g
e
s a
n
d
Ind
i
rect d
a
mag
e
s,
Term
in
atio
n
[6], [7
].
W
e
b Serv
ice Lev
e
l
Agree
m
en
t is
propo
sed
to m
o
n
itor th
e SLA and
to
m
a
n
a
g
e
th
e
SLAs
in a
distributed E
n
vironm
ent. SL
A
web services ar
e
de
scribed a
s
Web Service Le
vel
Agreem
ent and
We
b
Service Agree
m
ent.
3.
FRAMEWORK OF WEB SERVICE
LE
VEL AGREE
M
ENT
(WSLA)
Parties, SLA
param
e
ters, Serv
ice Lev
e
l Ob
jectiv
es (SLOs) are th
ree en
tities o
f
W
e
b
Serv
ice Lev
e
l
Agreem
ent [8]. The
r
e a
r
e three type
s o
f
Parties
n
a
m
e
ly
Serv
ice
Prov
ider, Se
rvice Consum
er and T
h
ird
Part
i
e
s. T
h
i
r
d
part
i
e
s m
a
y
vary
t
o
t
a
ke
de
ci
si
ons
o
n
vi
o
l
at
i
ons ei
t
h
er
by
t
h
e Se
r
v
i
ce Pr
ovi
der
o
r
S
e
rvi
c
e
Cons
um
er.
In Web
Se
rvice Level Agreements
(WSLA),
S
L
A
p
a
r
a
me
t
e
r
s
ar
e
u
s
ed
to
me
a
s
u
r
e
th
e s
e
rv
ic
e
param
e
ters. There are t
w
o
kinds of Metrics such as
C
o
m
posi
t
e
m
e
t
r
ics and R
e
so
urce m
e
trics.
Fro
m
th
e
service
provider’s
res
o
urces
resource Metri
c
s are
accesse
d.
Co
m
p
o
s
ite
Metrics are com
b
in
atio
n
of
variou
s
resource
m
e
tri
c
s an
d
Serv
ice Lev
e
l Obj
ectives are set of
expressi
ons as
if-th
en
structure.
W
e
b services are
pr
o
v
i
d
e
d
t
o
t
h
e
co
nsum
ers wi
t
h
di
ffe
re
nt
ser
v
i
ce l
e
vel
s
by
s
u
i
n
g t
h
e
aut
o
m
a
t
e
d m
a
nagem
e
nt
an
d
ser
v
i
ce l
e
vel
agreem
ents. Web se
rvice level agreem
ent language s
p
eci
fi
es th
at to
supp
ly th
e resources b
a
sed
on
the SLA.
Work
lo
ad
m
a
n
a
g
e
m
e
n
t
g
i
v
e
s th
e priority to
th
e
requ
ests asso
ciated
wi
th
serv
ice lev
e
l ag
reem
en
ts an
d
t
o
m
o
n
ito
r th
e ag
reem
en
t with th
e
serv
ice lev
e
l ag
r
eem
ent.
Pri
n
ciples of
The
WSL
A
Fram
ework
a
r
e
SL
A
Param
e
ters, Bu
sin
e
ss Metrics, Resou
r
ce
Metrics, an
d
Co
m
p
o
s
ite Metrics. Resou
r
ce m
e
trics are d
i
rectly
accessed
from the controlled resources whi
c
h are t
h
ere i
n
the se
rvice
provider’s
tier suc
h
as routers and
serve
r
s. C
o
m
posi
t
e
m
e
t
r
i
c
s are ge
nerat
e
d b
y
gat
h
eri
n
g va
ri
o
u
s res
o
u
r
ces
. C
o
m
posi
t
e
m
e
t
r
i
c
s are defi
n
e
d by
serv
ice lev
e
l ag
reem
en
ts. Busin
e
ss m
e
trics form
th
e c
ons
um
er ri
sk m
a
nagem
e
nt
pol
i
c
y
whi
c
h i
s
av
ai
l
a
bl
e
within the cus
t
om
er service dom
ain.
Service provide
r
perform
s
a
ma
ppi
ng to ass
u
re that service level
ag
reem
en
t and
to
fu
lfill bu
siness go
als.
W
e
b serv
ice lev
e
l ag
reem
en
t lan
g
u
a
g
e
is
XML sch
e
m
a
.
4.
PRESENT
A
NATO
M
Y
O
F
CLO
U
D SE
RVI
CE
AG
R
EEMENT
Cloud Consum
ers com
p
are the agreem
en
ts between
th
e
d
i
stin
ct p
u
b
lic
cl
o
ud pr
o
v
i
d
e
r
s. A
co
ns
um
er
h
a
s t
o
b
e
carefu
l wh
ile selectin
g th
e lang
u
a
ge fo
r th
e
a
g
ree
m
ents. Som
e
times there a
r
e
voca
b
ulary e
r
rors
tha
t
alter the m
eaning
of a cla
u
s
e
. Before
si
gning the contra
ct catch the er
ro
rs a
nd c
o
rre
ct the err
o
r
s
.
Clou
d
agreem
ent is divided i
n
to t
h
ree types suc
h
a
s
End-Us
e
r
agree
m
ent, Acce
pting the
Policies and Service
Level
Agreem
ent [9]. In End-Use
r
Agreem
ent
,
B
u
si
ness
ser
v
i
ce m
a
nagem
e
nt
co
nsi
s
t
s
t
h
e p
o
l
i
c
i
e
s of
cl
ou
d
providers. Custom
er Agreem
ent ful
f
ills the requirem
en
t
of “Te
r
m
s
of Service”.
Public cloud custom
er
agreem
ents consists
of t
h
e c
r
itical sections
as follo
ws
offere
d se
rvices
, breaki
n
g the
service tem
porarily,
p
a
yin
g
th
e fee, term
in
atin
g
t
h
e term
s and
co
nd
itio
ns
,
d
i
sclai
m
er, li
m
i
te
d
liab
ility an
d security. Offered
serv
ices
d
e
scri
b
e
s th
at,
h
o
w cu
sto
m
er u
tilize
s
th
e pu
b
lic clou
d
o
f
feri
n
g
s in ter
m
s o
f
sup
p
l
yin
g
th
e serv
ice an
d
Serv
ice D
e
scr
i
p
tio
n d
e
tails.
Pay
m
en
t o
f
f
e
e r
e
f
e
r
s
th
at, it is th
e m
e
th
od
of
sp
en
d
i
n
g
th
e f
ee
f
o
r
th
e cloud
services s
u
c
h
as Service Charge
Sche
dule, buyi
ng t
h
e
servi
ce a
nd
d
e
fray
m
ent
Term
s and C
o
n
d
i
t
i
ons
.
Tem
porary s
u
s
p
ension of a
se
rvice is
a
proce
ss where
the se
rvice
provide
r
sus
p
ends t
h
e
usage
of t
h
e cl
oud by
specific cons
umer for a time
base
d o
n
t
h
e dra
w
back
of
anom
al
ous
us
age of
distributed environm
ent and
saf
e
ty pr
ob
lems.
Term
in
atin
g
the Term
s an
d
Co
nd
itio
ns d
e
scrib
e
s th
at
Service p
r
o
v
i
d
e
rs t
e
rm
in
ate serv
ices u
s
ed
b
y
th
e clien
t
an
d
Clo
s
es th
e Acco
un
t
o
f
clien
t
s. Disclaim
er
is d
e
scri
p
tio
n
o
f
serv
ices wh
ich are
no
t in
cluded
i
n
th
e agreem
en
t in
term
s o
f
Warran
ties and
Disclai
m
er. Li
mited
liab
ilit
y sp
ecifies a li
mit th
at a cu
st
omer can
clai
m
o
n
th
e li
mited
paym
ent
. Security in
clu
d
e
s th
at Respo
n
s
i
b
ility, p
r
o
t
ectin
g
th
e Dat
a
, Priv
acy p
l
an
and
Cu
sto
m
er
respo
n
s
i
b
ility
. Acc
e
pt
abl
e
Use P
o
l
i
c
i
e
s refers t
h
a
t
Pu
bl
i
c
cl
ou
d
pr
o
v
i
d
er
s an
d
C
l
ou
d C
l
i
e
nt
s
sho
u
l
d
ag
ree th
e term
s and
C
o
nd
ition
s
.
In
Serv
ice
Lev
e
l
Agreem
en
ts of Cloud
p
h
a
se, Serv
ice
Lev
e
l
Agreemen
ts of
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 1
,
Febru
a
ry
2
015
:
15
8
–
16
5
16
2
Clo
u
d
are accep
ted
fo
r
b
o
t
h Clien
t
an
d
p
r
o
v
i
d
e
r th
at d
e
fin
e
s th
e set o
f
o
b
j
ectiv
es
co
n
t
ains av
ailab
ility,
perform
a
nce, Security and privacy
policies.SLA C
h
allenges are SLA
Architecture, SL
A Based Sc
he
duling
Pol
i
ces, an
d R
e
so
urce Al
l
o
c
a
t
i
on o
f
SLA
.
Thr
o
u
g
h
o
u
t
t
h
e w
o
rl
d
w
i
d
e
,
C
l
oud com
p
u
t
i
ng of
fe
rs pa
ym
ent
-
o
r
ien
t
ed
an
d
o
r
g
a
n
i
zatio
n
a
l
–
Qu
ality co
mp
u
ting
serv
ic
es to
end
-
u
s
ers. Serv
ice Pro
v
i
d
e
rs
d
e
p
l
oy d
a
ta
p
r
o
cessi
n
g
centers in
v
a
riou
s
lo
catio
n
s
fo
r pro
v
i
d
i
ng
b
ackup
for en
su
ring
t
h
e
reliab
ility.
5.
CHALLENGES OF
SERVICE
LEVEL AGREEME
N
TS
(SL
A
)
Reso
urce m
a
n
a
g
e
m
e
n
t
issu
es are in
vo
lv
ed
in
d
e
liv
ering
the serv
ices for a
m
i
l
lio
n
s
o
f
u
s
er’s serv
ices
v
i
a d
a
ta cen
ter. Ch
allen
g
e
s
of serv
i
ce level agreem
ents are as follows da
t
a
pr
ocessi
n
g
o
f
ri
sk m
a
nage
m
e
nt
,
con
s
um
er-d
ri
v
e
n ser
v
i
ce m
a
n
a
gem
e
nt
, and i
nde
pe
nde
nt
res
o
u
r
ce m
a
nage
m
e
nt
,
m
easuring t
h
e service,
syste
m
d
e
sign
and
reit
eratio
n v
a
l
u
atio
n resou
r
ce allo
catio
n in
SLA with
v
i
rt
u
a
lizatio
n
.
5.1 Consumer
-Dri
ven Servi
ce
Manage
me
nt
Three
user-ce
n
tric objectives
are use
d
to satisfy
the custom
er requirem
ent
wh
ich
in
clud
es g
e
ttin
g
th
e
feed
bac
k
fr
om
t
h
e cust
om
ers, pr
o
v
i
d
i
n
g t
h
e
rel
i
a
bl
e comm
unication between the
c
u
st
om
ers, increas
ing the
access efficiency to
unde
rstand the
sp
e
c
ific necessities of
the c
u
stom
er
and believe the
custom
er.
Whe
n
t
h
e
service is developing, if cus
t
om
er expectations are
consi
d
ere
d
, the
n
those expectations of custom
er are
im
ported into
service
provider.
If t
h
ese e
xpectations ar
e i
m
ple
m
ented by service
prov
i
d
er shou
ld
satisfy th
e
client’s re
quire
ments. Reasonable expectations of c
u
st
om
er are accepte
d a
nd a
d
opte
d
by
the provide
r
then it
enters i
n
to c
u
stom
er contra
ct. Custom
er-dri
ven co
ntrac
t
s characte
r
istics are
as
fo
llo
ws, singu
lar, on
-
au
tho
r
itativ
e,
b
oun
d
e
d
stab
i
lity
an
d
immu
tab
ility co
m
p
lete an
d
clo
s
ed
. Cu
st
o
m
er-driv
e
n
ag
reem
e
n
t is
co
m
p
leted
an
d clo
s
ed
t
o
th
e
co
m
p
lete co
lle
ctio
n
of fun
c
tio
n
a
lity d
e
m
a
n
d
e
d
b
y
ex
isting
cu
st
o
m
ers. Prov
i
d
er
ag
reem
en
ts are sin
g
l
e in
th
eir ex
pressi
o
n
o
f
b
u
s
i
n
e
ss fun
c
tio
n
a
lity av
ailab
l
e to
th
e syste
m
, wh
ereas n
o
n
-
au
tho
r
itativ
e are d
e
ri
v
e
d
fro
m
co
m
b
in
atio
n
of ex
isting
co
ns
um
er expectations
. Cust
om
er dri
v
en a
g
reem
ent is
u
n
c
h
a
n
g
e
ab
le
with
resp
ect to
th
e
p
a
rticu
l
ar set o
f
c
u
st
om
er agreem
e
n
ts. Vali
dity
of c
u
stom
er drive
n
agreem
ent according to a
speci
fied set of
custom
er agre
e
m
ents is
effe
ctively bounde
d to t
h
e forwa
r
d a
nd
b
ackward
co
mp
atib
le agreemen
t in
sp
ace an
d ti
m
e
. The co
m
p
atib
ilit
y o
f
co
nsu
m
er agreem
en
t re
m
a
in
s
unc
ha
nge
d f
o
r speci
fi
ed
cu
stomer expectations.
5.
2 Da
ta
Pr
oc
essi
ng of
Ri
s
k
Ma
na
geme
nt
The R
i
sk M
a
nagem
e
nt
pr
o
cess co
nt
ai
ns
t
h
e f
o
l
l
o
wi
ng
way
s
suc
h
as
Ide
n
t
i
f
y
i
n
g
t
h
e R
i
sk
a
n
d
assesses th
e
risk
, id
en
tify th
e t
ech
n
i
q
u
e
s to
man
a
g
e
t
h
e risk
s an
d
rev
i
ew the risk
m
a
n
a
g
e
men
t
p
l
an
.
Quality o
f
servi
ce c
o
n
d
i
t
i
ons
of
g
r
i
d
se
rvi
ce c
u
st
om
er
s nee
d
t
h
e
fo
r
m
of ser
v
i
ce l
e
vel
ag
reem
ent
s
am
ong
user
s a
n
d
service provi
d
ers. Haza
rds for a Su
cces
sful service level agreem
ent
supp
lying in grid com
puting, firstly
f
a
ilu
r
e
s
of
co
mp
u
t
ation
a
l nodes. Gr
id
serv
ice p
r
ov
id
er
s
r
e
qu
ir
e the r
i
sk
analysis to
ev
aluate ex
p
ected
losses in
resource m
a
na
gem
e
nt. Deciding that whether to accept or
reject service level agreem
e
n
t request is an issue
for
se
rvice provide
r
s, since re
source
a
r
e
disrupte
d
a
n
d una
v
ailable hazard
[31].
5.3
Independe
n
t Res
o
urce Manageme
nt
Data p
r
o
cessi
ng
cen
ter sh
ou
l
d
m
a
in
tain
th
e reservatio
n
p
r
o
cess withou
t in
terru
p
tion
b
y
m
a
n
a
g
i
n
g
th
e presen
t serv
ice requ
isitio
n
an
d im
p
r
o
v
e
th
e fu
ture se
rv
ice requ
isitio
n
an
d altering
th
e price for the n
e
wly
receive
d reque
s
ts
[10]. SL
A Risk
Mana
ge
ment Challenges are
I
n
f
o
r
m
a
t
i
on
Secu
ri
t
y
a
n
d
Ri
sk M
a
na
geme
nt
,
SLA an
d E
x
ce
pt
i
on m
a
nage
m
e
nt
[11]
, [
12]
.
Developing SLA in Dat
a
Center
Service Level agreem
ents helps
t
h
at
t
h
e I
D
C
P
r
om
i
s
es t
h
at
w
h
at
i
s
P
o
ssi
bl
e
t
o
del
i
v
er,
del
i
vers
w
h
at
I
D
C
i
s
pr
om
i
s
ed [1
5]
. B
a
si
c
pr
obl
em
with
clou
d
app
licatio
n
s
is to co
m
p
are th
e task
s
n
a
m
e
l
y
a
s
fo
llows
d
a
ta tran
sm
issio
n
or co
m
p
u
t
atio
n
in
to
a
po
ol
of
reso
u
r
c
e
s t
h
at
sho
u
l
d
m
eet
t
h
e cl
oud
appl
i
cat
i
ons c
o
st
, pe
rf
orm
a
nce, securi
t
y
pa
ram
e
t
e
rs. In resou
r
ce
m
a
nagem
e
nt
paradi
gm
, i
n
t
e
ract
i
ons o
f
re
so
u
r
ces are m
a
ppe
d t
o
a
po
ol
o
f
pl
at
fo
rm
i
ndep
e
nde
nt
ser
v
i
ce
l
e
vel
agreem
ents [30]. Service level agreem
ents
do
not guara
n
t
ee the respons
e ti
me and also it is difficult
because
of
un
pre
d
i
c
t
a
b
l
e t
r
affi
c pat
t
e
r
n
s.
Ada
p
t
i
v
e r
e
so
urce m
a
nag
e
m
e
nt
i
n
crease
s
t
h
e usa
g
e o
f
web a
ppl
i
cat
i
o
n by
max
i
m
i
zin
g
th
e resource u
tilizatio
n
.
Arc
h
i
t
ect
ure
of
reso
u
r
ce
m
a
nagem
e
nt
wi
t
h
t
h
e c
o
o
p
e
rat
i
on
of
com
put
i
n
g
syste
m
s v
i
a n
u
m
ero
u
s
v
i
rtu
a
l
mach
in
es in
creases th
e p
e
rfo
rman
ce o
f
co
m
p
u
t
ation
a
l syste
m
s will
i
m
p
r
ov
e th
e
u
tilizatio
n
of
resou
r
ces d
e
sign
ed fo
r
o
n
-d
eman
d
u
tiliza
tio
n
o
f
resou
r
ce.
Arch
itectu
r
e of reso
urce m
a
n
a
g
e
m
e
n
t
has be
nefi
t
s
o
f
som
e
co
m
ponent
s i
n
vi
rt
ua
l
i
zed pl
at
fo
rm
, cl
oud c
o
m
put
i
ng pl
at
f
o
rm
and
gri
d
com
put
i
n
g
pl
at
fo
rm
reduc
es t
h
e o
v
er
hea
d
o
f
com
put
at
i
onal
sy
st
em
s. Arc
h
i
t
ect
ure
of
reso
u
r
ce m
a
nagem
e
nt
wi
t
h
t
h
e
co
op
eratio
n of co
m
p
u
ting
sy
ste
m
s h
a
s
h
i
ghest CPU u
tilizatio
n
and
b
e
st
p
e
rform
a
n
ce.
Virtu
a
l m
ach
ine b
a
sed
reso
u
r
ce p
r
ovi
si
oni
ng i
s
a
d
o
p
t
e
d i
n
di
st
ri
b
u
t
ed en
vi
r
onm
ent
s
.
Al
l
o
cat
i
o
n
s
of
vi
rt
ual
m
a
chi
n
es
by
usi
n
g st
at
i
c
sche
duling m
echanism
,
res
o
urces are
not c
o
m
p
letely u
tilized. Virtual
m
a
chine uses
opt
i
m
a
l cloud res
o
urce
pr
o
v
i
s
i
oni
ng i
n
dy
n
a
m
i
c resou
r
ce al
l
o
cat
i
o
n. Se
rvi
ce l
e
v
e
l
agreem
ent
s
sho
u
l
d
be
use
d
t
o
gi
ve t
r
adi
ng
of
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Service Level
Agree
m
ents
in
Cloud Computi
n
g and Bi
g
Dat
a
(
K
Ra
d
ha)
16
3
reso
u
r
ces base
d u
p
on t
h
e ec
o
nom
y
m
odel
s
. Ser
v
i
ce l
e
vel
ag
reem
ents are able to find
the
i
r needs and identify
th
e prov
id
er abilit
ies. Cu
sto
m
er driv
en
ag
reemen
ts h
a
s two
ad
v
a
n
t
ag
es
su
ch
as
d
e
liv
ery an
d
sp
ecificati
o
n
of
serv
ice
fu
n
c
ti
o
n
a
lity aro
und th
e
b
u
sin
e
ss
v
a
lu
e
driv
ers
an
d m
i
n
i
mal set o
f
req
u
i
remen
t
s are d
e
v
e
l
o
p th
e
service
whe
n
c
u
stom
er expect
a se
rvice to
devel
o
p,
o
p
era
t
e and
depl
oy
i
n
a
m
a
nageabl
e
m
a
nner. C
u
st
om
er
d
r
i
v
en
agreemen
t is su
itab
l
e to
sing
le organ
i
zatio
n.
The
s
e agre
em
ents
efficiently
con
t
ro
l th
e in
terrup
ted
changes
to t
h
e
agreem
ents. T
h
is a
g
reem
ent
o
p
tim
izes th
e bo
und
ing
b
e
tween
th
e serv
ices.
5.4 Resource Allocation
in
SLA with Virtualiz
ation
Virtu
a
lizatio
n p
r
ov
id
es
co
m
p
u
tin
g
res
o
urces
. He
nce a is
ola
t
ed physi
cal machine is ca
pa
ble to
work
like m
a
ny logical Virtual
Machines. Virtual m
achines
are ab
le to
p
r
ov
id
es m
u
lti
p
l
e op
eratin
g
syste
m
envi
ronm
ents and capa
b
le to
configure
Virt
ual Machi
n
es
t
o
use di
st
i
n
ct
s
e
gre
g
at
i
o
n of r
e
so
urces o
n
t
h
e
sam
e
phy
si
cal
m
achi
n
e.
Vi
rt
ual
m
achi
n
es
assi
g
n
e
d
vari
ous
res
o
u
r
ce m
a
nage
m
e
nt
pol
i
c
i
e
s
are
pr
ovi
di
n
g
fo
r t
h
e
v
a
ri
o
u
s
u
s
er
req
u
i
rem
e
n
t
s to
su
ppo
rt th
e im
p
l
e
m
en
t resou
r
ce allo
cation in
SLA.
Supp
lyin
g
th
e
reso
urces
efficiently is a
challengi
ng
problem
in distri
bute
d
e
nvi
ronment because
of
its C
h
a
ngi
ng
nature a
n
d necessity
o
f
su
ppo
rting
th
e
di
v
e
rse
ap
plicatio
n
s
with
d
i
stin
ct p
e
rforman
ce require
m
ents. Resour
ce allo
catio
n
prob
lem
i
n
dat
a
cent
e
r
pr
ovi
des
vari
ous
ki
nds
of
appl
i
cat
i
o
n w
o
r
k
l
o
a
d
s s
p
eci
fi
cal
l
y
reci
pro
cat
ed an
d ent
e
rp
ri
se
ap
p
lication
s
. Ad
m
i
ssio
n
co
n
t
ro
l an
d
sch
e
d
u
l
in
g
m
ech
an
ism is p
r
o
p
o
s
ed
for u
tilizin
g
the resou
r
ces and
end-
u
s
er
SLA n
e
eds [25
]
.
An
op
timal j
o
in
t m
u
lti
p
l
e reso
urce al
lo
catio
n
m
e
th
od
is u
s
ed
i
n
Al
lo
catio
n
o
f
reso
urce
m
o
d
e
l o
f
d
i
strib
u
t
ed env
i
ronmen
t. Th
e
reso
urces
wh
ich
Allo
cated
are
co
mmitted
to
ev
ery serv
ice
req
u
e
st.
Th
ese m
e
th
o
d
s are d
ecrease th
e p
r
ob
ab
ilit
y o
f
requ
is
itio
n
lo
ss and
red
u
ce th
e to
tal
resou
r
ce. Reso
urce
allo
catio
n
ap
pro
a
ch
is pro
posed
for th
e mu
lti-d
i
m
e
n
s
io
nal resou
r
ce all
o
catio
n
prob
lem to
ex
ecu
te
u
s
er’s
ap
p
lication
s
.
5.
5
Mea
s
uri
n
g
the
Ser
v
i
ce
Vari
ous
Ser
v
i
c
e pr
o
v
i
d
e
r
s are
pr
o
v
i
d
i
n
g
di
st
i
n
ct
com
puting services
. Se
rvice
m
easurem
e
n
t acquire
s
m
o
st suitable services t
o
sat
i
sfy the cons
umer needs.
E
v
aluating t
h
e se
rvice
perform
a
nce is re
quire
d
for
ori
g
i
n
al
cl
o
ud
fo
ot
p
r
i
n
t
s
f
r
o
m
vari
o
u
s
p
ubl
i
c
doc
um
ent
s
t
o
m
odel
t
h
e appl
i
cat
i
on an
d ser
v
i
ce nee
d
s. P
r
e
s
ent
l
y
th
ere are no
measu
r
em
en
ts are av
ailab
l
e to
d
e
termin
e u
tilizat
io
n
–
b
a
sed
m
o
n
ito
ring
o
f
resou
r
ces fo
r th
e
d
i
stribu
ted
environ
m
en
t. Th
ere are v
a
riou
s
data-in
t
en
si
v
e
ap
p
lication
s
and
wo
rkflow app
licatio
n
s
,
reliab
ility
and
security. There is
neces
sity to
predict
the collection of
se
rvice m
easurem
ents for exact assessment of
co
n
t
ro
lling
th
e resou
r
ce
p
r
incip
l
es. Ben
c
hmark
p
r
ov
id
es
th
e fu
ture pred
ictio
n
of con
s
u
m
er requ
ire
m
en
ts.
Service
m
easurem
ent has inform
ation on t
h
e confi
g
ura
tio
n
o
f
the curren
t syste
m
an
d
ru
n
tim
e in
fo
rm
at
io
n
metrics as p
a
rt of th
e serv
ice
lev
e
l agreem
en
t. It m
e
ters
th
e
p
a
ram
e
ters of
serv
ice lev
e
l ag
reem
en
t b
y
d
i
rectly
accessing
form
managed resources [32].
Measurea
bl
e
qualities of se
rvice le
vel agreem
ents are
quality,
av
ailab
ility, co
st, cap
acity an
d
laten
c
y wh
ereas un
-m
easureab
le
q
u
a
lities are security, in
terop
e
rab
ility an
d
m
o
d
i
fiab
ility.
5.
6
S
y
s
t
em De
si
gn and
Rei
t
e
r
ati
o
n V
a
l
u
a
t
i
o
n
R
e
sou
r
ce m
a
nagem
e
nt
pl
ans
are e
v
al
uat
e
d
t
h
r
o
u
g
h
di
ffe
re
nt
ki
nd
s
of s
o
urces
an
d c
o
ns
um
ers wi
t
h
v
a
rian
t serv
ice prerequ
i
site to
p
r
ov
e t
h
e e
fficiency. It is
tedious
to
pe
rf
orm
per
f
o
r
m
a
nce as
sessm
ent
o
f
m
o
n
ito
rin
g
th
e reso
urce
p
l
ans in
rep
e
tition and
ad
m
i
n
i
strab
l
e fash
ion sin
ce
res
ources are
tran
sferred and
serv
ice
req
u
i
si
tio
n
s
will come fro
m
v
a
ri
o
u
s
con
s
u
m
ers at
an
y stag
e. Mo
n
itoring
t
h
e resou
r
ce st
rateg
i
es
p
e
rf
or
m
a
n
ce can
b
e
ev
alu
a
ted
b
y
D
i
scr
e
p
a
n
c
y-
ev
en
t simu
latio
n
.
CloudSi
m
esti
mate
t
h
e r
e
so
ur
ce mo
n
itor
i
ng
pl
ans
per
f
o
r
m
a
nce. C
l
o
u
d
S
i
m
i
s
a t
ool
ki
t
i
s
used t
o
m
odel
and si
m
u
l
a
t
i
on of
di
st
ri
but
e
d
en
vi
r
onm
ent
reso
u
r
ces a
n
d
sche
dul
i
n
g t
h
e
ap
pl
i
cat
i
ons.
C
l
ouSi
m
const
r
uct
t
h
e si
m
u
lat
i
on
fram
e
wo
rks
t
o
cal
c
u
l
a
t
e
t
h
e
p
e
rform
a
n
ce of m
a
n
a
g
e
m
e
n
t
of resou
r
ce po
licies [2
4
]
. U
tility
Co
m
p
u
tin
g
p
r
ov
id
es su
bscri
p
tio
n–orien
t
ed
servi
ces
. SLA
com
pone
nt
s ar
e Scope
, Li
m
i
tat
i
ons,
Val
i
d
i
t
y
, Pur
p
ose, Pa
r
t
i
e
s, Servi
ce L
e
vel
ob
ject
i
v
es
(SLO
)
Pen
a
lties, Op
ti
o
n
a
l Serv
ices, Ad
m
i
n
i
stratio
n.
SLAs
are
created
,
m
o
n
ito
red
an
d u
tilized
in
u
tility co
m
p
u
tin
g
envi
ro
nm
ent
[1
6]
.
6.
SERVICE LE
VEL AGREE
M
ENT
FOR
SERVICE
-
O
R
IENTE
D
S
Y
STEMS
Serv
ice–o
r
ien
t
ed
arch
itectu
r
e is to
d
e
sig
n
an
d
b
u
ilt th
e Serv
ice Bases Syste
m
s. SLA sho
u
l
d
sp
ecify
th
e Qu
ality o
f
Serv
ice (Qo
S
)
related
to
sp
ecific ro
le
s.
Service Based
syste
m
s p
r
ov
id
e
valu
e-ad
d
e
d
serv
ices.
Web
Se
rvi
ce Descri
pt
i
o
n
La
ng
ua
ge (
W
SD
L), Si
m
p
l
e
Ob
j
ect
Access Protocol (SOAP)
t
echnologies a
r
e use
d
by Service-bas
e
d system
s. There are three is
sues s
u
ch
a
s
S
L
A l
i
f
e cy
cl
e m
a
nagem
e
nt
, SLA
vi
ol
at
i
o
n
–
i
m
pact
an
alysis,
SLA v
e
rification
and
Gen
e
ration
,
[1
7
]
.
Cl
o
u
d
com
p
u
tin
g
system
s are Utilit
y co
m
p
u
tin
g
reliab
l
e d
a
ta
stora
g
e system
s. Each Consum
er has Service Le
vel Agreem
ent specifies that, it has obli
g
ation
on
perform
a
nce and se
rvice’s
Quality
that is retrieve
d from
the syst
em
.
Cloud Com
puting system
pays
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 1
,
Febru
a
ry
2
015
:
15
8
–
16
5
16
4
penal
i
zat
i
o
n i
f
t
h
e co
ns
um
er’s
re
quest
s a
r
e n
o
t
sat
i
s
fi
e
d
on t
h
ei
r se
r
v
i
ce t
i
m
e
[18
]
. Vari
ous
r
u
l
e
s ar
e
i
n
t
r
o
d
u
ced
f
o
r
B
u
si
ness
R
u
l
e
s fo
r m
a
xim
i
zi
ng t
h
e R
e
ve
nue
o
f
Se
rvi
c
e
pr
o
v
i
d
e
r’s i
n
cl
ou
d c
o
m
put
i
ng
f
o
r
Neg
o
t
i
a
t
i
on a
n
d e
x
ecut
i
o
n t
i
m
e t
o
achi
e
ve
t
h
e b
u
si
ness
l
e
vel
o
b
ject
i
v
e
s
am
ong t
h
e
m
a
rket
an
d re
sou
r
c
e
layers. Clo
u
d
Perform
a
n
ce is
ad
d
e
d
To
Measu
r
e an
d
Us
e th
e Qo
S to
d
e
termin
e th
e av
ailab
ility,
liab
i
lity
in
case of failure
[20], [21].
Cloud
service Brokeri
ng allows t
o
access
t
h
e Cloud Res
o
urce
continuously am
ong
t
h
e C
ons
um
ers and Ser
v
i
ce p
r
o
v
i
d
e
r
s, m
a
nagem
e
nt
and
m
oni
t
o
ri
n
g
o
f
d
e
pl
oy
ed cl
o
u
d
servi
ces
on va
ri
o
u
s
C
l
ou
d ser
v
i
ce
pr
o
v
i
d
er
s fi
nal
l
y
C
l
oud
res
o
urces
sh
o
u
l
d
s
a
t
i
s
fy
t
h
e C
u
st
om
ers t
hose a
r
e speci
fi
e
d
by
SLA
[2
2]
. M
a
i
n
pu
rp
ose
o
f
t
h
e s
e
rvi
ce-
o
r
i
e
nt
ed
m
odel
i
s
m
a
nagi
ng
a
n
d
co
nt
r
o
l
l
i
ng t
h
e e
vol
ut
i
o
n
of
se
rvi
c
e
s
.
Ser
v
i
ce–
ori
e
nt
ed m
odel
uses t
h
e servi
ces t
o
devel
op t
h
e l
e
s
s
cost
, ev
ol
va
b
l
e, i
n
t
e
ro
pera
bl
e and m
a
gni
t
ude of
appl
i
cat
i
o
ns. Ser
v
i
ces can b
e
di
sco
v
ere
d
,
pu
bl
i
s
he
d,
des
c
ri
be
d. Ser
v
i
c
e
–
o
r
i
e
nt
e
d
m
e
chani
s
m
di
scov
ers an
d
in
vo
k
e
s th
e network av
ailable serv
ices to
co
m
p
lete
so
me task
. Serv
ice orien
t
ed en
gin
eering
and
serv
ice
m
o
d
e
l is essen
tial to
create sem
a
n
tic serv
ices an
d business process
specifications
.
Se
rvice
–
orie
nted
mech
an
ism
d
i
sco
v
e
rs and
invo
k
e
s th
e n
e
t
w
ork av
ailab
l
e
serv
ices to co
m
p
lete so
m
e
task
.
Best p
r
actices
are used
t
o
d
e
scrib
e
t
h
e pro
c
ess d
e
v
e
lop
i
ng th
at can
b
e
u
s
ed
m
u
ltip
le p
a
rtn
e
rs. Best
practices are as
follows Ide
n
tifying th
e Cloud
Actors, Ev
alu
a
tio
n
o
f
bu
si
ness lev
e
l p
o
licies, Metrics, Secu
rity,
Ide
n
t
i
f
y
i
ng
ser
v
i
ce m
a
nagem
e
nt
re
q
u
i
r
em
ent
s
, Pre
p
a
r
e a
n
d
m
a
nage ser
v
i
c
e fai
l
u
re
s [
1
4]
,
[1
9]
.
7.
SERVICE LE
VEL AGREE
M
ENTS IN B
I
G DAT
A
The
dat
a
g
r
o
w
i
n
g ra
pi
dl
y
by
t
h
e co
ns
u
m
ers, b
u
si
nes
s
and
g
ove
r
n
m
e
nt
ge
nerat
e
d
cont
e
n
t
.
T
o
main
tain
th
is d
a
ta th
ere are
ru
les
with
serv
ice lev
e
l agree
m
en
ts to
pro
t
ect th
e d
a
ta. Cap
acity, scalab
ility,
secu
rity, priv
acy, av
ailab
ilit
y are th
e issues o
f
d
a
ta
storag
e an
d
d
a
ta g
r
o
w
t
h
[2
3
]
. In
serv
ice orien
t
ed
arch
itecture env
i
ro
n
m
en
t qu
ality attrib
u
t
e need
s
p
l
ay m
a
j
o
r ro
le in selectin
g
t
h
e serv
i
ce. Sp
ecificatio
n of
serv
ice lev
e
l
ag
reem
en
ts p
r
o
v
i
d
e
s th
at to en
sure th
e serv
ices are
p
r
o
v
i
d
e
d
with
av
ailab
ility, sec
u
rity,
per
f
o
r
m
a
nce.
8.
CO
NCL
USI
O
N
Due t
o
ra
pi
d
gr
owt
h
o
f
C
l
ou
d C
o
m
put
i
n
g M
a
rket
, i
t
is pr
ovi
di
n
g
t
h
e new ser
v
i
c
e
s
wi
t
h
t
h
e
i
n
t
e
ract
i
on of cl
ou
d
ser
v
i
ce
pr
o
v
i
d
er
s and services
. Service level agree
m
en
t is p
r
ov
ided
fo
r th
e cu
sto
m
ers
who
are
u
tilizi
n
g
t
h
e clo
u
d
serv
ice to
p
r
o
v
i
de th
e q
u
a
lity
o
f
serv
ice. SLA
web
serv
i
ces are d
e
scrib
e
d
as
W
e
b
Service Level
Agreem
ent and
W
e
b Se
rvice
Agreem
ent.
Web
Ser
v
i
ce L
e
vel
A
g
reem
ent
i
s
di
vi
de
d i
n
t
o
t
h
ree
en
tities su
ch
as Parties, SLA p
a
ram
e
ters, Serv
ice Lev
e
l Obj
ectiv
es
(SLOs).
C
h
allen
g
e
s of Serv
ice lev
e
l
agreem
ent
s
(S
LA) as C
ons
um
er-dri
ven
Ser
v
i
ce M
a
na
gem
e
nt
, Dat
a
pr
ocessi
n
g
o
f
ri
sk m
a
nag
e
m
e
nt
,
In
de
pen
d
e
n
t
R
e
sou
r
ce m
a
n
a
gem
e
nt
. Servi
ce l
e
vel
ag
reem
ents are im
ple
m
ented in Service orie
nte
d
architecture.
REFERE
NC
ES
[1]
Peter Me
ll
, T
i
m
o
th
y Granc
e
,
“
NIST Definition o
f
Cloud Computing
”, Special Pub
lication
800-145
.
[2]
Judith Hurwitz,
Robin Bl
oor
, Marcia Kaufman,
C
l
oud Computing
for dummies
, HP
S
p
eci
al
edi
tion.
[3]
C.
N.
Hoefer,
G.
Karagiannis,
“
Taxonomy of clou
d computing
services
”.
[4]
Brusse
ls,
“
Cloud Computing S
e
rvice Level
Agreements
”, Jun
e
20
13.
[5]
Shu Zhang, Meina Song, Junde
Song, “
A Life Cycle Based
SLA
M
anagement Architecture
Desig
n
”.
[6]
“
User Guide: Cl
oud Computing
Contracts,
SLAs
and Terms
&
Co
nditions of
Use
”, August 2011.
[7]
S
t
eve M
c
Dona
l
,
“
Legal and Quasi-Legal Issues in Cloud Computing Contracts
”.
[8]
Pankesh Pate
l,
Ajith R
a
nab
a
hu,
Am
it Sheth,
“
Serv
ice
Lev
e
l
Agree
m
ent in
Cloud
C
o
mputing
”.
[9]
“
Public C
l
oud S
e
rvice
Agreemen
ts
”: What to
Exp
ect and
What
to
Negotiate, March 30, 2013.
[10]
Rajkumar Bu
yya, Saur
abh Kumar Ga
rg
, Rodrig
o N. Calheiros,
“
SLA-Oriented Resource Provis
ioning
for
Clou
d
Computing: Cha
lleng
es,
Ar
chitecture, and
Solu
tio
ns
”.
[11]
J
ean-Henr
y
M
o
r
i
n, J
o
ce
l
y
n Aub
e
rt, B
e
nj
am
in Gate
au, “
Towards Cloud Computing SLA Risk Ma
nagement: Issues
and Challenges
”
[12]
Mitchell Co
chran,
Paul d. Witman, “
Governance and
service level agreemen
t issues in
a
Cloud computin
g
envir
onment
”.
[13]
“
Creating Effective Cloud Co
mputing C
ontracts for the Federal Government
”, Feder
a
l C
l
oud Complian
ce
Committee, Feb
r
uar
y
24, 2012
.
[14]
Judith M. M
y
erson, “
Best practices to d
e
velop
SLAs for cloud
computing
”, 07
Janu
ar
y
2013.
[15]
Edward Wusten
Hoff, “
Ser
v
ic
e L
eve
l Agr
eement
i
n
the
Data C
e
nte
r
”.
[16]
Linlin
W
u
, R
a
jk
um
ar Bu
yya
,
Ser
v
ic
e Le
vel
Agre
e
m
ent (
S
LA)
in Utilit
y Comput
ing
Systems
.
[17]
Azlan Ism
a
il
, “
Temporal-aware Service
Le
vel Ag
reement manag
ement fo
r S
e
rvice-based systems
”.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Service Level
Agree
m
ents
in
Cloud Computi
n
g and Bi
g
Dat
a
(
K
Ra
d
ha)
16
5
[18]
Hadi Goudarzi,
Moham
m
a
d
Ghasem
azar, assoud Pedram
, “
S
L
A-based Optimization of Power and Migration Cost
in Cloud
Computing
”.
[19]
Andrew Ge
yer,
Melinda
McL
e
ll
an, Hunton
,
W
illiam
s
LLP,
“
Strategies fo
r Evaluating C
l
oud Computin
g
Agr
eements
”
[20]
Ma
rio Ma
c
´
ı
as, J. Oriol Fit´
o and Jordi Guitart,
“
Rule-based SLA Management f
o
r Revenu
e
Max
i
mization in Clo
ud
Computing Markets
”.
[21]
Lee Gillam, Bin
Li, and
John O'Loughlin, “
Addin
g
Cloud Performance to
S
e
rvic
e Level
Agre
ement
s
”.
[22]
Fouled Jrad, Jie
Tao
and Ach
i
m
Streit, “
SLA bas
ed service brokering in
interclou
d
environments
”.
[23]
Alnafoosi, A.B
,
Steinbach, T.
An
integr
ated fr
am
ework for evalu
a
ting big-d
a
t
a
s
t
orage s
o
lutions
-
IDA cas
e s
t
u
d
y
,
Science
and Info
rmation C
onfer
ence (SAI), 2013,947 – 956
[24]
S.
K.
Garg and
R.
Buyy
a.
Netw
ork Cloud Sim:
Modeling
Parallel App
lica
tions in Cloud Simula
tions
. Proceeding
s
of the 4
t
h I
E
EE/ACM Intern
ational Conf
eren
ce
on Utility
an
d Cloud Computing (UCC 20
11), Melbourn
e
,
Australia, December, 2011
[25]
S.
K.
Garg,
S.
K.
Gopalaiy
engar
,
and R. Bu
yy
a. “
SLA-based Resource Prov
isionin
g
for Heterogen
e
ous Workloads in
a Virtualized
Cloud Datacenter
”. P
r
o
ceed
ing
s
of the
11th
Intern
ation
a
l
Conferenc
e
on
Algorithm
s
a
nd
Architectures
for
Parallel Process
i
ng (ICA3
PP 2011), Melbou
rne,
Australia, Octob
e
r, 2011
.
[26]
Mitchell Coch
r
a
n, Pau
l
d. witman, “
governance and s
e
rvice level agreement issues in a
cloud computin
g
envir
onment
”,
a
public
ation
of
th
e a
ssociation
of
management.
[27]
Funmilade Faniy
i
, Rami Bahsoo
na, Ge
org
i
os Th
eodoropoulos, “A Dy
n
a
mic Da
ta-Driven Simulation Approach f
o
r
Preventing Service Level Agreement
Violations
in Cloud Federation”,
Internation
a
l Conference o
n
Computational
Scien
c
e, ICCS
2
012
.
[28]
Sobir Bazarbay
e
v
, Faraz Faghr
i,
Mark Overholt,
“Failure as
a
Ser
v
ice (FaaS) for
Hadoop Cluster
”
.
[29]
P. Rubach, M. S
obolewski,
Au
to
nomic SLA management in fe
derated computing
environments
,
i
n
: P
r
oceedings
o
f
the 2009 Intern
ation
a
l Confere
n
ce on Parall
el
Processing W
o
rkshops, ICPPW
’09, IEEE Com
puter Societ
y
,
Washington, DC
, USA, 2009, pp. 314–321.
[30]
Karl Cz
ajkows
k
i
,
Ian
F
o
s
t
er,
Car
l
Kes
s
e
lm
an
, V
o
lker S
a
nd
er, S
t
even
Tueck
e,
“
SNAP: A Proto
c
o
l
for Nego
tiating
Service Level Agreements and
Coordinating Re
source Management in Distribu
ted Systems
”, In
8th Workshop on
Job Scheduling
Strategies
for
Parallel Processing.
[31]
D. Battr_e, G. B
i
rkenheu
e
r,
et al,” Apply
i
ng
Risk
Manage
ment to Support
SLA
Provisioning”.
[32]
Alexander Keller, Heiko Ludwig, “
The WSLA Framework: Specif
y
ing and Mo
n
itoring Servi
ce
Level Agre
emen
ts
for We
b Se
rv
ic
e
s
”.
BIOGRAP
HI
ES OF
AUTH
ORS
K. Radha Pursuing PhD fro
m KL University
, Guntur. Completed M.Tech
from Anurag
Engine
ering Col
l
ege
,
Kodad and
B.Te
ch Degre
e
from
G. Nara
ya
nam
m
a
Institute
of Techno
log
y
&S
cien
ces
for
W
o
m
e
n, Hy
d
.
W
o
rked as
an As
s
i
s
t
ant P
r
ofes
s
o
r for 6
years
in L
a
qs
h
y
a
Ins
tiut
e
o
f
Techno
log
y
and
Sciences, Kham
mam.
Dr. B. Thirum
al
a Rao is
working as
a P
r
ofess
o
r
in KL University, Guntur
. Ph.D. in Com
puter
Science & Engineering
in
the
area of
Cloud C
o
mputing from Achar
y
a Nag
a
rjuna Univers
ity
.M.Tech (CSE) from Jawaharlal
Nehru University
, H
y
derab
a
d.
B.Tech (CSE)
from Achar
y
a
Nagarjuna Univ
ersity
.
Shaik Masthan
Babu working as an Assistant Prof
essor in Sri Sa
i Educ
ation
a
l So
cie
t
y’s since 6
y
e
ars and Pursuing PhD in KL
University
, Gun
t
ur. M.Tech fro
m Anurag Engineering Co
lleg
e
,
Kodad. B
.
Tech f
r
om Khader Memorial
College o
f
Engin
eering
,
Devarakond
a.
Evaluation Warning : The document was created with Spire.PDF for Python.