Negotiation-Based Flexible SLA Establishment with SLA-driven Resource Allocation in Cloud Computing

Abstract

As various consumers tend to use personalized Cloud services, Service Level Agreements (SLAs) emerge as a key aspect in Cloud and Utility computing. The objectives of this doctoral research are 1) to support a flexible establishment of SLAs that enhances the utility of SLAs for both providers and consumers, and 2) to manage Cloud resources to prevent SLA violations. Because consumers and providers may be independent bodies, some mechanisms are necessary to resolve different preferences when they establish a SLA. Thus, we designed a Cloud SLA negotiation mechanism for interactive and flexible SLA establishment. The novelty of this SLA negotiation mechanism is that it can support advanced multi-issue negotiation that includes time slot and price negotiations. In addition, to prevent SLA violations, we provided a SLA-driven resource allocation scheme that selects a proper data center among globally distributed centers operated by a provider. Empirical results showed that the proposed SLA negotiation mechanism supports faster agreements and achieves higher utilities. Also, the proposed SLA-driven resource allocation scheme performs better in terms of SLA violations and the provider's profits.

Full text

Negotiation-based Flexible SLA Establishment with
SLA-driven Resource Allocation in Cloud Computing
Seokho Son
School of Information and Communications
Gwangju Institute of Science and Technology
Gwangju, Republic of Korea
shson@gist.ac.kr
Sung Chan Jun
School of Information and Communications
Gwangju Institute of Science and Technology
Gwangju, Republic of Korea
scjun@gist.ac.kr
Abstract—As various consumers tend to use personalized
Cloud services, Service Level Agreements (SLAs) emerge as a
key aspect in Cloud and Utility computing. The objectives of
this doctoral research are 1) to support a flexible establishment
of SLAs that enhances the utility of SLAs for both providers
and consumers, and 2) to manage Cloud resources to prevent
SLA violations. Because consumers and providers may be
independent bodies, some mechanisms are necessary to resolve
different preferences when they establish a SLA. Thus, we
designed a Cloud SLA negotiation mechanism for interactive
and flexible SLA establishment. The novelty of this SLA
negotiation mechanism is that it can support advanced multi-
issue negotiation that includes time slot and price negotiations.
In addition, to prevent SLA violations, we provided a SLA-
driven resource allocation scheme that selects a proper data
center among globally distributed centers operated by a
provider. Empirical results showed that the proposed SLA
negotiation mechanism supports faster agreements and
achieves higher utilities. Also, the proposed SLA-driven
resource allocation scheme performs better in terms of SLA
violations and the provider’s profits.
Keywords-Cloud Computing; SLA Negotiation; Cost Models of
Cloud; Cloud Resource Allocation; Distributed Data Centers
I. INTRODUCTION
Cloud computing is an evolving paradigm to provide
consumers with a new utility as various computing services
(e.g., Software, Infrastructure, and Platform as a Service). In
the Cloud market, consumers are varied and thus have
personalized budget plans and requirements for service
quality. Also, Cloud service providers (CSPs) have different
resource capacities and marketing strategies. As various
consumers tend to use personalized services, Service Level
Agreements (SLAs) emerge as a key aspect in Cloud
computing. There are some standards to support SLAs such
as Web Service Agreement Specification (WS-Agreement)
[1]. However, SLA-driven Cloud computing designed to
enhance utility for both consumers and CSPs is not maturely
developed at this time.
Therefore, the objectives of this doctoral research
focused on two aspects of SLA-driven Cloud computing: 1)
supporting SLA establishment (SLA-E) that enhances the
utility of the agreements for both CSPs and consumers (i.e.,
negotiation-based SLA-E), and 2) supporting SLA
management (SLA-M) to prevent SLA violations (i.e., SLA-
driven resource allocation).
As participants in a Cloud may be independent bodies, in
order to establish a flexible SLA, some mechanisms must be
in place to resolve the different preferences of those entities.
A negotiation mechanism is effective in resolving those
different preferences. Whereas it is essential for both a
consumer and a CSP to reach an agreement on the price of a
service, when to use the service, and Cloud Quality of
Service (QoS) issues, to date there is little or no negotiation
support for Cloud service reservations with respect to
concurrent price, time slot, and QoS negotiation. The
purpose of this dissertation—to design a negotiation
mechanism that facilitates SLA-E—includes: 1) the design
of a multi-attribute negotiation mechanism that takes into
account concurrently: price, time slot and QoS, 2) tradeoff
algorithms that facilitate decision making in a multi-attribute
negotiation, and 3) a one-to-many negotiation mechanism to
facilitate distributed Cloud resource allocation.
In addition to facilitating SLA-E, it is important for CSPs
to manage limited resources to guarantee the SLAs. Existing
CSPs have been deploying and operating data centers
globally. Because the resource capacity of a data center is
limited, distributing the load to global data centers will
provide stable services. Although various load-balancing
algorithms have been developed, it is important to avoid
SLA violations (e.g., response time) when a CSP allocates
the load to data centers around the world. Considering load
balancing and guaranteed SLA, therefore, this dissertation
proposes 4) an SLA-driven Cloud computing to facilitate
resource allocation that takes into account the workload and
geographical location of distributed data centers.
II.SIGNIFICANCE OF OUR RESEARCH
Buyya et al. [2] addressed the necessity of SLA-driven
(oriented) resource allocation to realize Cloud and Utility
computing. They present the challenges and architectural
elements of SLA-driven resource management. Along with
[2], this research aims to enhance SLA-driven Cloud
computing. Whereas [2] provide a SLA-driven Cloud
framework that incorporates the challenges, it is important to
establish a well-adjusted and mutually agreeable SLA before
managing Cloud resources. Accordingly, we focused on both
SLA-E and SLA-M in Cloud computing (Fig. 1).
2013 13th IEEE/ACM International Symposium on Cluster, Cloud, and Grid Computing
978-0-7695-4996-5/13 $26.00 © 2013 IEEE
DOI 10.1109/CCGrid.2013.81
168

·;
·; ·; ·; ·; ·; ·; ·;
·; ·; ·; ·; ·; ·; ·; ·;
kGjG
OjGG
wG
tP
}G
tG
O}tP
jG
t
zshGl
zshGt
}tGt
w
zshGl
jV
i
yGh
zshGt
h
zG
·;
·; ·; ·; ·; ·; ·; ·;
·; ·; ·; ·; ·; ·; ·; ·;
·;
·; ·; ·; ·; ·; ·; ·;
·; ·; ·; ·; ·; ·; ·; ·;
·;
zshGl zshGl zshGl
Figure 1. Focus of the doctoral dissertation.
To enhance the utility of SLA-E, we designed a multi-
attribute negotiation mechanism that considers price, time
slot and Cloud QoS concurrently. Whereas CSPs such as
EC2 provide a pre-defined SLA that incorporates fixed price
(EC2 also supports auction-based spot price), fixed response
time, and some selective performance options, this may
restrict diversifying service types and expressing required
service level exactly. Thus, SLAs should be variable and
flexible to personalize service qualities by budget plans.
Hence, the significance of this research in leveraging
such limitations is that with the proposed multi-attribute
negotiation, CSPs can support flexible and interactive SLAs.
As the proposed mechanism includes time slot negotiation
capability, consumers and CSPs can express their temporal
preferences in SLAs. Lastly, it can be used as a pricing
method that takes into account changing price rates
according to market (supply/demand) and resource capability.
Also, it is important to guarantee the established SLAs.
As such, we developed an SLA-driven Cloud framework that
includes the automated SLA negotiation mechanism and a
workload- and location-aware resource allocation (i.e., initial
VM placement in a data center). Using the proposed system,
a consumer can establish the SLA with respect to service
price, time slot, and response time through an automated
SLA negotiation; further, a CSP can facilitate load balancing
using a pricing strategy. We documented the effectiveness of
SLA negotiation and SLA-driven resource allocation in
terms of SLA violations and the CSP’s profits in where a
CSP operates multiple data centers worldwide.
Because we provide a negotiation-based pricing model to
Clouds, our research is relevant to 1) the area of economic
and utility computing models for Clouds. Also, the SLA-M
scheme is included in 2) the topics on scheduling, load
balancing and resource management paradigms (both are
included in the CCGrid symposium topic areas).
III. RELATED WORK
As this work explores the issue of designing the
negotiation-based SLA-E and SLA-driven resource
allocations, areas related to this work include: 1) automated
negotiation mechanisms and frameworks applied to
Grid/Cloud and 2) SLA-driven resource allocation schemes.
1) Automated negotiation in Grid/Cloud computing:
There are several automated negotiation mechanisms for
Grid/Cloud (see [3] for a survey). Although there are single-
issue ([4][5]) and multi-issue negotiation mechanisms [6][7])
for Grid resource negotiation, none of these works considers
time slot negotiation. In many existing negotiation
mechanisms, a utility function is used to characterize a price
utility. The difference between this work and previous
researches that consider single [4][5] and multi-issue
negotiations without a specific tradeoff algorithm [6][7] is
that this work considers a price, time slot, and Cloud QoS
issue negotiation concurrently with the design of utility
functions and an advanced tradeoff algorithm. Venugopal et
al. [8] adopted a protocol for negotiating SLAs based on
Rubinstein’s alternating offers protocol [9] for the advance
reservation of Grid/Cloud resources. Whereas [8] proposed
time slot-based resource allocations, the resource allocation
focuses on finding a time slot that can be co-allocated; that
form of time slot negotiation is not addressed.
For SLA specifications, a meta-negotiation was proposed
by Brandic et al. [10] to allow two parties to reach an
agreement on what specific negotiation protocols to use
before starting the actual negotiation. [11] proposed a
declarative rule-based SLA language for describing SLAs
generically. Whereas [10][11] do not focus on specifying
negotiation strategies or designing utility functions for each
negotiation term, [12] proposed a framework for a Web
service composition that provides SLA negotiation for QoS
constraints. In [12], a utility function-based decision making
model is presented. [12] designed a single attribute utility
function for linear and monotonic QoS attributes. This
function is appropriate for generic attributes (e.g., price).
However, here we consider a time slot attribute that is
difficult to represent as a linear and monotonic utility
function. Also, we designed a trade-off algorithm to enhance
negotiation utility and speed.
2) SLA-driven Cloud computing that includes load
balancing in global data centers: Sotomayor et al. [13]
compared OpenNebula with several well-known virtual
infrastructure managers, including Amazon EC2, vSphere,
Nimbus, Eucalyptus, and oVirt. The comparison includes
resource allocation policies such as static-greedy, round
robin, and resource placement considering average CPU load.
While the placement focused on selecting a physical
machine at a data center, they did not focus on placement to
select a proper data center among global data centers. With
data centers, we need to consider SLA violations (e.g.,
response time) because of the network speed. Moreover,
existing CSPs such as EC2 do not employ sophisticated VM
placement for global data centers, and users themselves
manually select a data center at which to place their VMs.
Buyya et al. investigated energy-aware resource
provisioning and allocation algorithms to improve the energy
efficiency of the data center without violating the negotiated
SLA [14]. Whereas [14] provides a research direction for
resource allocation in Cloud, [14] does not consider a CSP
that operates distributed data centers to balance the resource
load and response time by geographical distance, and does
not provide a specific SLA negotiation. Le et al. [15]
considered load placement policies to manage center
temperatures among CSPs operating multiple data centers
worldwide. However, While [15] proposes dynamic load
distribution policies, there is no focus on the SLA guarantees.
169

IV. RESEARCH ACCOMPLISHMENTS
A. SLA-driven Cloud computing
ʹΝΠΦΕΓΣΠΜΖΣ ʹΝΠΦΕΡΣΠΧΚΕΖΣ
ʹΝΠΦΕΡΣΠΧΚΕΖΣ
ʹΝΠΦΕΡΣΠΧΚΕΖΣ
ʹΝΠΦΕΡΣΠΧΚΕΖΣ
ʹΝΠΦΕΡΣΠΧΚΕΖΣ
͵ΒΥΒΔΖΟΥΖΣ
͵ΒΥΒΔΖΟΥΖΣ
͵ΒΥΒΔΖΟΥΖΣ
ʹΝΠΦΕΔΠΟΤΦΞΖΣ
ʹΝΠΦΕΓΣΠΜΖΣʹΝΠΦΕΔΠΟΤΦΞΖΣ
ʹΝΠΦΕΓΣΠΜΖΣʹΝΠΦΕΔΠΟΤΦΞΖΣ
΄ͽͲͶΤΥΒΓΝΚΤΙΞΖΟΥ
XaXGzshGl
XauGzshGl
XaXGzshGl XaOXatPGzshGt
jGj
zshGlG
MGtGG
jGj
XaXGzshG
l
XauGzshG
l
XaOXatPGzshG
t
΄ͽͲΕΣΚΧΖΟ΃ΖΤΠΦΣΔΖ
ͲΝΝΠΔΒΥΚΠΟ
΄ͽͲͶΤΥΒΓΝΚΤΙΞΖΟΥ
΄ͽͲͶΤΥΒΓΝΚΤΙΞΖΟΥ
Figure 2. Design models for SLA-driven Cloud computing.
Fig. 2 shows design models included in this research.
Each model consists of consumer, broker, and CSP. This
includes SLA-E between a consumer and CSP (1:1 SLA-E),
SLA-E for a consumer among multiple CSPs (1:N SLA-E),
and a 1:1 SLA-E model with resource allocation to multiple
data centers (SLA based 1:1:M). In this paper, we introduce
the methodology and the major research accomplishments
published in [16][17] for 1) Negotiation-based SLA-E and 2)
SLA-driven resource allocation in global data centers.
B. Negotiation-based SLA-E
A negotiation mechanism consists of a protocol, strategy,
and utility functions. The protocol is a set of communication
rules for negotiations. The negotiation mechanism in this
work follows Rubinstein’s alternating offers protocol [9],
which permits agents to make counter-offers to their
opponents in alternate rounds. Both agents generate counter-
offers and evaluate their opponent’s offer. Counter-proposals
are generated by the strategy (concession and tradeoff). A
concession algorithm determines the degree of concession
for each negotiation round, and a tradeoff algorithm is
required to generate a proposal in multi-issue negotiation.
The tradeoff algorithm generates a proposal by combining
proposals for individual issues. If the negotiation issues are
price and response time, a proposal is a combination of price
and response time (e.g., low price with slow response, or
high price with fast response). Unlike existing mechanisms
can make only one proposal at a time, in this study, agents
are allowed to make multiple proposals concurrently in a
round that generated the same aggregated utility (i.e., ‘burst
proposal’ [16]), differing only in terms of individual utilities.
The utility function
()Ux
represents an agent’s level of
satisfaction with negotiation outcome x (e.g.,
()UP
for price).
For a decision-making, agents evaluate proposals according
to the utility function. To define a price utility function, the
negotiator needs to specify the most and the least preferred
price. In general, the range of the utility function is
min
{0} [ ,1]u
,
where
min
()UP u
and
() 1UP
represents the least and the most
preferred price, respectively.
The time slot utility function defined in [16] supports
participants in representing the temporal preferences for
leasing/lending services. A consumer can specify the time
slot utility function according to his/her work schedule, and a
CSP can specify the time slot utility according to the
expected resource demands at any given time. CSPs may
charge a higher price at peak time and a lower price at off
peak, and consumers may need to pay a higher price to use a
service in more desirable time slots. Fig. 3 shows an example
of generated time slot utility function. The consumer who
uses this function will have the highest utility at 15T.
XUW
W
TS
WU^\
WUY\
min
u
T
C
T
U
()
C
T
UT
5
15
25
35
40
C
FT
C
LT
Figure 3. Example of generated time slot utility function.
The service response time represents the minimum
response time that a CSP offers. Let the initial response time
(IRT) and reserve response time (RRT) be the most and least
preferred response time, respectively. The response time
given to a consumer can be evaluated by the response time
utility function of a consumer, as follows:
(1 ) ,
()
0 , o.w.
C
min min C C
C
RT CC
RRT RT
u u IRT RT RRT
URT RRT IRT

   





(1)
Finally, the aggregated utility function, which includes
service price, time slot, and response time, is as follows:

0, ( 0, 0, 0)
,,
, o.w.
PTS RT
Total
P P TS TS RT RT
UUorU
UPTSRT
wU w U w U
 

   

(2)
Fig. 4 shows empirical results of the proposed SLA
negotiation in terms of negotiation speed and utility using an
agent-based Cloud testbed [16]. The proposed burst mode
(B10, B50, and B100) and the adaptive burst mode (AB)
achieved a higher average total utility and a faster agreement
speed than related schemes (middle: M1, random: R1, and
heuristic: H1[12]) that can generate only one proposal in
each negotiation round.
0
0.2
0.4
0.6
0.8
1
0.1 0.3 0.5 0.7
0.9
Negotiation speed
Cloud Load
λ
c
=1/3, λ
p
=3.0, τ
c
=50, τ
p
=200
M1 R1 H1 B10
B50 B100 AB
0
0.2
0.4
0.6
0.8
1
0.1 0.3 0.5 0.7
0.9
Total utility
Cloud Load
λ
c
=1/3, λ
p
=3.0, τ
c
=50, τ
p
=200
M1 R1 H1 B10
B50 B100 AB
(a) (b)
Figure 4. Simulation results: effect of the proposed trade-off algorithm.
C. SLA-driven resource allocation in gloval data centers
zG
zshGu
|G

wG

wG
 zG

jGi jGzGw
zshGu
|G

wG

wG
 zG

y

zshGt
kGG

wGG

wGG

wGG
GOwttP
kG
y
G

ͿΖΘΠΥΚΒΥΚΠ
ΟΤΖΤΤΚΠΟ
΃ΖΤΖΣΧΒΥΚΠ
ΟΣΖ΢ΦΖΤΥ
͵ΚΤΔΠΧΖΣΪ
ΣΖ΢ΦΖΤΥ
wG
wtt
wG
kG
wG
wtt
wG
kG
wG
wtt
wG
Figure 5. SLA negotiation and management based Cloud framework.
The proposed framework [17] consists of a service
broker and CSP (Fig. 5). The broker connects a consumer to
the CSP who owns the service discovered and has SLA-E
capability through the SLA negotiation component. A CSP
170

consists of (1) reservation controller, (2) SLA negotiation,
(3) SLA-M, and (4) distributed data centers. The SLA-M
component, which is called the workload- and location-
aware resource allocation (WLARA), selects a center among
the global data centers to allocate the requested service. The
conditions (i.e., utility-based evaluation [17]) of selecting a
data center are based on workload and the SLA (service
response time in this work). Each data center includes a
physical machine manager, who manages the physical
computing nodes of a data center to evaluate the average
response time of a data center. Using the monitoring, SLA-M
selects a data center and specific physical computing node.
Fig. 6 shows the performance of WLARA and other
schemes in terms of SLA violations and placement failures
[17]. Fig. 6(a) shows agreed and measured response time in
WLARA. Consumers have different response time
thresholds according to the outcomes of the negotiated SLA.
In Fig. 6(b), with WLARA, the least number of SLA
violations is guaranteed, whereas the greedy, random, RR,
NIM, and IM (a similar way with EC2) schemes caused
more violations because WLARA considers both workload
and response time (including network delay) in a utility
function. Hence, WLARA can allocate a consumer’s request
to a data center that has a lower workload and is physically
closer to guarantee the response time threshold in the SLA.
0
100
200
300
400
1 301 601 901
Response time (ms)
User requests
Reponse Time (Measured and SLA)
Workload Delay
Network Delay
Agreed Response Time (SLA)
6200 6257
1279
6374
1439
2621
30
652 1952
1060
0
2000
4000
6000
8000
10000
Consumers (n)
Violations and Placement failures
SLA Violation
Placement Failure
(a) (b)
Figure 6. Simulation results: SLA-driven resource allocation.
V. CONCLUSIONS AND FUTURE WORK
The novelty and significance of this study are 1) the
design of a multi-issue negotiation mechanism that facilitates
the price, time slot, and QoS negotiation for SLA-E, and 2)
the development of a SLA-driven Cloud framework that
includes the SLA negotiation mechanism and a workload-
and location-aware resource allocation to manage SLAs.
The expected contributions of this research are as
follows: 1) while the variety of SLA options is limited within
enforced SLA strategies, the different preferences of a
consumer and CSP can be narrowed efficiently through the
proposed SLA negotiation; 2) the time slot negotiation can
provide a market-based pricing scheme, and we observed
that the proposed mechanism as a pricing scheme has
advantages over the pricing schemes used in EC2 [16]; 3) the
design of tradeoff algorithms considers the tradeoff
relationship among utilities to enhance utility and negotiation
speed. Also, to prevent SLA violations, 4) we provide a
SLA-driven resource allocation that selects a data center
among globally distributed data centers operated by a CSP.
Finally, the authors expect this work can be extended in
two ways: 1) considering and specifying additional
negotiation issues in Cloud SLAs and 2) deploying the
proposed system on a real infrastructure and evaluating the
performance with real workloads ([16] includes a case study).
ACKNOWLEDGMENTS
This work was supported by the National Research
Foundation of Korea (NRF) grant funded by the Korean
government (MEST; No. 2010-0026438) and by PLSI
supercomputing resources of the Korea Institute of Science
and Technology Information. Thanks to Prof. Kwang Mong
Sim for his valuable advice.
REFERENCES
[1] A. Andrieux, K. Czajkowski, A. Dan, K. Keahey, H. Ludwig, T.
Nakata, J. Pruyne, J. Rofrano, S. Tuecke, and M. Xu, Web Services
Agreement Specification (WS-Agreement), Open Grid Forum, 2006.
[2] R. Buyya, S. Garg, and R. Calheiros, “SLA-Oriented resource
provisioning for cloud computing: challenges, architecture, and
solutions,” In Proc. International Conference on Cloud and Service
Computing, pp. 1–10, 2011.
[3] K. M. Sim, “Grid resource negotiation: survey and new directions,”
IEEE Trans. Syst., Man, Cybern. C, Applications and Reviews, vol.
40, no. 3, pp. 245–257, May 2010.
[4] R. Lawley, M. Luck, K. Decker, T. R. Payne, and L. Moreau,
“Automated negotiation between publishers and consumers of grid
notifications,” Parallel Proc. Lett., vol. 13, no. 4, pp. 537–548, 2003.
[5] K. M. Sim, “G-commerce, market-driven G-negotiation agents and
Grid resource management,” IEEE Trans. on Syst, Man and Cybern.
B, Cybernetics, vol. 36, no. 6, pp. 1381–1394, Dec. 2006.
[6] F. Lang, “Developing dynamic strategies for multi-issue automated
contracting in the agent based commercial grid,” in Proc. IEEE Int.
Symp. Cluster Comput. Grid (CCGrid05), Workshop Agent -Based
Grid Econ. Cardiff, U.K. pp. 342–349, May 2005.
[7] H. Gimpel, H. Ludwig, A. Dan, and R. Kearney, “PANDA”:
Specifying policies for automated negotiations of service contracts,”
in Proc. ICSOC, vol. 2910, LNCS, New York, pp. 287–302, 2003.
[8] S. Venugopal, X. Chu, R. Buyya, “A negotiation mechanism for
advance resource reservation using the alternate offers protocol,” in
Proc. 16th Int. Workshop on Quality of Service (IWQoS 2008),
Twente, The Netherlands. June 2008.
[9] A. Rubinstein, “Perfect equilibrium in a bargaining model,”
Econometrica, vol. 50, no. 1, pp. 97–109, 1982.
[10] I. Brandic, D. Music, and S. Dustdar, “Service mediation and
negotiation bootstrapping as first achievements towards self-
adaptable grid and cloud services,” In Proc. Grids meet Autonomic
Computing Workshop (GMAC ) 2009, Spain, June 2009.
[11] A. Paschke, J. Dietrich, K. Kuhla, “A Logic Based SLA Management
Framework,” in Proc. Semantic Web and Policy Workshop (SWPW),
4th Semantic Web Conference (ISWC 2005), Galway, Ireland, 2005.
[12] J. Yan, R. Kowalczyk, J. Lin, M. B. Chhetri, S. K. Goh, and J. Zhang,
“Autonomous service level agreement negotiation for service
composition provision,” Future Generation Computer Systems, vol.
23, no. 6, pp. 748–759, July 2007.
[13] B. Sotomayor, R. S. Montero, I. M. Llorente, and I. Foster, “Virtual
infrastructure management in private and hybrid clouds,”IEEE
Internet Computing, vol. 13, no. 5, pp. 14–22, 2009.
[14] R. Buyya, A. Beloglazov, and J. Abawajy, “Energy-efficient
management of data center resources for cloud computing: A vision,
architectural elements, and open challenges,”In Proc. the
International Conference on Parallel and Distributed Processing
Techniques and Applications (PDPTA2010), pp. 6–20
[15] K. Le, J. Zhang, J. Meng, R. Bianchini, T. D. Nguyen, and Y. Jaluria,
“Reducing electricity cost through virtual machine placement in high
performance computing clouds,” In Proc. 2011 Super Computing
(SC11), Washington, USA.
[16] S. Son and K. M. Sim, “A price and time slot negotiation mechanism
for cloud service reservations,”IEEE Trans. on Systems, Man and
Cybernetics, B: Cybernetics, vol. 42, no. 3, pp. 713–728, June 2012.
[17] S. Son, G. Jung, and S. C. Jun, “An SLA-based cloud computing that
facilitates resource allocation in the distributed data centers of a cloud
provider,” The Journal of Supercomputing, Jan. 2013. doi:
10.1007/s11227-012-0861-z.
171