VM Placement Algorithm .pdf



Nom original: VM Placement Algorithm.pdf

Ce document au format PDF 1.6 a été généré par Adobe Acrobat Pro DC 15.10.20060, et a été envoyé sur fichier-pdf.fr le 19/07/2016 à 10:39, depuis l'adresse IP 41.230.x.x. La présente page de téléchargement du fichier a été vue 3739 fois.
Taille du document: 790 Ko (7 pages).
Confidentialité: fichier public




Télécharger le fichier (PDF)










Aperçu du document


2016 International Conference on Digital Economy (ICDEc)

VM Placement Algorithm Based on Recruitment
Process within Ant Colonies
Faten Ben Hassen

Zaki Brahmi and Hajer Toumi

Higher Institute of Computer and
Communication Techniques Hammam Sousse
Sousse University,Tunis
Email: faten.bh90@gmail.com

RIADI-GD Laboratory, National School of Computer Sciences
Manouba University, Manouba,Tunisia
Email: zakibrahmi@gmail.com
Email: hajertoumi1990@outlook.com
to shut down the empty PM (that does not contain any VM).
There’s a clear difference between the bin packing problem
and VM placement problem. While the former allows placing
items one above the other, within the latter, this is not an
option. This is because once a resource is utilized or occupied
by a VM, it cannot be reused by any other VM [23].
Therefore, it’s easy to observe that placing several VM in the
same PM will increase its resources exploitation provided that
the overall resources needed by the set of VM doesn’t exceed
the capacity delivered by the PM. However it’s more beneficial
to consider the VM that will ensure load balance.
We explain: if we consider, for simplicity, two resources
CPU and MEMORY, then, the PM capacity can be presented
according two dimensions (Fig. 1). To ensure load balance, it’s

Abstract—The Cloud computing forges the shape of the
current era and the following ones based on delocalized IT infrastructure and sharing resources. However, the rebellious rise of
cloud computing comes with concerns over energy consumption.
Numerous reports which inspected Cloud energy consumption
showed that the Cloud is an energy monster, specifically the
data centers that holds 2% of overall energy consumed in the
world on 2011 [6]. More closely, it has been proven that the
servers (Physical machines PM) are the most energy-hungry
elements of the data center [7]. Server consolidation based on
virtualization is a key mechanism for energy consumption taming.
Within this context, our aim in this paper is to propose a VM
Placement Algorithm Based on Recruitment process within Ant
Colonies proposed by Bonabeau [12] that seeks to maximize PM
resources exploitation along with maximizing resources balance.
The exprimental results showed that our algorithm generates
always a significantly good solution.
Keywords—VM Placement, Maximizing resources utilisation,
Load balance, Recruitment process

I.

I NTRODUCTION

Cloud computing was built upon a decentralized architecture that relies on sharing resources. This trend has received
a very warm welcome from the IT industry and becomes the
pedestal of new technologies. Yet, this trend comes with a
price that is payed at the expense of the environment and the
operational cost of cloud. In fact, cloud computing is very
hungry for energy [7]. This hunger focuses more specifically
on data centers. In order to mitigate the energy consumption,
cloud uses virtualization.
Virtualization is one of the technologies that constitutes the
cloud computing. The basic idea behind it, is to make it possible to run multiple operating systems and multiple applications
on the same PM at the same time [30]. The key component
of virtualization is the virtual machine (VM) that hosts the
operating system. The virtualization technology brings us to
PM consolidation where the main objective is to maximize
resources exploitation by multiplexing several VM in the same
PM. Such technique allows keeping the minimum number of
active MP and thus lower energy consumption.Our concern
in this paper is the VM placement step that is an NP-hard
problem [19].
This problem is similar to bin packing problem [23] where the
objective is to place a set of items within the minimum number
of bins. By analogy to VM placement, the main objective is to
place the set of VM in the minimum number of PM in order
978-1-5090-2230-4/16/$31.00 ©2016 IEEE

Fig. 1: Simulation results for the network.
crucial to place a VM that its workload respects the shape of
the remained capacity. This will prevent from maximizing the
use of a resource at the expense of the other and thus it will
increase the likelihood of placing more VM in the same PM.
VM placement problem is considered prominently important in
cloud computing field and thus it drew the attention of several
researchers [23] [11] [19] [12] [39] [37] [5] [21] [13] [14] [1]
[2] [9] [35] [36] [15] [4] [25] [10] [32] [8] [18] [34].
In this paper, we focus on VM placement problem based on
maximizing PM resources utilization and load balance objectives. We propose VM Placement algorithm (VMPRP) based
on Recruitment Process within ant colonies that considers three
type of resources (CPU, memory and I/O).
In the next section we shall discuss related works. Section 3
aims to describe VM placement problem statement. Section
4 will discuss our proposed solution based on the recruitment
process within Ant Colonies. Section 5 will present experiment
results and analyses. Finally we wind up with conclusion and
1

based on vector algebra multidimensional based load balancing
proposed by [23]. To ensure resource balance they measure the
resource imbalance vector (RIV) using magRIV. [23] proposes
a novel methodology based on vector arithmetic to ensure
resources balance. Aside, they proposed a strategy to achieve
efficient utilization of PM while underwriting PM balance per
resource using resources utilization metric. Therefore, they
also proposed a vector based consolidation heuristic that seeks
to find the highest load PM that has complementary resource
usage with respect to the VM, in this case, the proposed
heuristic outbalance the number of active PM over the load
balance of PM.
We can clearly see that the works that address the load
balance are far outnumbered by the works that only deal with
maximizing resources usage. Moreover, only few works have
adopted a decentralized approach despite that such approach
can reduce the computing process by dividing the computation
burden over more than one performer.
In this paper, we consider the multi-dimensional version of VM
placement problem where PM capacity and VM requirements
are presented according to multiple resources. Mainly, our aim
in this paper is to propose a decentralized VM placement
approach that seeks to maximize PM’s resources exploitation
along with maximizing resources balance. The proposed algorithm is based on recruitment process within ant colonies and
considers three types of resources (CPU, memory and I/O).For
each server, we introduce a threshold for each resource type
that must not be exceeded. Otherwise, to tackle load imbalance
issue, we consider vector magnitude to measure imbalance rate
as proposed by [21].

possible future research directions.
II.

R ELATED W ORKS

A major part research addresses this problem using heuristic or/and meta-heuristic to place VM in a minimum number of
PM Such as Greedy approaches, genetic approaches, particle
swarm optimization.
Earlier attempts used greedy heuristics such as first fit [32]
[15], best fit (BF) [34] [39] and modified best fit decreasing
(MBFD) [1] [26] [23] [3] [2] [35] [36].
[4] proposed a based knapsack problem approach and an evolutionary computation (EC) heuristic that periodically perform
the VM consolidation to optimize the current placement and
maintain the minimum number of active PM. [17] formulated
VM placement problem as stochastic integer programming and
proposed a placement algorithm based on knapsack problem
approach. Besides, various works were directed towards genetic approach. [29] formulated VM placement problem as
vector packing problem with conflict and tries to solve it
with a variant from the traditional genetic algorithm (GA),
a grouping genetic algorithm. [18] proposed an improved
genetic algorithm based on fuzzy multi-objective evaluation.
[10] used a genetic approach to solve multidimensional bin
packing VM placement. [37] proposed a genetic distributed
approach to mitigate processing time. [9] proposed a novel
genetic approach which is a variation of the genetic algorithm approach based on distributed scheme in an attempt
to overcome typical limitations of genetic algorithms such as
premature convergence and high processing time.
Despite the fact that the GA has been proven better than the
greedy approach [24], the research proved that ACO (Ant
Colony Optimization) is more effective approach. [21] [5] [14]
propose an ACO algorithm to solve VM placement problem
as an instance of multi-dimensional bin packing. In their
next work [13] extend their research along with distributed
approach. [38] proposed a multi-objective ACO algorithm
seeking to find pareto set (non-dominant solutions) that minimizes resources wastage along with power consumption. This
multi-objectives version of ACO proves more efficiency than
other ACO algorithm and genetic algorithm.
[25] [8] proposed an approaches based on partical swam
optimisation (PSO) that simulate The behaviour of the insects.
[8] proposed a based particle swarm optimization within a
multi-objective version (MOPSO).
The above-mentioned works was headed for maximizing PM
resources utilization without considering load balance. Some
research tackles this aspect. [31] proposed best fit decreasing
heuristic that uses Bin Centric heuristic [28] to maintain
resource balance throughout VM placement. The heuristic uses
scalar size which is calculated as weight sums of the respective
vector components.
[11] used metric with weighted sum of resources to measure
load imbalance of PMs and a greedy approach to achieve VM
placement. [36] addressed the imbalanced issue of multidimensional resources using multi-dimensional space partition
model and propose algorithm (EAGLE) that can balance
the utilization of multidimensional resources and reduce the
number of active PM.
[21] tried to dim resource wastage while minimizing power
consumption. They proposed the later adaptation of ACO
(ACS: ant colony system) with balanced usage of resources

III.

VM P LACEMENT P ROBLEM S TATEMENT

Formally the VM Placement Problem VMPP is described
by the triple V M P P = hP M, V M, af f i


P M = {p1 , p2 , ..., pm } the set of m PMs.
Each PM pj ∈ P M (j ∈ [1..m]) possesses a ddimensional Server (PM) Capacity Vector SCVpj .
SCVpj = hCp1j , Cp2j , ..., Cpdj i where Cpkj denotes the
total capacity of resource k; k ∈ [1..d].



V M = {v1 , v2 , ..., vn } the set of n virtual machines.
Each virtual machine vi ∈ V M possesses a ddimensional Resources Demand Vector RDVvi .
RDVvi = hDv1i , Dv2i , ..., vvdi i where Dvki denotes the
total capacity of resource.
The total PM utilisation is described as the
sum of VM demands and it possesses a ddimensional PM Utilisation Vector SU Vpj .
SU Vpj = hUp1j , Up2j , ..., Updj i.
Upkj =

n
X

Dvki ∀vi ∈ V Mpj

(1)

i=1

V Mpj denotes the VM set that belongs to PM pj .


af f designates the VM packing function such that
af f : P M × V M =⇒ {hvi − Pj i}; vi ∈ V M , pj ∈
PM

Let’s

2

Aj depict if the PM pj is active or not

(
Aj =

0
if
1 if not

V Mpj = ∅

The objective function is defined as follows (3)
m
X
M in
Aj

A sub-response SRIvi ,pj is calculated which denotes the
remaining quantity of resources if vi migrates to pj . The PM
with a minimum SRI increases the most the resources use.
SRIvi ,pj = hEv1i ,pj , Ev2i ,pj , ..., Evdi ,pj i such that

(2)

Evki ,pj = skpj − Dvki

(3)

(7)

j=1

In the proposed solution, the VM individually select a PM
to where it must migrate. To complete this process each
vi calculates its internal response on the basis of received
stimulus. If the internal response is negative, the coupling
(vi , pj ) violates the SLA constraint.
The Internal Response IRvi ,pj is calculated as follows:

SC
n
X

xvi .pj ≤ 1∀i ∈ [1..m]

(4)

j=1

Upkj ≤ βpkj ; k ∈ [1..d]

(5)

The constraint (4) imposes that the VM must be placed in
only one PM at once. In order to comply the SLA constraints,
a d-dimensional Resource ustilisation threshold Vector RUT is
introduced. RU Tpj = hβp1j , βp2j , ..., βpdj i is fixed for each PM
and must not be exceeded during the VM packing process.
βpkj denotes the use threshold of resource k. The constraint
(5) insists that the overall PM utilization must be lower than
the fixed threshold β.
IV.

IRvi ,pj =

The proposed solution is mainly based on Swarm Intelligence. As a matter of fact, Swarm is a large number of
homogeneous simple agents interacting locally among themselves and their environment, with no center to allow a global
interesting behavior to emerge [16]. Swarm-based algorithms
have recently emerged as a family of nature inspired population
based algorithms that are capable of producing low, cost, fast
and robust solutions to several complex problems [20] [?].
Nowadays, several models of swarm intelligence based on
different natural systems have been proposed in the literature.
Some of most famous algorithms are ACO, PSO, Articial Bee
Colony, Cat Swarm Optimization and Articial Immune System.
More closely, we adopt recruitment process in ant colonies.
More specifically, we assume that the PM will recruit a set of
VM such that a maximized and balanced resources exploitation
across different computing resources is maintained with the
goal of minimizing power consumption and resource wastage.
However, The VM packing process must not overload the PM
and violate the SLA (Service Level Agreement) constraint. In
our proposed solution we adopted the two following heuristics:
Each VM migrates to the PM that has the minimum
unused resources without exceeding resources use
threshold. The purpose of this heuristic is to maximize
physical resources exploitation.



Each VM migrates to the PM that decreases further
the imbalance rate of resources use. This is feasible
by using magnitude imbalance vector.

M=

k=1
X
d

Dvki
d

(9)

In our case, d = 3 (3 types of resources).
When selecting the PM for packing a VM, the couple VM,PM
that has the smaller magnitude of RIV is the one that mostly
balances the resources utilization of the server across different
dimensions.The magnitude of RIV is given as follows:
mag(RIVvi ,pj ) =
q
(Ev1i ,pj + M )2 + (Ev2i ,pj + M )2 + ... + (Evdi ,pj + M )2
(10)
Each VM calculates the tendency Tvi ,pj which is determined
by both stimulus and internal response. The tendency reflects
the ability of the PM sj to host vi and is given by:

(spj )2

if f (SRIvi ,pj ) ≥ 0
2 +(IR
2
(s
)
p
vi ,pj )
j
Tvi ,pj =

0 else
(11)
(
0 if ∃Evki ,pj < 0
Pd
f (SRIvi ,pj ) =
(12)
k
else
k=1 Evi ,pj

Our model considers 3-dimensional resources (CPU, memory,
and I/O) as relevant PM resources. Each PM pj ∈ P M
produces and distributes a stimulus across the set of VM. The
stimulus Spj is determined by the available resources quantity
and is calculated as follows.
Spj = hs1pj , s2pj , ..., sdpj i such that skpj denotes the available
quantity of the resource k and is given by:
skpj = (1 − βpkj )Cpkj − (Cpkj − Upkj )

(8)

f (SRIvi ,pj ) is a function that calculates the sum of Evki ,pj for
k ∈ [1..d].
RIV captures the degree of imbalance in the current utilization
of a PM. In fact, capturing the measure of overall resources
utilization across multiple resource types is one of the most
important factors, saturation of only one resource type can lead
to no further improvement in utilization while leaving other
types of resources underutilized [14].
We resort to a d-dimensional Resources Imbalance Vector RIV
as follows:
RIVvi ,pj = h(Ev1i ,pj + M ), (Ev2i ,pj + M ), ..., (Evdi ,pj + M )i

P ROPOSED S OLUTION



f (SRIvi ,pj ) + mag(RIV vi , pj )
2

The tendency increases and opts toward 1 once the IR
decreases and opts toward 0. The V M seeks to maximize
resources utilization of P M , thus it must migrate toward the
P M with maximum tendency. However, if f (SRIvi ,pj ) < 0,
that means that the couple (vi , pj ) violates pj capacity. In
such case the tendency will immediately takes the value 0.
f (SRIvi ,pj ) receives 0 once there’s a Evki ,pj < 0.

(6)
3

V.

A. Proposed Algorithm VMPRP
Since our proposed solution focuses on maximum and
balanced resources utilisation, each VM must migrate to the
PM that provides maximum tendency value. The algorithm
pseudo code is shown Algorithm 1.

P ERFORMANCE E VALUATION

Due to the lack of access to large test beds or real Cloud
infrastructure, we resorted to simulation-based evaluation using the CloudSim. In order to evaluate the performance of
the proposed VMPRP solution, we compare it with the two
following solutions:

Algorithm 1 VMPRP Algorithm
Require: V M = {v1 , v2 , .., vi }, P M = {p1 , p2 , .., pj }
Ensure: V M P M = {hvi , pj i}: the set of pairs solutions
{Initialization}
1: V M S := ∅
2: L.tendencies := ∅
3: for all pj ∈ P M do
4:
Broadcast(spj ) {This function allows to a PM to
broadcast a stimulus to the set of VM}
5: end for
{these instructions are executed in by all VM }
6: for all pj do
7:
SRIvi ,pj := ComputeSRI(Spj , Dvi ) {This function
takes the stimilus and vi ’s demand as input and computes the remaining quatity of resources SRIvi ,pj if vi
is placed in pj }
8:
if f (SRIvi ,pj ) ≥ 0 then
9:
10:
11:
12:
13:
14:
15:
16:
17:
18:
19:

f (SRIv

,p

)+mag(RIVv

i j
i
IRvi ,pj :=
2
2
(sp−j )
Tvi ,pj := (sp−j )2 +(IR
2
vi ,pj )
else
Tvi ,pj := 0
end if
L.tendencies.add(Tvi ,pj )
end for
T max := F indM ax(L.tendencies)
if T max ≥ 0 then
V M P M.add(hvi , pj i)
end if

Search of the maximum value of tendency

A random solution based on a random selection of a
PM to host the VM. The algorithm 2. Presents the
pseudo code of the random algorithm.

Since the size of the VM placement problem depends
strongly of the number of both VM and PM,we prepared two
test problems groups. In the first test group, the number of PM
was set at 1000, while the number of VM varies between 1500
and 6000 with a pitch of 500. In a second group, the number of
VM has been set at 3500, while the PM varies between 800 and
2600 with a pitch of 200. For each test problem, we randomly
generated the requested resources (CPU, memory and I/O)
in the interval [1000, 5000] and the resources capacity in
the interval [10000, 50000]. Simulation is conducted through
the 20 test problems previously set, and each simulation was
repeated several times, the results are generated after taking the
average.The results are calculated according to the following
evaluation criteria:

The complexity of our algorithm deponds on the complexity of:




Require: V M = {v1 , v2 , , vi }, P M = {p1 , p2 , , pj }
Ensure: V M P M = {hvi , pj i}: the set of pairs solutions
{Initialization}
1: V M S := ∅
2: for all ∀vi ∈ V M do
3:
Randomly choosing a PM pj ∈ P M
4:
if PM capacity exceeds VM needs then
5:
V M P M.ADD(vi , pj )
6:
else
7:
return to 2
8:
end if
9: end for

B. Algorithm Complexity

Computation of the tendency

The time-shared solution offered by CloudSim [27]
where requested resources are allocated to each VM
for a certain period of time.

Algorithm 2 random algorithm

,pj )





The first function seeks to compute the tendency for each
pj to receive vi . This calcul is performed after computing
IR. Once the tendency is computed, it will be added to
L.tendencies list. This steps will be repeated for each pj .
Thus the complexity of this function is O(m) such that m is
the number of PM.
The next step is a sequential search of maximum tendency
value T.max within L.tendencies list. The size of this list is
m which corresponds to PMs number. Each item within the
list must be examined, thus we obtain m comparisons.
Therefore according the two above function (computation and
search) we obtain a complexity O(m).
These steps are executed, in parallel, by each vi . Hence, the
algorithm must be executed n times where n is the number of
VM. Accordingly, our algorithm’s complexity is analyzed as
in O(n × m).
This complexity is polynomial according to the number of VM



The number of active PM



The Packing Efficiency: The PE of any solution produced by a virtual machine consolidation algorithm is
given by [21]
nV M
(13)
nActiveS
The unused resources: UR is the percentage of unused
resources in active PM and is given by [21]
P1
P1
n SCV −
m RDV
RU =
× 100
(14)
P1
n SCV
PE =





The run time.

Fig. 1 shows two graphic representations of active PM number:
the first representation is based on the number of VM, while
4

Fig. 2: Representation of active PM number

Fig. 3: Packing efficiency (PE)

the second representation is based on the number of PM in the
system.

600 with an incremental of 500 and the computation curve
for case when the number of PMs increased from 800 to
2600 with an incremental of 200 as shown in Fig.4. It can
be observed from the two curves that the computation time of
the VMPRP increases very close to linearly when the problem
size increases, whether by increasing VM or PM number. Thus,
it can be concluded that the VMPRP is scalable.

Results in Fig.1 show that the proposed VMPRP solution
activates the least number of PM compared to other solutions.This is because VMPRP tries to maximize the use of
PM resources. In large size problems VMPRP used only 600
PMs to host 6000VMs. It can be seen from the results in Fig.2
that the VMPRP algorithm has reached the highest packing
efficiency value for twenty different scenarios.In all cases,
VMPRP solution ensures an efficiency value higher than 8.

VI.

During the simulation of the twenty tests, we collected
the amount of unused resources produced by each consolidation solution. The results are shown in Fig.3. The unused
resources are represented as a percentage of the active PM
total capacity. The proposed VMPRP solution represented
a waste of resources lower than 10% while the CloudSim
solution and random solution had a high waste of resources
reaching even 85%. In order to check the scalability of the
VMPRP algorithm, we drew the computation time curve for
the case when the number of VMs increased from 1500 to

C ONCLUSION AND F UTURE WORKS

In this paper, we presented proposed a VM placement
policy based on Swarm Intelligence guaranteeing maximum
and balanced resources utilization and thus minimizing both
power consumption and resources wastage. We presented performance evaluation by comparing the proposed policy with
other two policies. The experimental results showed that our
VMPRP algorithm generated always significantly better quality
of solutions than other algorithms. As future work, we plan
to extend the proposed policy to take into account the extra
energy consumption in the communication between VM.
5

Fig. 4: The unused resources

Fig. 5: Computation time

R EFERENCES
[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

B. Xu, Z. Peng, F. Xiao, A. Gates and J. Yu, ”Dynamic deployment of
virtual machines in cloud computing using multi-objective optimization”,
Soft Comput, vol. 19, no. 8, pp. 2265-2273, 2014.

A. Beloglazov, J. Abawajy and R. Buyya, ”Energy-aware resource
allocation heuristics for efficient management of data centers for Cloud
[9] C. Joseph, K. Chandrasekaran and R. Cyriac, ”A Novel Family Genetic
computing”, Future Generation Computer Systems, vol. 28, no. 5, pp.
Approach for Virtual Machine Allocation”, Procedia Computer Science,
755-768, 2012.
vol. 46, pp. 558-565, 2015.
A. Beloglazov, and R. Buyya, ”Energy-Efficient Management of Virtual
[10] D. Wilcox, A. McNabb, and K. Seppi, ”Solving virtual machine packing
Machines in Data Centers for Cloud Computing”, the unisversity of
with a Reordering Grouping Genetic Algorithm”, in IEEE Congress on
Melbourne, Australia, 2013.
Evolutionary Computation (CE, New Orleans, LA, 2011, pp. 362-369.
A. Beloglazov, and R. Buyya, ”Adaptive threshold-based approach for
[11]
E. Arzuaga, and D. Kaeli, ”Quantifying load imbalance on virtualized
energy-efficient consolidation of virtual machines in cloud data centers”,
enterprise servers”, in 10 Proceedings of the first joint WOSP/SIPEW
in 10 Proceedings of the 8th International Workshop on Middleware for
international conference on Performance engineering, . New York, NY,
Grids, Clouds and e-Science., New York, NY, USA, 2010.
USA, 2010, pp. 235-242.
A. Vigliotti and D. Batista, ”Energy-Efficient Virtual Machines Place[12]
E. Bonabeau, G. Theraulaz and J. Deneubourg, ”Quantitative Study of
ment”, in Brazilian Symposium on Computer Networks and Distributed
the
Fixed Threshold Model for the Regulation of Division of Labour
Systems (SBRC), Florianopolis, 2014, pp. 1 - 8.
in Insect Societies”, Proceedings of the Royal Society B: Biological
A. Esnault, E. Feller and c. Morin, ”Energy-aware distributed ant
Sciences, vol. 263, no. 1376, pp. 1565-1569, 1996.
colony based virtual machine consolidation in iaas clouds bibliographic”,
[13]
E. Feller, C. Morin and A. Esnault, ”A case for fully decentralized
InformaticsMathematics, pp. 1-13, 2012.
dynamic VM consolidation in clouds”, in IEEE 4th International ConferA.
Farthouat,
”Le
cloud
computing,
un
nuage
pas
ence on Cloud Computing Technology and Science (CloudCom), Taipei,
si
vert”,
Novethic.fr,
2016.
[Online].
Available:
http:
2012, pp. 26 - 33.
//www.novethic.fr/lapres-petrole/energies-fossiles/isr-rse/
[14]
E. Feller, L. Rilling and C. Morin, ”Energy-Aware Ant Colony Based
le-cloud-computing-un-nuage-pas-si-vert-132429.html.
[Accessed:
Workload Placement in Clouds.”, in 12th IEEE/ACM International Con29- Mar- 2015].
ference on Grid Computing (GRID), Lyon, 2011, pp. 26-33.
A. Garric, ”Les centres de donnes informatiques avalent des
[15]
G. Lovasz, F. Niedermeier and H. de Meer, ”Performance tradeoffs of
quantits croissantes d’energie”, MPlanete, 2013. [Online].
energy-aware virtual machine consolidation”, Cluster Computing, vol.
Available:
http://www.lemonde.fr/planete/article/2013/07/01/
les-centres-de-donnees-informatiques-gros-consommateurs-d-energie34397683244.16, no. 3, pp. 481-496, 2012.
html. [Accessed: 01- Apr- 2015].
[16] H. Ahmed and J. Glasgow, ”Swarm Intelligence:Concepts, Models and

6

[17]

[18]

[19]

[20]

[21]

[22]
[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]
[34]

[35]

Applications”, Queens University. Kingston, Ontario, Canada, 2012.
H. Van, F. Tran and J. Menaud, ”Performance and Power Management
for Cloud Infrastructures”, in 3rd International Conference on Cloud
Computing (CLOUD), Miami, 2010, pp. 329- 336.
J. Xu and J. Fortes, ”Multi-Objective Virtual Machine Placement in
Virtualized Data Center Environments”, in IEEE/ACM Intl Conference
on and Intl Conference on Cyber, Physical and Social Computing (CPSCom)Green Computing and Communications (GreenCom), Hangzhou:,
2010, pp. 79-188.
M. Barati and S. Sharifian, ”A hybrid heuristic-based tuned support
vector regression model for cloud load prediction”, J Supercomput, vol.
71, no. 11, pp. 4235-4259, 2015.
M. Dorigo, E. Bonabeau and G. Theraulaz, ”Ant algorithms and
stigmergy”, Future Generation Computer Systems, vol. 16, no. 8, pp.
851-871, 2000.
M. Ferdaus, M. Murshed, R. Calheiros and R. Buyya, ”Virtual Machine
Consolidation in Cloud Data Centers Using ACO Metaheuristic”, in
Euro-Par 2014 Parallel Processing, 1st ed., M. Ferdaus, M. Murshed,
R. Calheiros and R. Buyya, Ed. Springer International Publishing, 2014,
pp. 306-317.
M. Kovanica, Swarm Algorithms: Simulation and Generation.. Case
Western Reserve University, 2006, p. 252.
M. Mishra and A. Sahoo, ”On Theory of VM Placement: Anomalies
in Existing Methodologies and Their Mitigation Using a Novel Vector
Based Approach”, in IEEE International Conference on Cloud Computing (CLOUD), Washington, DC, 2011, pp. 275 - 282.
M. Tang and S. Pan, ”A Hybrid Genetic Algorithm for the EnergyEfficient Virtual Machine Placement Problem in Data Centers”, Neural
Process Lett, vol. 41, no. 2, pp. 211-221, 2014.
M. Wang, X. Meng and L. Zhang, ”Consolidating virtual machines
with dynamic bandwidth demand in data centers”, in Proceedings IEEE
INFOCOM, Shanghai:, 2011, pp. 71-75.
N. Gondhi and P. Kailu, ”Prediction Based Energy Efficient Virtual
Machine Consolidation in Cloud Computing”, in Second International
Conference on Advances in Computing and Communication Engineering
(ICACCE), Dehradun:, 2015, pp. 437-441.
R. Calheiros, R. Ranjan, A. Beloglazov, C. De Rose and R. Buyya,
”CloudSim: a toolkit for modeling and simulation of cloud computing environments and evaluation of resource provisioning algorithms”, Softw:
Pract. Exper., vol. 41, no. 1, pp. 23-50, 2010.
R. Panigrahy, K. Talwar, L. Uyeda and U. wieder, Research at Microsoft, ”Heuristics for Vector Bin Packing”, Research at Microsoft,
2011.
S. Agrawal, S. Bose and S. Sundarrajan, ”Grouping genetic algorithm
for solving the server consolidation problem with conflicts”, in the first
ACM/SIGEVO Summit on Genetic and Evolutionary Computation, New
York, NY, USA, 2009, pp. 1-8.
S. Angeles, ”Virtualization vs. Cloud Computing: What’s the
Difference?”, Business News Daily, 2014. [Online]. Available:
http://www.businessnewsdaily.com/5791-virtualization-vs-cloudcomputing.html. [Accessed: 29- Mar- 2015].
S. Filiposka, A. Mishev and C. Juiz, ”Balancing Performances in Online
VM Placement”, in ICT Innovations 2015, 1st ed., S. Filiposka, A.
Mishev and C. Juiz, Ed. Springer International Publishing, 2015, pp.
153-162.
T. Wood, G. Tarasuk-Levin, P. Shenoy, P. Desnoyers, E. Cecchet and M.
Corner, ”Memory buddies”, ACM SIGOPS Operating Systems Review,
vol. 43, no. 3, p. 27, 2009.
W. Vogels, ”Beyond server consolidation”, Queue, vol. 6, no. 1, p. 20,
2008.
X. Zhang, Q. Yue and Z. He, ”Dynamic Energy-Efficient Virtual
Machine Placement Optimization for Virtualized Clouds”, in Proceedings
of the 2013 International Conference on Electrical and Information
Technologies for Rail Transportation (EITRT2013), 2nd ed., X. Zhang,
Q. Yue and Z. He, Ed. Springer-Verlag Berlin Heidelberg., 2014, pp.
439-448.
X. Li, Z. Qian, R. Chi, B. Zhang and S. Lu, ”Balancing Resource
Utilization for Continuous Virtual Machine Requests in Clouds”, in Sixth
International Conference on Innovative Mobile and Internet Services in
Ubiquitous Computing (IMIS), Palermo, 2012, pp. 266-273.

[36]

[37]

[38]

[39]

[40]

7

X. Li, Z. Qian, S. Lu and J. Wu, ”Energy efficient virtual machine
placement algorithm with balanced and improved resource utilization in
a data center”, Mathematical and Computer Modelling, vol. 58, no. 5-6,
pp. 1222-1235, 2013.
Y. Dong, G. Xu and X. Fu, ”A Distributed Parallel Genetic Algorithm
of Placement Strategy for Virtual Machines Deployment on Cloud
Platform”, The Scientific World Journal, vol. 2014, pp. 1-12, 2014.
Y. Gao, H. Guan, Z. Qi, Y. Hou and L. Liu, ”A multi-objective
ant colony system algorithm for virtual machine placement in cloud
computing”, Journal of Computer and System Sciences, vol. 79, no. 8,
pp. 1230-1242, 2013.
Z. Cao and S. Dong, ”An energy-aware heuristic framework for virtual
machine consolidation in Cloud computing”, J Supercomput, vol. 69, no.
1, pp. 429-451, 2014.
Z. Gong, X. Gu and J. Wilkes, ”PRESS: PRedictive Elastic Re- Source
Scaling for cloud systems.”, in International Conference on Network and
Service Management (CNSM), Niagara Falls, ON DOI:, 2010, pp. 9 16.



Documents similaires


vm placement algorithm
icmitieee2017
art 3a10 1007 2fs11554 009 0137 x
ijcai07 259
uk cloud hosting providers
10 1 1 95 9702


Sur le même sujet..