Drops division and replication of data in cloud for optimal performance and security

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation
information: DOI 10.1109/TCC.2015.2400460, IEEE Transactions on Cloud Computing
IEEE TRANSACTIONS ON CLOUD COMPUTING 1
DROPS: Division and Replication of Data in
Cloud for Optimal Performance and Security
Mazhar Ali, Student Member, IEEE, Kashif Bilal, Student Member, IEEE, Samee U. Khan, Senior
Member, IEEE, Bharadwaj Veeravalli, Senior Member, IEEE, Keqin Li, Senior Member, IEEE, and
Albert Y. Zomaya, Fellow, IEEE
Abstract—Outsourcing data to a third-party administrative control, as is done in cloud computing, gives rise to security concerns.
The data compromise may occur due to attacks by other users and nodes within the cloud. Therefore, high security measures are
required to protect data within the cloud. However, the employed security strategy must also take into account the optimization
of the data retrieval time. In this paper, we propose Division and Replication of Data in the Cloud for Optimal Performance and
Security (DROPS) that collectively approaches the security and performance issues. In the DROPS methodology, we divide a
file into fragments, and replicate the fragmented data over the cloud nodes. Each of the nodes stores only a single fragment
of a particular data file that ensures that even in case of a successful attack, no meaningful information is revealed to the
attacker. Moreover, the nodes storing the fragments, are separated with certain distance by means of graph T-coloring to prohibit
an attacker of guessing the locations of the fragments. Furthermore, the DROPS methodology does not rely on the traditional
cryptographic techniques for the data security; thereby relieving the system of computationally expensive methodologies. We
show that the probability to locate and compromise all of the nodes storing the fragments of a single file is extremely low. We
also compare the performance of the DROPS methodology with ten other schemes. The higher level of security with slight
performance overhead was observed.
Index Terms—Centrality, cloud security, fragmentation, replication, performance.
!
1 INTRODUCTION
THE cloud computing paradigm has reformed the
usage and management of the information tech-
nology infrastructure [7]. Cloud computing is char-
acterized by on-demand self-services, ubiquitous net-
work accesses, resource pooling, elasticity, and mea-
sured services [22, 8]. The aforementioned character-
istics of cloud computing make it a striking candidate
for businesses, organizations, and individual users
for adoption [25]. However, the benefits of low-cost,
negligible management (from a users perspective),
and greater flexibility come with increased security
concerns [7].
Security is one of the most crucial aspects among
those prohibiting the wide-spread adoption of cloud
computing [14, 19]. Cloud security issues may stem
M. Ali, K. Bilal, and S. U. Khan are with the Department
of Electrical and Computer Engineering, North Dakota
State University, Fargo, ND 58108-6050, USA. E-mail:
{mazhar.ali,kashif.bilal,samee.khan}@ndsu.edu
B. Veeravallii is with the Department of Electrical and Com-
puter Engineering, The National University of Singapore. E-mail:
elebv@nus.edu.sg
K. Li is with the Department of Computer Science, State University
of New York , New Paltz, NY 12561. E-mail: lik@ndsu.edu
A.Y. Zomaya is with the School of Information Technologies, The
University of Sydney, Sydney, NSW 2006, Australia. E-mail: al-
bert.zomaya@sydney.edu.au
due to the core technology′
s implementation (virtual
machine (VM) escape, session riding, etc.), cloud ser-
vice offerings (structured query language injection,
weak authentication schemes, etc.), and arising from
cloud characteristics (data recovery vulnerability, In-
ternet protocol vulnerability, etc.) [5]. For a cloud to be
secure, all of the participating entities must be secure.
In any given system with multiple units, the highest
level of the system′
s security is equal to the security
level of the weakest entity [12]. Therefore, in a cloud,
the security of the assets does not solely depend on
an individual’s security measures [5]. The neighboring
entities may provide an opportunity to an attacker to
bypass the users defenses.
The off-site data storage cloud utility requires users
to move data in cloud’s virtualized and shared envi-
ronment that may result in various security concerns.
Pooling and elasticity of a cloud, allows the physi-
cal resources to be shared among many users [22].
Moreover, the shared resources may be reassigned to
other users at some instance of time that may result
in data compromise through data recovery method-
ologies [22]. Furthermore, a multi-tenant virtualized
environment may result in a VM to escape the bounds
of virtual machine monitor (VMM). The escaped VM
can interfere with other VMs to have access to unau-
thorized data [9]. Similarly, cross-tenant virtualized
network access may also compromise data privacy
and integrity. Improper media sanitization can also
leak customer′
s private data [5].
For More Details Contact G.Venkat Rao
PVR TECHNOLOGIES 8143271457

Fig. 1: The DROPS methodology
The data outsourced to a public cloud must be
secured. Unauthorized data access by other users and
processes (whether accidental or deliberate) must be
prevented [14]. As discussed above, any weak entity
can put the whole cloud at risk. In such a scenario,
the security mechanism must substantially increase
an attacker’s effort to retrieve a reasonable amount
of data even after a successful intrusion in the cloud.
Moreover, the probable amount of loss (as a result of
data leakage) must also be minimized.
A cloud must ensure throughput, reliability, and
security [15]. A key factor determining the throughput
of a cloud that stores data is the data retrieval time
[21]. In large-scale systems, the problems of data re-
liability, data availability, and response time are dealt
with data replication strategies [3]. However, placing
replicas data over a number of nodes increases the
attack surface for that particular data. For instance,
storing m replicas of a file in a cloud instead of one
replica increases the probability of a node holding file
to be chosen as attack victim, from 1
n
to m
n
, where n
is the total number of nodes.
From the above discussion, we can deduce that
both security and performance are critical for the
next generation large-scale systems, such as clouds.
Therefore, in this paper, we collectively approach
the issue of security and performance as a secure
data replication problem. We present Division and
Replication of Data in the Cloud for Optimal Perfor-
mance and Security (DROPS) that judicially fragments
user files into pieces and replicates them at strategic
locations within the cloud. The division of a file into
fragments is performed based on a given user criteria
such that the individual fragments do not contain any
meaningful information. Each of the cloud nodes (we
use the term node to represent computing, storage,
physical, and virtual machines) contains a distinct
fragment to increase the data security. A successful
attack on a single node must not reveal the loca-
tions of other fragments within the cloud. To keep
an attacker uncertain about the locations of the file
fragments and to further improve the security, we
select the nodes in a manner that they are not adjacent
and are at certain distance from each other. The node
separation is ensured by the means of the T-coloring
[6]. To improve data retrieval time, the nodes are se-
lected based on the centrality measures that ensure an
improved access time. To further improve the retrieval
time, we judicially replicate fragments over the nodes
that generate the highest read/write requests. The
selection of the nodes is performed in two phases.
In the first phase, the nodes are selected for the initial
placement of the fragments based on the centrality
measures. In the second phase, the nodes are selected
for replication. The working of the DROPS methodol-
ogy is shown as a high-level work flow in Fig. 1. We
implement ten heuristics based replication strategies
as comparative techniques to the DROPS methodol-
ogy. The implemented replication strategies are: (a)
A-star based searching technique for data replication
problem (DRPA-star), (b) weighted A-star (WA-star),
(c) A -star, (d) suboptimal A-star1 (SA1), (e) subop-
timal A-star2 (SA2), (f) suboptimal A-star3 (SA3), (g)
Local Min-Min, (h) Global Min-Min, (i) Greedy algo-
rithm, and (j) Genetic Replication Algorithm (GRA).
The aforesaid strategies are fine-grained replication
techniques that determine the number and locations
of the replicas for improved system performance. For
our studies, we use three Data Center Network (DCN)
architectures, namely: (a) Three tier, (b) Fat tree, and
(c) DCell. We use the aforesaid architectures because
they constitute the modern cloud infrastructures and
the DROPS methodology is proposed to work for the
cloud computing paradigm.
Our major contributions in this paper are as follows:
We develop a scheme for outsourced data that
takes into account both the security and per-
formance. The proposed scheme fragments and
replicates the data file over cloud nodes.
The proposed DROPS scheme ensures that even
in the case of a successful attack, no meaningful
information is revealed to the attacker.
We do not rely on traditional cryptographic tech-
niques for data security. The non-cryptographic
nature of the proposed scheme makes it faster to
perform the required operations (placement and
retrieval) on the data.
We ensure a controlled replication of the file frag-
ments, where each of the fragments is replicated
only once for the purpose of improved security.
The remainder of the paper is organized as follows.
Section 2 provides an overview of the related work in
the field. In Section 3, we present the preliminaries.
The DROPS methodology is introduced in Section 4.
Section 5 explains the experimental setup and results,
and Section 6 concludes the paper.

2 RELATED WORK
Juels et al. [10] presented a technique to ensure the
integrity, freshness, and availability of data in a cloud.
The data migration to the cloud is performed by the
Iris file system. A gateway application is designed
and employed in the organization that ensures the
integrity and freshness of the data using a Merkle
tree. The file blocks, MAC codes, and version numbers
are stored at various levels of the tree. The proposed
technique in [10] heavily depends on the user′
s em-
ployed scheme for data confidentiality. Moreover, the
probable amount of loss in case of data tempering as a
result of intrusion or access by other VMs cannot be
decreased. Our proposed strategy does not depend
on the traditional cryptographic techniques for data
security. Moreover, the DROPS methodology does
not store the whole file on a single node to avoid
compromise of all of the data in case of successful
attack on the node.
The authors in [11] approached the virtualized and
multi-tenancy related issues in the cloud storage by
utilizing the consolidated storage and native access
control. The Dike authorization architecture is pro-
posed that combines the native access control and
the tenant name space isolation. The proposed system
is designed and works for object based file systems.
However, the leakage of critical information in case of
improper sanitization and malicious VM is not han-
dled. The DROPS methodology handles the leakage
of critical information by fragmenting data file and
using multiple nodes to store a single file.
The use of a trusted third party for providing
security services in the cloud is advocated in [22]. The
authors used the public key infrastructure (PKI) to en-
hance the level of trust in the authentication, integrity,
and confidentiality of data and the communication
between the involved parties. The keys are generated
and managed by the certification authorities. At the
user level, the use of temper proof devices, such as
smart cards was proposed for the storage of the keys.
Similarly, Tang et. al. have utilized the public key
cryptography and trusted third party for providing
data security in cloud environments [20]. However,
the authors in [20] have not used the PKI infrastruc-
ture to reduce the overheads. The trusted third party
is responsible for the generation and management of
public/private keys. The trusted third party may be a
single server or multiple servers. The symmetric keys
are protected by combining the public key cryptogra-
phy and the (k, n) threshold secret sharing schemes.
Nevertheless, such schemes do not protect the data
files against tempering and loss due to issues arising
from virtualization and multi-tenancy.
A secure and optimal placement of data objects in a
distributed system is presented in [21]. An encryption
key is divided into n shares and distributed on differ-
ent sites within the network. The division of a key into
n shares is carried out through the (k, n) threshold
secret sharing scheme. The network is divided into
clusters. The number of replicas and their placement
is determined through heuristics. A primary site is
selected in each of the clusters that allocates the repli-
cas within the cluster. The scheme presented in [21]
combines the replication problem with security and
access time improvement. Nevertheless, the scheme
focuses only on the security of the encryption key.
The data files are not fragmented and are handled as
a single file. The DROPS methodology, on the other
hand, fragments the file and store the fragments on
multiple nodes. Moreover, the DROPS methodology
focuses on the security of the data within the cloud
computing domain that is not considered in [21].
3 PRELIMINARIES
Before we go into the details of the DROPS methodol-
ogy, we introduce the related concepts in the follow-
ing for the ease of the readers.
3.1 Data Fragmentation
The security of a large-scale system, such as cloud de-
pends on the security of the system as a whole and the
security of individual nodes. A successful intrusion
into a single node may have severe consequences, not
only for data and applications on the victim node, but
also for the other nodes. The data on the victim node
may be revealed fully because of the presence of the
whole file [17]. A successful intrusion may be a result
of some software or administrative vulnerability [17].
In case of homogenous systems, the same flaw can
be utilized to target other nodes within the system.
The success of an attack on the subsequent nodes
will require less effort as compared to the effort on
the first node. Comparatively, more effort is required
for heterogeneous systems. However, compromising
a single file will require the effort to penetrate only
a single node. The amount of compromised data can
be reduced by making fragments of a data file and
storing them on separate nodes [17, 21]. A successful
intrusion on a single or few nodes will only provide
access to a portion of data that might not be of
any significance. Moreover, if an attacker is uncertain
about the locations of the fragments, the probability
of finding fragments on all of the nodes is very low.
Let us consider a cloud with M nodes and a file
with z number of fragments. Let s be the number of
successful intrusions on distinct nodes, such that s>z.
The probability that s number of victim nodes contain
all of the z sites storing the file fragments (represented
by P(s,z)) is given as:
P(s,z) =
(
s
z
)(
M − s
s − z
)
(
M
s
)
. (1)

If M = 30, s = 10, and z = 7, then P(10,7) = 0.0046.
However, if we choose M = 50, s = 20, and z = 15,
then P(20,15) = 0.000046. With the increase in M, the
probability of a state reduces further. Therefore, we
can say that the greater the value of M, the less prob-
able that an attacker will obtain the data file. In cloud
systems with thousands of nodes, the probability for
an attacker to obtain a considerable amount of data,
reduces significantly. However, placing each fragment
once in the system will increase the data retrieval
time. To improve the data retrieval time, fragments
can be replicated in a manner that reduces retrieval
time to an extent that does not increase the aforesaid
probability.
3.2 Centrality
The centrality of a node in a graph provides the
measure of the relative importance of a node in the
network. The objective of improved retrieval time
in replication makes the centrality measures more
important. There are various centrality measures; for
instance, closeness centrality, degree centrality, be-
tweenness centrality, eccentricity centrality, and eigen-
vector centrality. We only elaborate on the closeness,
betweenness, and eccentricity centralities because we
are using the aforesaid three centralities in this work.
For the remainder of the centralities, we encourage
the readers to review [24].
3.2.1 Betweenness Centrality
The betweenness centrality of a node n is the number
of the shortest paths, between other nodes, passing
through n [24]. Formally, the betweenness centrality
of any node v in a network is given as:
Cb(v) = ∑
a≠v≠b
δab(v)
δab
, (2)
where δab is the total number of shortest paths be-
tween a and b, and δab(v) is the number of shortest
paths between a and b passing through v. The variable
Cb(v) denotes the betweenness centrality for node v.
3.2.2 Closeness Centrality
A node is said to be closer with respect to all of
the other nodes within a network, if the sum of the
distances from all of the other nodes is lower than
the sum of the distances of other candidate nodes
from all of the other nodes [24]. The lower the sum
of distances from the other nodes, the more central is
the node. Formally, the closeness centrality of a node
v in a network is defined as:
Cc(v) =
N − 1
∑
a≠v
d(v,a)
, (3)
where N is total number of nodes in a network and
d(v,a) represents the distance between node v and
node a.
TABLE 1: Notations and their meanings
Symbols Meanings
M Total number of nodes in the cloud
N Total number of file fragments to be placed
Ok k-th fragment of file
ok Size of Ok
Si i-th node
si Size of Si
ceni Centrality measure for Si
colSi Color assigned to Si
T A set containing distances by which assignment of
fragments must be separated
ri
k Number of reads for Ok from Si
Ri
k Aggregate read cost of ri
k
wi
k Number of writes for Ok from Si
Wi
k Aggregate write cost of wi
k
NNi
k Nearest neighbor of Si holding Ok
c(i,j) Communication cost between Si and Sj
Pk Primary node for Ok
Rk Replication schema of Ok
RT Replication time
3.2.3 Eccentricity
The eccentricity of a node n is the maximum distance
to any node from a node n [24]. A node is more central
in the network, if it is less eccentric. Formally, the
eccentricity can be given as:
E(va) = maxbd(va,vb), (4)
where d(va,vb) represents the distance between node
va and node vb. It may be noted that in our evaluation
of the strategies the centrality measures introduced
above seem very meaningful and relevant than using
simple hop-count kind of metrics.
3.3 T-coloring
Suppose we have a graph G = (V,E) and a set T
containing non-negative integers including 0. The T-
coloring is a mapping function f from the vertices of V
to the set of non-negative integers, such that f(x)- f(y)
∉ T, where (x,y) ∈ E. The mapping function f assigns
a color to a vertex. In simple words, the distance
between the colors of the adjacent vertices must not
belong to T. Formulated by Hale [6], the T-coloring
problem for channel assignment assigns channels to
the nodes, such that the channels are separated by a
distance to avoid interference.
4 DROPS
4.1 System Model
Consider a cloud that consists of M nodes, each with
its own storage capacity. Let Si
represents the name
of i-th node and si denotes total storage capacity of
Si
. The communication time between Si
and Sj
is
the total time of all of the links within a selected path

from Si
to Sj
represented by c(i, j). We consider N
number of file fragments such that Ok denotes k-th
fragment of a file while ok represents the size of k-th
fragment. Let the total read and write requests from Si
for Ok be represented by ri
k and wi
k, respectively. Let
Pk denote the primary node that stores the primary
copy of Ok. The replication scheme for Ok denoted by
Rk is also stored at Pk. Moreover, every Si
contains
a two-field record, storing Pk for Ok and NNi
k that
represents the nearest node storing Ok. Whenever
there is an update in Ok, the updated version is
sent to Pk that broadcasts the updated version to all
of the nodes in Rk. Let b(i,j) and t(i,j) be the total
bandwidth of the link and traffic between sites Si
and Sj
, respectively . The centrality measure for Si
is represented by ceni. Let colSi store the value of
assigned color to Si
. The colSi can have one out of two
values, namely: open color and close color. The value
open color represents that the node is available for
storing the file fragment. The value close color shows
that the node cannot store the file fragment. Let T be
a set of integers starting from zero and ending on a
prespecified number. If the selected number is three,
then T = {0,1,2,3}. The set T is used to restrict the
node selection to those nodes that are at hop-distances
not belonging to T. For the ease of reading, the most
commonly used notations are listed in Table 1.
Our aim is to minimize the overall total network
transfer time or replication time (RT) or also termed
as replication cost (RC). The RT is composed of two
factors: (a) time due to read requests and (b) time due
to write requests. The total read time of Ok by Si
from
NNi
k is denoted by Ri
k and is given by:
Ri
k = ri
kokc(i,NNi
k). (5)
The total time due to the writing of Ok by Si
ad-
dressed to the Pk is represented as Wi
k and is given:
Wi
k = wi
kok(c(i,Pk) + ∑
(j∈Rk),j≠i
c(Pk,j)). (6)
The overall RT is represented by:
RT =
M
∑
i=1
N
∑
k=1
(Ri
k + Wi
k) (7)
The storage capacity constraint states that a file frag-
ment can only be assigned to a node, if storage
capacity of the node is greater or equal to the size
of fragment. The bandwidth constraint states that
b(i,j) ≥ t(i,j)∀i, ∀j. The DROPS methodology as-
signs the file fragments to the nodes in a cloud that
minimizes the RT, subject to capacity and bandwidth
constraints.
4.2 DROPS
In a cloud environment, a file in its totality, stored
at a node leads to a single point of failure [17]. A
successful attack on a node might put the data con-
fidentiality or integrity, or both at risk. The aforesaid
scenario can occur both in the case of intrusion or
accidental errors. In such systems, performance in
terms of retrieval time can be enhanced by employing
replication strategies. However, replication increases
the number of file copies within the cloud. Thereby,
increasing the probability of the node holding the file
to be a victim of attack as discussed in Section 1.
Security and replication are essential for a large-scale
system, such as cloud, as both are utilized to provide
services to the end user. Security and replication must
be balanced such that one service must not lower the
service level of the other.
In the DROPS methodology, we propose not to
store the entire file at a single node. The DROPS
methodology fragments the file and makes use of the
cloud for replication. The fragments are distributed
such that no node in a cloud holds more than a
single fragment, so that even a successful attack on
the node leaks no significant information. The DROPS
methodology uses controlled replication where each
of the fragments is replicated only once in the cloud to
improve the security. Although, the controlled repli-
cation does not improve the retrieval time to the level
of full-scale replication, it significantly improves the
security.
In the DROPS methodology, user sends the data file
to cloud. The cloud manager system (a user facing
server in the cloud that entertains user’s requests)
upon receiving the file performs: (a) fragmentation,
(b) first cycle of nodes selection and stores one frag-
ment over each of the selected node, and (c) second
cycle of nodes selection for fragments replication.
The cloud manager keeps record of the fragment
placement and is assumed to be a secure entity.
The fragmentation threshold of the data file is spec-
ified to be generated by the file owner. The file owner
can specify the fragmentation threshold in terms of
either percentage or the number and size of different
fragments. The percentage fragmentation threshold,
for instance, can dictate that each fragment will be
of 5% size of the total size of the file. Alternatively,
the owner may generate a separate file containing
information about the fragment number and size, for
instance, fragment 1 of size 5,000 Bytes, fragment 2
of size 8,749 Bytes. We argue that the owner of the
file is the best candidate to generate fragmentation
threshold. The owner can best split the file such that
each fragment does not contain significant amount
of information as the owner is cognizant of all the
facts pertaining to the data. The default percentage
fragmentation threshold can be made a part of the
Service Level Agreement (SLA), if the user does not
specify the fragmentation threshold while uploading
the data file. We primarily focus the storage system
security in this work with an assumption that the
communication channel between user and the cloud

is secure.
Algorithm 1 Algorithm for fragment placement
Inputs and initializations:
O = {O1,O2,...,ON }
o = {sizeof(O1),sizeof(O2),....,sizeof(ON )}
col = {open color,close color}
cen = {cen1,cen2,...,cenM }
col ← open color∀ i
cen ← ceni∀ i
Compute:
for each Ok ∈ O do
select Si
Si
← indexof(max(ceni))
if colSi = open color and si >= ok then
Si
← Ok
si ← si − ok
colSi ← close color
Si
’← distance(Si
,T) /*returns all nodes at
distance T from Si and stores in temporary set Si’*/
colSi′ ← close color
end if
end for
Once the file is split into fragments, the DROPS
methodology selects the cloud nodes for fragment
placement. The selection is made by keeping an equal
focus on both security and performance in terms of
the access time. We choose the nodes that are most
central to the cloud network to provide better access
time. For the aforesaid purpose, the DROPS method-
ology uses the concept of centrality to reduce access
time. The centralities determine how central a node is
based on different measures as discussed in Section
3.2. We implement DROPS with three centrality mea-
sures, namely: (a) betweenness, (b) closeness, and (c)
eccentricity centrality. However, if all of the fragments
are placed on the nodes based on the descending
order of centrality, then there is a possibility that
adjacent nodes are selected for fragment placement.
Such a placement can provide clues to an attacker as
to where other fragments might be present, reducing
the security level of the data. To deal with the security
aspects of placing fragments, we use the concept of
T-coloring that was originally used for the channel
assignment problem [6]. We generate a non-negative
random number and build the set T starting from
zero to the generated random number. The set T is
used to restrict the node selection to those nodes that
are at hop-distances not belonging to T. For the said
purpose, we assign colors to the nodes, such that,
initially, all of the nodes are given the open color.
Once a fragment is placed on the node, all of the nodes
within the neighborhood at a distance belonging to
T are assigned close color. In the aforesaid process,
we lose some of the central nodes that may increase
the retrieval time but we achieve a higher security
level. If somehow the intruder compromises a node
and obtains a fragment, then the location of the
other fragments cannot be determined. The attacker
can only keep on guessing the location of the other
fragments. However, as stated previously in Section
3.1, the probability of a successful coordinated attack
is extremely minute. The process is repeated until all
of the fragments are placed at the nodes. Algorithm
1 represents the fragment placement methodology.
In addition to placing the fragments on the central
nodes, we also perform a controlled replication to
increase the data availability, reliability, and improve
data retrieval time. We place the fragment on the
node that provides the decreased access cost with an
objective to improve retrieval time for accessing the
fragments for reconstruction of original file. While
replicating the fragment, the separation of fragments
as explained in the placement technique through T-
coloring, is also taken care off. In case of a large
number of fragments or small number of nodes, it
is also possible that some of the fragments are left
without being replicated because of the T-coloring.
As discussed previously, T-coloring prohibits to store
the fragment in neighborhood of a node storing a
fragment, resulting in the elimination of a number of
nodes to be used for storage. In such a case, only for
the remaining fragments, the nodes that are not hold-
ing any fragment are selected for storage randomly.
The replication strategy is presented in Algorithm 2.
To handle the download request from user, the
cloud manager collects all the fragments from the
nodes and re-assemble them into a single file. After-
wards, the file is sent to the user.
Algorithm 2 Algorithm for fragment′
s replication
for each Ok in O do
select Si
that has max(Ri
k + Wi
k)
if colSi = open color and si >= ok then
Si
← Ok
si ← si − ok
colSi ← close color
Si
’← distance(Si
,T) /*returns all nodes at
distance T from Si and stores in temporary set Si’*/
colSi′ ← close color
end if
end for
4.3 Discussion
A node is compromised with a certain amount of an
attacker’s effort. If the compromised node stores the
data file in totality, then a successful attack on a cloud
node will result in compromise of an entire data file.
However, if the node stores only a fragment of a file,
then a successful attack reveals only a fragment of
a data file. Because the DROPS methodology stores
fragments of data files over distinct nodes, an attacker
has to compromise a large number of nodes to obtain
meaningful information. The number of compromised

TABLE 2: Various attacks handled by DROPS methodology
Attack Description
Data Recovery Rollback of VM to some previous state. May expose previously stored data.
Cross VM attack Malicious VM attacking co-resident VM that may lead to data breach.
Improper media sanitization Data exposure due to improper sanitization of storage devices.
E-discovery Data exposure of one user due to seized hardware for investigations related to some other users.
VM escape A malicious user or VM escapes from the control of VMM. Provides access to storage and compute devices.
VM rollback Rollback of VM to some previous state. May expose previously stored data.
nodes must be greater than n because each of the com-
promised node may not give fragment in the DROPS
methodology as the nodes are separated based on
the T-coloring. Alternatively, an attacker has to com-
promise the authentication system of cloud [23]. The
effort required by an attacker to compromise a node
(in systems dealing with fragments/shares of data) is
given in [23] as:
EConf = min(EAuth,n × EBreakIn), (8)
where EConf is the effort required to compromise
the confidentiality, EAuth is the effort required to
compromise authentication, and EBreakIn is the effort
required to compromise a single node. Our focus in
this paper is on the security of the data in the cloud
and we do not take into account the security of the
authentication system. Therefore, we can say that to
obtain n fragments, the effort of an attacker increases
by a factor of n. Moreover, in case of the DROPS
methodology, the attacker must correctly guess the
nodes storing fragments of file. Therefore, in the worst
case scenario, the set of nodes compromised by the
attacker will contain all of the nodes storing the file
fragments. From Equation (1), we observe that the
probability of the worst case to be successful is very
low. The probability that some of the machines (av-
erage case) storing the file fragments will be selected
is high in comparison to the worst case probability.
However, the compromised fragments will not be
enough to reconstruct the whole data. In terms of
the probability, the worst, average, and best cases are
dependent on the number of nodes storing fragments
that are selected for an attack. Therefore, all of the
three cases are captured by Equation (1).
Besides the general attack of a compromised node,
the DROPS methodology can handle the attacks in
which attacker gets hold of user data by avoiding or
disrupting security defenses. Table 2 presents some of
the attacks that are handled by the DROPS methodol-
ogy. The presented attacks are cloud specific that stem
from clouds core technologies. Table 2 also provides a
brief description of the attacks. It is noteworthy that
even in case of successful attacks (that are mentioned),
the DROPS methodology ensures that the attacker
gets only a fragment of file as DROPS methodology
stores only a single fragment on the node. Moreover,
the successful attack has to be on the node that stores
the fragment.
5 EXPERIMENTAL SETUP AND RESULTS
The communicational backbone of cloud computing is
the Data Center Network (DCN) [2]. In this paper, we
use three DCN architectures namely: (a) Three tier, (b)
Fat tree, and (c) DCell [1]. The Three tier is the legacy
DCN architecture. However, to meet the growing de-
mands of the cloud computing, the Fat tree and Dcell
architectures were proposed [2]. Therefore, we use
the aforementioned three architectures to evaluate the
performance of our scheme on legacy as well as state
of the art architectures. The Fat tree and Three tier
architectures are switch-centric networks. The nodes
are connected with the access layer switches. Multiple
access layer switches are connected using aggregate
layer switches. Core layers switches interconnect the
aggregate layer switches.. The Dcell is a server centric
network architecture that uses servers in addition
to switches to perform the communication process
within the network [1]. A server in the Dcell architec-
ture is connected to other servers and a switch. The
lower level dcells recursively build the higher level
dcells. The dcells at the same level are fully connected.
For details about the aforesaid architectures and their
performance analysis, the readers are encouraged to
read [1] and [2].
5.1 Comparative techniques
We compared the results of the DROPS methodol-
ogy with fine-grained replication strategies, namely:
(a) DRPA-star, (b) WA-star, (c) A -star, (d) SA1, (e)
SA2, (f) SA3, (g) Local Min-Min, (h) Global Min-
Min, (i) Greedy algorithm, and (j) Genetic Replication
Algorithm (GRA). The DRPA-star is a data replication
algorithm based on the A-star best-first search algo-
rithm. The DRPA-star starts from the null solution
that is called a root node. The communication cost
at each node n is computed as: cost(n) = g(n) + h(n),
where g(n) is the path cost for reaching n and h(n) is
called the heuristic cost and is the estimate of cost
from n to the goal node. The DRPA-star searches
all of the solutions of allocating a fragment to a
node. The solution that minimizes the cost within
the constraints is explored while others are discarded.
The selected solution is inserted into a list called

the OPEN list. The list is ordered in the ascending
order so that the solution with the minimum cost
is expanded first. The heuristic used by the DRPA-
star is given as h(n) = max(0,(mmk(n)g(n))), where
mmk(n) is the least cost replica allocation or the max-
min RC. Readers are encouraged to see the details
about DRPA-star in [13]. The WA-Star is a refinement
of the DRPA-star that implements a weighted func-
tion to evaluate the cost. The function is given as:
f(n) = f(n) + h(n) + (1 − (d(n)/D)h(n). The variable
d(n) represents the depth of the node n and D denotes
the expected depth of the goal node [13]. The A -star
is also a variation of the DRPA-star that uses two lists,
OPEN and FOCAL. The FOCAL list contains only
those nodes from the OPEN list that have f greater
than or equal to the lowest f by a factor of 1 + .
The node expansion is performed from the FOCAL list
instead of the OPEN list. Further details about WA-
Star and A -star can be found in [13]. The SA1 (sub-
optimal assignments), SA2, and SA3 are DRPA-star
based heuristics. In SA1, at level R or below, only the
best successors of node n having the least expansion
cost are selected. The SA2 selects the best successors
of node n only for the first time when it reaches
the depth level R. All other successors are discarded.
The SA3 works similar to the SA2, except that the
nodes are removed from OPEN list except the one
with the lowest cost. Readers are encouraged to read
[13] for further details about SA1, SA2, and SA3. The
LMM can be considered as a special case of the bin
packing algorithm. The LMM sorts the file fragments
based on the RC of the fragments to be stored at a
node. The LMM then assigns the fragments in the
ascending order. In case of a tie, the file fragment
with minimum size is selected for assignment (name
local Min-Min is derived from such a policy). The
GMM selects the file fragment with global minimum
of all the RC associated with a file fragment. In case
of a tie, the file fragment is selected at random. The
Greedy algorithm first iterates through all of the M
cloud nodes to find the best node for allocating a
file fragment. The node with the lowest replication
cost is selected. The second node for the fragment
is selected in the second iteration. However, in the
second iteration that node is selected that produces
the lowest RC in combination with node already
selected. The process is repeated for all of the file
fragments. Details of the greedy algorithm can be
found in [18]. The GRA consists of chromosomes rep-
resenting various schemes for storing file fragments
over cloud nodes. Every chromosome consists of M
genes, each representing a node. Every gene is a N
bit string. If the k-th file fragment is to be assigned
to Si
, then the k-th bit of i-th gene holds the value
of one. Genetic algorithms perform the operations of
selection, crossover, and mutation. The value for the
crossover rate (µc) was selected as 0.9, while for the
mutation rate (µm) the value was 0.01. The use of the
values for µc and µm is advocated in [16].The best
chromosome represents the solution. GRA utilizes mix
and match strategy to reach the solution. More details
about GRA can be obtained from [16].
5.2 Workload
The size of files were generated using a uniform dis-
tribution between 10Kb and 60 Kb. The primary nodes
were randomly selected for replication algorithms. For
the DROPS methodology, the Si′
s selected during the
first cycle of the nodes selection by Algorithm 1 were
considered as the primary nodes.
The capacity of a node was generated using a
uniform distribution between (1
2
CS)C and (3
2
CS)C,
where 0 ≤ C ≥ 1. For instance, for CS = 150 and
C = 0.6 the capacities of the nodes were uniformly
distributed between 45 and 135. The mean value of
g in the OPEN and FOCAL lists was selected as the
value of , for WA-star and A -star, respectively. The
value for level R was set to ⌊d
2
⌋, where d is the depth
of the search tree(number of fragments).
The read/write (R/W) ratio for the simulations
that used fixed value was selected to be 0.25 (The
R/W ratio reflecting 25% reads and 75% writes within
the cloud). The reason for choosing a high workload
(lower percentage of reads and higher percentage
of writes) was to evaluate the performance of the
techniques under extreme cases. The simulations that
studied the impact of change in the R/W ratio used
various workloads in terms of R/W ratios. The R/W
ratios selected were in the range of 0.10 to 0.90. The
selected range covered the effect of high, medium, and
low workloads with respect to the R/W ratio.
5.3 Results and Discussion
We compared the performance of the DROPS method-
ology with the algorithms discussed in Section 5.1.
The behavior of the algorithms was studied by: (a)
increasing the number of nodes in the system, (b)
increasing the number of objects keeping number
of nodes constant, (c) changing the nodes storage
capacity, and (d) varying the read/write ratio. The
aforesaid parameters are significant as they affect the
problem size and the performance of algorithms [13].
5.3.1 Impact of increase in number of cloud nodes
We studied the performance of the placement tech-
niques and the DROPS methodology by increasing the
number of nodes. The performance was studied for
the three discussed cloud architectures. The numbers
of nodes selected for the simulations were 100, 500,
1,024, 2,400, and 30,000. The number of nodes in
the Dcell architecture increases exponentially [2]. For
a Dcell architecture, with two nodes in the Dcell0,
the architecture consists of 2,400 nodes. However,
increasing a single node in the Dcell0, the total nodes
increases to 30, 000 [2]. The number of file fragments

20
30
40
50
60
70
80
100 500 1024 2400 30000
RCsavings(%)
No. of nodes
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(a)
10
20
30
40
50
60
70
80
100 500 1024 2400 30000
RCsavings(%)
No. of nodes
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(b)
Fig. 2: (a) RC versus number of nodes (Three tier) (b) RC versus number of nodes (Fat tier)
20
30
40
50
60
70
80
100 500 1024 2400 30000
RCsavings(%)
No. of nodes
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(a)
18
19
20
21
22
23
24
25
100 500 1024 2400 30000
RCsavings(%)
No. of nodes
DROPS-BC
DROPS-CC
DROPS-EC
(b)
Fig. 3: (a) RC versus number of nodes (Dcell) (b) RC versus number of nodes for DROPS variations with
maximum available capacity constraint (Three tier)
18
19
20
21
22
23
24
100 500 1024 2400 30000
RCsavings(%)
No. of nodes
DROPS-BC
DROPS-CC
DROPS-EC
(a)
18
20
22
24
26
28
30
100 500 1024 2400 30000
RCsavings(%)
No. of nodes
DROPS-BC
DROPS-CC
DROPS-EC
(b)
Fig. 4: RC versus number of nodes for DROPS variations with maximum available capacity constraints (a) Fat
tree (b) Dcell
was set to 50. For the first experiment we used
C = 0.2. Fig. 2 (a), Fig. 2 (b), and Fig. 3 (a) show
the results for the Three tier, Fat tree, and Dcell
architectures, respectively. The reduction in network
transfer time for a file is termed as RC. In the figures,
the BC stands for the betweenness centrality, the CC
stands for closeness centrality, and the EC stands for
eccentricity centrality. The interesting observation is
that although all of the algorithms showed similar
trend in performance within a specific architecture,
the performance of the algorithms was better in the
Dcell architecture as compared to three tier and fat
tree architectures. This is because the Dcell archi-
tecture exhibits better inter node connectivity and
robustness [2]. The DRPA-star gave best solutions as
compared to other techniques and registered consis-
tent performance with the increase in the number
of nodes. Similarly, WA-star, A -star, GRA, greedy,
and SA3 showed almost consistent performance with
various number of nodes. The performance of LMM
and GMM gradually increased with the increase in
number of nodes since the increase in the number of
nodes increased the number of bins. The SA1 and SA2
also showed almost constant performance in all of the
three architectures. However, it is important to note
that SA2 ended up with a decrease in performance

10
20
30
40
50
60
70
80
50 100 200 300 400 500
RCsavings(%)
No. of fragments
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(a)
20
30
40
50
60
70
80
50 100 200 300 400 500
RCsavings(%)
No. of fragments
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(b)
Fig. 5: (a) RC versus number of file fragments (Three tier) (b) RC versus number of file fragments (Fat tier)
as compared to the initial performance. This may be
due to the fact that SA2 only expands the node with
minimum cost when it reaches at certain depth for
the first time. Such a pruning for the first time, might
have purged nodes by providing better global access
time. The DROPS methodology, did not employ full-
scale replication. Every fragment is replicated only
once in the system. The smaller number of replicas of
any fragment and separation of nodes by T-coloring
decreased the probability of finding that fragment by
an attacker. Therefore, the increase in the security
level of the data is accompanied by the drop in
performance as compared to the comparative tech-
niques discussed in this paper. It is important to note
that the DROPS methodology was implemented using
three centrality measures namely: (a) betweenness, (b)
closeness, and (c) eccentricity. However, Fig. 2(a) and
Fig. 2(b) show only a single plot. Due to the inherent
structure of the Three tier and Fat tree architectures,
all of the nodes in the network are at the same
distance from each other or exist at the same level.
Therefore, the centrality measure is the same for all of
the nodes. This results in the selection of same node
for storing the file fragment. Consequently, the per-
formance showed the same value and all three lines
are on the same points. However, this is not the case
for the Dcell architecture. In the Dcell architecture,
nodes have different centrality measures resulting in
the selection of different nodes. It is noteworthy to
mention that in Fig 3(a), the eccentricity centrality
performs better as compared to the closeness and be-
tweenness centralities because the nodes with higher
eccentricity are located closer to all other nodes within
the network. To check the effect of closeness and
betweenness centralities, we modified the heuristic
presented in Algorithm 1. Instead of selecting the
node with criteria of only maximum centrality, we
selected the node with: (a) maximum centrality and
(b) maximum available storage capacity. The results
are presented in Fig. 3 (b), Fig. 4 (a), and Fig. 4 (b). It is
evident that the eccentricity centrality resulted in the
highest performance while the betweenness centrality
showed the lowest performance. The reason for this
is that nodes with higher eccentricity are closer to all
other nodes in the network that results in lower RC
value for accessing the fragments.
5.3.2 Impact of increase in number of file fragments
The increase in number of file fragments can strain
the storage capacity of the cloud that, in turn may
affect the selection of the nodes. To study the impact
on performance due to increase in number of file
fragments, we set the number of nodes to 30,000. The
numbers of file fragments selected were 50, 100, 200,
300, 400, and 500. The workload was generated with
C = 45% to observe the effect of increase number
of file fragments with fairly reasonable amount of
memory and to discern the performance of all the
algorithms. The results are shown in Fig. 5 (a), Fig. 5
(b), and Fig. 6 (a) for the Three tier, Fat tree, and Dcell
architectures, respectively. It can be observed from the
plots that the increase in the number of file fragments
reduced the performance of the algorithms, in general.
However, the greedy algorithm showed the most
improved performance. The LMM showed the highest
loss in performance that is little above 16%. The loss in
performance can be attributed to the storage capacity
constraints that prohibited the placements of some
fragments at nodes with optimal retrieval time. As
discussed earlier, the DROPS methodology produced
similar results in three tier and fat tree architectures.
However, from the Dcell architecture, it is clear that
the DROPS methodology with eccentricity centrality
maintains the supremacy on the other two centralities.
5.3.3 Impact of increase in storage capacity of nodes
Next, we studied the effect of change in the nodes
storage capacity. A change in storage capacity of the
nodes may affect the number of replicas on the node
due to storage capacity constraints. Intuitively, a lower
node storage capacity may result in the elimination
of some optimal nodes to be selected for replication
because of violation of storage capacity constraints.
The elimination of some nodes may degrade the per-
formance to some extent because a node giving lower
access time might be pruned due to non-availability

10
20
30
40
50
60
70
80
90
50 100 200 300 400 500
RCsavings(%)
No. of fragments
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(a)
10
20
30
40
50
60
70
80
20 25 30 35 40 45
RCsavings(%)
Node storage capacity
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(b)
Fig. 6: (a) RC versus number of file fragments (Dcell) (b) RC versus nodes storage capacity (Three tier)
10
20
30
40
50
60
70
80
20 25 30 35 40 45
RCsavings(%)
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(a)
10
20
30
40
50
60
70
80
20 25 30 35 40 45
RCsavings(%)
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(b)
Fig. 7: (a) RC versus nodes storage capacity (Fat tree) (b) RC versus nodes storage capacity (Dcell)
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90
RCsavings(%)
R/W ratio
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(a)
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90
RCsavings(%)
R/W ratio
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
(b)
Fig. 8: (a) RC versus R/W ratio (Three tree) (b) RC versus R/W ratio (Fat tree)
20
30
40
50
60
70
80
90
10 20 30 40 50 60 70 80 90
RCsavings(%)
R/W ratio
DRPA-star
LMM
WA-star
GMM
Aeps-star
SA1
SA2
SA3
Greedy
GRA
DROPS-BC
DROPS-CC
DROPS-EC
AƐ
Fig. 9: RC versus R/W ratio (Dcell)
of enough storage space to store the file fragment.
Higher node storage capacity allows full-scale repli-
cation of fragments, increasing the performance gain.
However, node capacity above certain level will not
change the performance significantly as replicating
the already replicated fragments will not produce con-
siderable performance increase. If the storage nodes
have enough capacity to store the allocated file frag-
ments, then a further increase in the storage capacity
of a node cannot cause the fragments to be stored
again. Moreover, the T-coloring allows only a single
replica to be stored on any node. Therefore, after a
certain point, the increase in storage capacity might
not affect the performance.
We increase the nodes storage capacity incremen-
tally from 20% to 40%. The results are shown in Fig.
6 (b), Fig. 7 (a), and Fig. 7 (b). It is observable from

0
15
30
45
60
75
90
500 1024 2400 30000
Failednodes(%)
No. of nodes
Three tier
Fat tree
Dcell
Fig. 10: Fault tolerance level of DROPS
the plots that initially, all of the algorithms showed
significant increase in performance with an increase
in the storage capacity. Afterwards, the marginal in-
crease in the performance reduces with the increase in
the storage capacity. The DRPA-star, greedy, WA-star,
and A -star showed nearly similar performance and
recorded higher performance. The DROPS methodol-
ogy did not show any considerable change in results
when compared to previously discussed experiments
(change in number of nodes and files). This is because
the DROPS methodology does not go for a full-scale
replication of file fragments rather they are replicated
only once and a single node only stores a single
fragment. Single time replication does not require
high storage capacity. Therefore, the change in nodes
storage capacity did not affect the performance of
DROPS to a notable extent.
5.3.4 Impact of increase in the read/write ratio
The change in R/W ratio affects the performance of
the discussed comparative techniques. An increase in
the number of reads would lead to a need of more
replicas of the fragments in the cloud. The increased
number of replicas decreases the communication cost
associated with the reading of fragments. However,
the increased number of writes demands that the
replicas be placed closer to the primary node. The
presence of replicas closer to the primary node results
in decreased RC associated with updating replicas.
The higher write ratios may increase the traffic on the
network for updating the replicas.
Fig. 8 (a), Fig. 8 (b), and Fig. 9 show the perfor-
mance of the comparative techniques and the DROPS
methodology under varying R/W ratios. It is ob-
served that all of the comparative techniques showed
an increase in the RC savings up to the R/W ratio of
0.50. The decrease in the number of writes caused the
reduction of cost associated with updating the replicas
of the fragments. However, all of the comparative
techniques showed some sort of decrease in RC saving
for R/W ratios above 0.50. This may be attributed to
the fact that an increase in the number of reads caused
more replicas of fragments resulting in increased cost
of updating the replicas. Therefore, the increased cost
of updating replicas underpins the advantage of de-
creased cost of reading with higher number of replicas
at R/W ratio above 0.50. It is also important to men-
tion that even at higher R/W ratio values the DRPA-
star, WA-star, A -star, and Greedy algorithms almost
maintained their initial RC saving values. The high
performance of the aforesaid algorithms is due to the
fact that these algorithms focus on the global RC value
while replicating the fragments. Therefore, the global
perception of these algorithms resulted in high perfor-
mance. Alternatively, LMM and GMM did not show
substantial performance due to their local RC view
while assigning a fragment to a node. The SA1, SA2,
and SA3 suffered due to their restricted search tree
that probably ignored some globally high performing
nodes during expansion. The DROPS methodology
maintained almost consistent performance as is ob-
servable from the plots. The reason for this is that the
DROPS methodology replicates the fragments only
once, so varying R/W ratios did not affect the results
considerably. However, the slight changes in the RC
value are observed. This might be due to the reason
that different nodes generate high cost for R/W of
fragments with different R/W ratio.
As discussed earlier, the comparative techniques
focus on the performance and try to reduce the RC
as much as possible. The DROPS methodology, on
the other hand, is proposed to collectively approach
the security and performance. To increase the security
level of the data, the DROPS methodology sacrifices
the performance to certain extent. Therefore, we see a
drop in the performance of the DROPS methodology
as compared to discussed comparative techniques.
However, the drop in performance is accompanied by
much needed increase in security level.
Moreover, it is noteworthy that the difference in
performance level of the DROPS methodology and
the comparative techniques is least with the reduced
storage capacity of the nodes (see Fig. 6 (b), Fig. 7
(a), and Fig. 7 (b)). The reduced storage capacity pro-
scribes the comparative techniques to place as many
replicas as required for the optimized performance. A
further reduction in the storage capacity will tend to
even lower the performance of the comparative tech-
niques. Therefore, we conclude that the difference in
performance level of the DROPS methodology and the
comparative techniques is least when the comparative
techniques reduce the extensiveness of replication for
any reason.
Due to the fact that the DROPS methodology re-
duces the number of replicas, we have also investi-
gates the fault tolerance of the DROPS methodology.
If two nodes storing the same file fragment fail, the
result will be incomplete or faulty file. We randomly
picked and failed the nodes to check that what per-
centage of failed nodes will result in loss of data or

TABLE 3: Average RC (%) savings for increase in number of nodes
Architec-
ture
DRPA LMM wa-star GMM A -star SA1 SA2 SA3 Greedy GRA DROPS-
BC
DROPS-
CC
DROPS-
EC
Three
tier
74.70 36.23 72.55 45.62 71.82 59.86 49.09 64.38 69.1 66.1 24.41 24.41 24.41
Fat
tree
76.76 38.95 75.22 45.77 73.33 60.89 52.67 68.33 71.64 70.54 23.28 23.28 23.28
Dcell 79.6 44.32 76.51 46.34 76.43 62.03 54.90 71.53 73.09 72.34 23.06 25.16 30.20
TABLE 4: Average RC (%) savings for increase in number of fragments
Architec-
ture
BC
DROPS-
CC
DROPS-
EC
Three
tier
74.63 40.08 69.69 48.67 68.82 60.29 49.65 62.18 71.25 64.44 23.93 23.93 23.93
Fat
tree
75.45 44.33 70.90 52.66 70.58 61.12 51.09 64.64 71.73 66.90 23.42 23.42 23.42
Dcell 76.08 45.90 72.49 52.78 72.33 62.12 50.02 64.66 70.92 69.50 23.17 25.35 28.17
TABLE 5: Average RC (%) savings for increase in storage capacity
Architec-
ture
BC
DROPS-
CC
DROPS-
EC
Three
tier
72.37 28.26 71.99 40.63 71.19 59.29 48.67 61.83 72.09 63.54 19.89 19.89 19.89
Fat
tree
69.19 28.34 70.73 41.99 66.20 60.28 51.29 61.83 69.33 62.16 21.60 21.60 21.60
Dcell 73.57 31.04 71.37 42.41 67.70 60.79 50.42 63.78 69.64 64.03 21.91 22.88 24.68
TABLE 6: Average RC (%) savings for increase in R/W ratio
Architec-
ture
BC
DROPS-
CC
DROPS-
EC
Three
tier
77.28 32.54 76.32 53.20 75.38 55.13 49.61 59.74 73.64 58.27 24.08 24.08 24.08
Fat
tree
76.29 31.47 74.81 52.08 73.37 53.33 49.35 57.87 71.61 57.47 23.68 23.68 23.68
Dcell 78.72 33.66 78.03 55.82 76.47 57.44 52.28 61.94 74.54 60.16 23.32 23.79 24.23
selection of two nodes storing same file fragment.
The numbers of nodes used in aforesaid experiment
were 500, 1,024, 2,400, and 30, 000. The number of file
fragments was set to 50. The results are shown in Fig.
10. As can be seen in Fig. 10, the increase in number of
nodes increases the fault tolerance level. The random
failure has generated a reasonable percentage for a
soundly decent number of nodes.
We report the average RC (%) savings in Table 3, Ta-
ble 4, Table 5, and Table 6. The averages are computed
over all of the RC (%) savings within a certain class of
experiments. Table 3 reveals the average results of all
of the experiments conducted to observe the impact of
increase in the number of nodes in the cloud for all of
the three discussed cloud architectures. Table 4 depicts
the average RC (%) savings for the increase in the
number of fragments. Table 5 and Table 6 describe the
average results for the increase the storage capacity
and R/W ratio, respectively. It is evident from the
average results that the Dcell architecture showed
better results due to its higher connectivity ratio.
6 CONCLUSIONS
We proposed the DROPS methodology, a cloud stor-
age security scheme that collectively deals with the
security and performance in terms of retrieval time.
The data file was fragmented and the fragments are
dispersed over multiple nodes. The nodes were sepa-
rated by means of T-coloring. The fragmentation and
dispersal ensured that no significant information was
obtainable by an adversary in case of a successful
attack. No node in the cloud, stored more than a single
fragment of the same file. The performance of the
DROPS methodology was compared with full-scale
replication techniques. The results of the simulations
revealed that the simultaneous focus on the security
and performance, resulted in increased security level
of data accompanied by a slight performance drop.
Currently with the DROPS methodology, a user has
to download the file, update the contents, and upload
it again. It is strategic to develop an automatic update
mechanism that can identify and update the required
fragments only. The aforesaid future work will save

the time and resources utilized in downloading, up-
dating, and uploading the file again. Moreover, the
implications of TCP incast over the DROPS method-
ology need to be studied that is relevant to distributed
data storage and access.
REFERENCES
[1] K. Bilal, S. U. Khan, L. Zhang, H. Li, K. Hayat, S. A. Madani,
N. Min-Allah, L. Wang, D. Chen, M. Iqbal, C. Z. Xu, and A. Y.
Zomaya, “Quantitative comparisons of the state of the art data
center architectures,” Concurrency and Computation: Practice and
Experience, Vol. 25, No. 12, 2013, pp. 1771-1783.
[2] K. Bilal, M. Manzano, S. U. Khan, E. Calle, K. Li, and A.
Zomaya, “On the characterization of the structural robustness
of data center networks,” IEEE Transactions on Cloud Computing,
Vol. 1, No. 1, 2013, pp. 64-77.
[3] D. Boru, D. Kliazovich, F. Granelli, P. Bouvry, and A. Y. Zomaya,
“Energy-efficient data replication in cloud computing datacen-
ters,” In IEEE Globecom Workshops, 2013, pp. 446-451. .
[4] Y. Deswarte, L. Blain, and J-C. Fabre, “Intrusion tolerance in dis-
tributed computing systems,” In Proceedings of IEEE Computer
Society Symposium on Research in Security and Privacy, Oakland
CA, pp. 110-121, 1991.
[5] B. Grobauer, T.Walloschek, and E. Stocker, “Understanding
cloud computing vulnerabilities,” IEEE Security and Privacy, Vol.
9, No. 2, 2011, pp. 50-57.
[6] W. K. Hale, “Frequency assignment: Theory and applications,”
Proceedings of the IEEE, Vol. 68, No. 12, 1980, pp. 1497-1514.
[7] K. Hashizume, D. G. Rosado, E. Fernndez-Medina, and E. B.
Fernandez, “An analysis of security issues for cloud comput-
ing,” Journal of Internet Services and Applications, Vol. 4, No. 1,
2013, pp. 1-13.
[8] M. Hogan, F. Liu, A.Sokol, and J. Tong, “NIST cloud computing
standards roadmap,” NIST Special Publication, July 2011.
[9] W. A. Jansen, “Cloud hooks: Security and privacy issues in
cloud computing,” In 44th Hawaii IEEE International Conference
onSystem Sciences (HICSS), 2011, pp. 1-10.
[10] A. Juels and A. Opera, “New approaches to security and
availability for cloud data,” Communications of the ACM, Vol.
56, No. 2, 2013, pp. 64-73.
[11] G. Kappes, A. Hatzieleftheriou, and S. V. Anastasiadis, “Dike:
Virtualization-aware Access Control for Multitenant Filesys-
tems,” University of Ioannina, Greece, Technical Report No.
DCS2013-1, 2013.
[12] L. M. Kaufman, “Data security in the world of cloud comput-
ing,” IEEE Security and Privacy, Vol. 7, No. 4, 2009, pp. 61-64.
[13] S. U. Khan, and I. Ahmad, “Comparison and analysis of
ten static heuristics-based Internet data replication techniques,”
Journal of Parallel and Distributed Computing, Vol. 68, No. 2, 2008,
pp. 113-136.
[14] A. N. Khan, M. L. M. Kiah, S. U. Khan, and S. A. Madani,
“Towards Secure Mobile Cloud Computing: A Survey,” Future
Generation Computer Systems, Vol. 29, No. 5, 2013, pp. 1278-1299.
[15] A. N. Khan, M.L. M. Kiah, S. A. Madani, and M. Ali, “En-
hanced dynamic credential generation scheme for protection
of user identity in mobile-cloud computing, The Journal of
Supercomputing, Vol. 66, No. 3, 2013, pp. 1687-1706 .
[16] T. Loukopoulos and I. Ahmad, “Static and adaptive dis-
tributed data replication using genetic algorithms,” Journal of
Parallel and Distributed Computing, Vol. 64, No. 11, 2004, pp.
1270-1285.
[17] A. Mei, L. V. Mancini, and S. Jajodia, “Secure dynamic frag-
ment and replica allocation in large-scale distributed file sys-
tems,” IEEE Transactions on Parallel and Distributed Systems, Vol.
14, No. 9, 2003, pp. 885-896.
[18] L. Qiu, V. N. Padmanabhan, and G. M. Voelker, “On the
placement of web server replicas,” In Proceedings of INFOCOM
2001, Twentieth Annual Joint Conference of the IEEE Computer and
Communications Societies, Vol. 3, pp. 1587-1596, 2001.
[19] D. Sun, G. Chang, L. Sun, and X. Wang, “Surveying and
analyzing security, privacy and trust issues in cloud computing
environments,” Procedia Engineering, Vol. 15, 2011, pp. 2852
2856.
[20] Y. Tang, P. P. Lee, J. C. S. Lui, and R. Perlman, “Secure overlay
cloud storage with access control and assured deletion,” IEEE
Transactions on Dependable and Secure Computing, Vol. 9, No. 6,
Nov. 2012, pp. 903-916.
[21] M. Tu, P. Li, Q. Ma, I-L. Yen, and F. B. Bastani, “On the
optimal placement of secure data objects over Internet,” In
Proceedings of 19th IEEE International Parallel and Distributed
Processing Symposium, pp. 14-14, 2005.
[22] D. Zissis and D. Lekkas, “Addressing cloud computing secu-
rity issues,” Future Generation Computer Systems, Vol. 28, No. 3,
2012, pp. 583-592.
[23] J. J. Wylie, M. Bakkaloglu, V. Pandurangan, M. W. Bigrigg,
S. Oguz, K. Tew, C. Williams, G. R. Ganger, and P. K. Khosla,
“Selecting the right data distribution scheme for a survivable
storage system,” Carnegie Mellon University, Technical Report
CMU-CS-01-120, May 2001.
[24] M. Newman, Networks: An introduction, Oxford University
Press, 2009.
[25] A. R. Khan, M. Othman, S. A. Madani, S. U. Khan,
“A survey of mobile cloud computing application
models,” IEEE Communications Surveys and Tutorials, DOI:
10.1109/SURV.2013.062613.00160.
Mazhar Ali is currently a PhD student at North Dakota State Uni-
versity, Fargo, ND, USA. His research interests include information
security, formal verification, modeling, and cloud computing systems.
Kashif Bilal did his PhD in Electrical and Computer Engineering
from the North Dakota State University, USA. His research interests
include data center networks, distributed computing, and energy
efficiency.
Samee U. Khan is an assistant professor at the North Dakota State
University. His research interest include topics, such as sustainable
computing, social networking, and reliability. He is a senior member
of IEEE, and a fellow of IET and BCS.
Bharadwaj Veeravalli is an associate professor at the National
University of Singapore. His main stream research interests include,
Scheduling problems, Cloud/Cluster/Grid computing, Green Stor-
age, and Multimedia computing. He is a senior member of the IEEE.
Keqin Li is a SUNY distinguished professor. His research interests
include mainly in the areas of design and analysis of algorithms,
parallel and distributed computing, and computer networking. He is
a senior member of the IEEE.
Albert Y. Zomaya is currently the chair professor of high per-
formance computing and networking in the School of Information
Technologies, The University of Sydney. He is a fellow of IEEE, IET,
and AAAS.

Drops division and replication of data in cloud for optimal performance and security

Recommended

More Related Content

What's hot (18)

Similar to Drops division and replication of data in cloud for optimal performance and security (20)

More from Pvrtechnologies Nellore (20)

Recently uploaded (20)

Drops division and replication of data in cloud for optimal performance and security