A novel cost-based replica server placement for optimal service quality in cloud-based content delivery network

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 13, No. 5, October 2023, pp. 5588~5598
ISSN: 2088-8708, DOI: 10.11591/ijece.v13i5.pp5588-5598  5588
Journal homepage: http://ijece.iaescore.com
A novel cost-based replica server placement for optimal service
quality in cloud-based content delivery network
Priyanka Dharmapal1
, Channakrishnaraju2
, Chethan Bommalingaiahanapalya Krishnamurthy3
1
Department of Computer Science and Engineering, Sri Siddhartha Academy of Higher Education University, Karnataka, India
2
Department of Computer Science and Engineering, Sri Siddhartha Institute of Technology, Karnataka, India
3
Department of Artificial Intelligence and Machine Learning, Vidyavardhaka College of Engineering,
Visveswaraiah Technological University, Karnataka, India
Article Info ABSTRACT
Article history:
Received Dec 10, 2022
Revised Apr 4, 2023
Accepted Apr 7, 2023
Replica server placement is one of the crucial concerns for a given
geographic diversity associated with placement problems in content delivery
network (CDN). After reviewing the existing literatures, it is noted that
studies are more for solving placement problem in conventional CDN and
not much over cloud-based CDN architectures, which some few studies are
reported towards replica selection are much in its nascent stages of
development. Moreover, such models are not benchmarked or practically
assessed to prove its effectiveness. Hence, the proposed study introduces a
novel design of computational framework associated with cloud-based CDN
which can facilitate cost-effective replica server management for enhanced
service delivery. Implemented using analytical research methodology, the
simulated study outcome shows that proposed scheme offers reduced cost,
reduced resource dependencies, reduced latency, and faster processing time
in contrast to existing models of replica server placement.
Keywords:
Cloud
Content delivery network
Cost
Replica server
Resource
This is an open access article under the CC BY-SA license.
Corresponding Author:
Priyanka Dharmapal
Research Scholar, Department of Computer Science and Engineering, Sri Siddhartha Academy of Higher
Education University
Tumkur, Karnataka-572017, India
Email: prsh2019@gmail.com
1. INTRODUCTION
With the rising demands of data and service availability, there is a need of an effective networking
technology that assists in the offering the seamless services irrespective of traffic condition in dedicated
manner. A content delivery network (CDN) is one such technology that assists in delivering the content
efficiently and quickly using surrogate server and content placement [1]–[5]. Although, there are several
reported advantage of adopting CDN system towards facilitating rate of data delivery, it is characterized by
various challenging condition too [6]. The primary challenge in deployment of CDN is its inclusion of large
sum of money with highly sophisticated process of deployment [7]. On the basis of multiple studies, it has
been noted that enabling technologies of CDN in current times make use of cloud computing more [8]. This
adoption offers beneficial characteristic of service provisioning, resource allocation, pay-as-per usage, and
hosting service supportability [9].
It should be noted that in the complete operation of CDN, the role of replica server is highly
important which is actually responsible for replicating the original content to content replica server.
Irrespective of various approaches being carried out towards content placement algorithm, there are still a
large trade-off between the demands of the user and the present state of operation of replica server. The
primary issue associated with the placement of replica server is associated with the optimization problem,

Int J Elec & Comp Eng ISSN: 2088-8708 
A novel cost-based replica server placement for optimal service quality in … (Priyanka Dharmapal)
5589
which could be either constraint or unconstrained type. The deployment cost of the replica server as well as
its associated delivery cost and update cost is not much emphasized upon in existing system with respect to
network topology, metric of network performance, latency, and count of hop. Apart from this, there are
various theoretical models presented in existing studies which has presented discussion about the facility
location with both capacitated and uncapacitated variant considering constraints as capacity of server.
Existing studies have also used K-median, minimum k-center, and k-cache location. for the purpose of
localization of replica server [10]. A thorough review of existing scheme also shows that majority of the
approaches associated with replica server placement is carried out on the basis of conventional CDN and not
towards cloud-based CDN. The prime difference between these two strategies is that conventional CDN are
bit centralized while cloud-based CDN are highly distributed and still it could be centralized to formulate a
large chain of network. It will mean that there are various inevitable challenges to be encountered when a
conventional CDN is exposed to challenges of heavier traffic condition as well as uncertain behavior of
various service and application running over it. In this entire scenario, it is also noted that cloud-based CDN
system also offers a less emphasis towards harnessing virtualized environment for increasing the coverage of
replica server. Although, it is nearly impossible to position the content server in multiple geographical places;
however, the cost can be significantly minimized when the existing virtualized environment (or machine) can
be classified into local and global form in order to offer connectivity in both smaller and larger scale. Hence,
it is essential to carry out an investigation in this direction in order to understand the impact of using local
virtual machine as well as proxies in order to facilitate better form of networking in cloud-based CDN
system. Further, it is also essential to understand the impact of spatial attributes towards the placement of the
replica server in cloud-based CDN and its possible impact in large scale network. Hence, in order to explore
these possibilities, the proposed study also carry out a deeper investigation of current literatures where it is
found that studies are more towards content placement and less towards replica server placement deploying
the virtualized environment.
The discussion of the relevant literature is as: our prior review has discussed about different
approaches and techniques required for improving the performance of CDN [11]. At present, there are
various further attempts that directly or indirectly contributes towards leveraging the performance of CDN.
The work carried out by Al-Abbasi et al. [12] have presented a probability-based model towards controlling
the stalling event during streaming services in CDN. A model is developed on the basis of restricted spaces in
caches along with allocation of CDN for better streaming delivery. Further, the work carried out by
Chuan et al. [13] have presented an optimized model for content placement considering belief propagation
system over cache in distributed network. The complete model is developed in a sequential flow where a
selection mechanism of content helper is carried out followed by caching the contents and delivering them at
proper destination. On the basis of states of communication channel and popularity, the modelling is carried
out in iterative manner emphasizing towards controlling energy consumption in wireless network. The work
carried out by Fan et al. [14] have presented an allocation model for replica server which is claimed to be
more energy efficient as well as reliable. The study model contributes to concurrent task processing over the
main server mapped with a dedicated storage point. The work carried out by Guerrero et al. [15] have
developed a model for data replication with a core intention towards leveraging data availability over a
sophisticated form of weighted network along with centrality factors. The study model also evaluates a graph
partitioned attributed while the data replicates are stored in fog devices. According to this model, one unit
stores fog information and a single replica of data while the other stores all file replica using greedy based
approach.
The work carried out by Yovita and Syambas [16] have discussed about the importance of the
caching mechanism which is highly essential for large distributed network. According to the study outcome,
there are yet unsolved problems towards caching mechanism. Kusuma et al. [17] have emphasized towards
the vertex markers in order to discuss possible scale of amendments towards grid methods.
Liu et al. [18] have presented a unique replica placement scheme that uses the existing software framework
for Hadoop in order to address the problems of large-scale voluminous data aggregated during data exchange.
The study considers three-dimensional raster data in order to perform better determination of location
associated with replica server placement while the study outcome show efficient performance with reduced
network overhead. Adoption of machine learning is also witnessed in existing scheme towards solving the
problem of selection of replica. This cadre of work is reported by Mostafa et al. [19] where a large scale
environment is considered by deploying artificial neural network (ANN) for profiling the behavior of the
location involved in the process. The outcome exhibits satisfactory predictive accuracy with sufficient
optimization of channel capacity. The study towards cost estimation is carried out by Nazir et al. [20] while
performing job scheduling over grid system. The work introduces a unique dynamic scheduling policy that is
centralized towards replica placement. The prime intention of this model is to reduce the cost of replica
placement by scheduling the data considering computing capacity of the node in order to explore the task to
be processed. A case study of telco CDN is reported in work of Safavi et al. [21] where an on-line learning

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 13, No. 5, October 2023: 5588-5598
5590
scheme towards replica placement is discussed considering the popularity of the contents. The spatial
patterns are computed that is associated with the request of the content in order to perform predictive
modelling on the basis of content popularity. As discussed by various work model, it was noted that content
caching is one of the essential operations for leveraging the perform of CDN. Study in such direction was
carried out by Saino et al. [22] focusing on both conventional and futuristic form of CDN where the authors
contributes towards introducing a unique placement of contents along with higher degree of flexibility
towards placement operation. The study outcome is exhibited to offer improved load balancing performance
and being less sensitive toward any form of traffic fluctuation.
The study of content placement is also carried out over 5G network services as noted in work of
Santos et al. [23] where the edge sites are targeted for placing the virtualized environment of CDN. The core
objective of the model is towards splitting the heavier multimedia file followed by localizing those splitted
files over dynamic and virtualized resources. Nearly similar form of approach is also carried out by Shankar
and Chitra [24] using integrated machine learning process. According to this model, the latency is addressed
by distributed storage of data followed by learning the placement coordinates using support vector machine
(SVM) and radial basis function (RBF). This learning scheme is capable of classifying data centers along
with prediction of possible traffic load by using either cloud or edge resources. Further, the optimization is
carried out using graph partitioning in dynamic approach in order to cater up the latency demands with
reduced cost of placement. Shao et al. [25] have carried out an investigation towards various processes of
selection of replica followed by techniques towards itself placement in edge environment as well as in
internet of things (IoT) environment. The review outcome stated the importance of provenance of data in
order to improve accuracies. The author also stated the ongoing challenges of replica placement is quite
complex to deal with especially in presence of dynamic CDN environment. Teng et al. [26] have presented a
scheme of content placement considering the problem of delay minimization by harnessing biconvexity.
The work carried out by Tran et al. [27] have presented a framework using software defined
network on CDN which primarily emphasizes on selection of server followed by constructing an intelligent
controller system which is primarily meant for resisting all possible overloading condition in SDN. The work
carried out by Xiong et al. [28] have presented a strategy for replication management in order to mitigate the
delay during access considering data with spatial and temporal characteristics on the basis of caching system.
According to the implementation model, the maximized popularity score is mined using correlational factor
associated with an access of user by adopting beneficial aspect of location. It also uses all connected files
obtained from the information of user access. This process finally yields replicates followed by selection of
cache node in order to fulfil the purpose of placement. The work carried out by Xu et al. [29] have presented
a scheme have presented an optimization framework using mixed integer linear programming where the
complete model initiates by optimal selection of replica server followed by undertaking decision towards
caching the content items within the replica server. Finally, all the varied servers are allocated with the loads
of content request followed by constructing a heuristic method. Another unique work has been presented by
Yu and Pan [30] where the problems associated with conventional hashing method for distributed data
storage in Hadoop is addressed. According to this study model, a hypergraph-based methodology for solving
the data placement has been discussed that takes into account varied metrics of performance connected with
the location information of the target data in cumulative fashion. The study model uses this method in order
to partition the selected dataset items and relocate them in different location of distributed storage units. The
study is also claimed to be capable enough for exploring the presence of redundant data.
After reviewing the above-mentioned existing studies, the conclusive statement of problem is as: it
is computationally challenging task to develop a generalized scheme of replica node placement over
distributed environment of cloud-based CDN. The prime reason behind this statement are as: i) majority of
the existing studies have emphasized on quality of service (QoS) attribute, which is essential; however,
consideration of user’s experience is also equally important which is not much reported to be emphasized, ii)
existing approaches has deployed a mechanism considering varied forms of files; however, such schemes are
not much effective when deployed under a situation of non-uniform distance between content server and
replica server; and iii) further, non-utilization of virtualized environment is less seen in the existing scheme.
The proposed solution presented in this paper discusses about a unique replica server placement by
harnessing the data availability potential of local virtualized machined and proxies connected to centralized
content server that is further maintained in a distributed form. The core idea of this work is mainly towards
offers similar user experience quality as well as service quality in presence of dynamic traffic condition in
cost effective manner.
The value added in proposed scheme are as: i) the proposed scheme presents a cost-based optimal
replica server placement strategy which equally contributes towards task allocation unlike existing schemes;
ii) adoption of non-iterative scheme towards flow processing considering proximity importance attribute

5591
among all the nodes connection in large network; and iii) faster processing time to process the query
irrespective of any form of traffic. The next section discusses about adopted method.
The organization of the proposed manuscript is as: section 1 discusses introduction while the
section 2 discusses about methodology being undertaken, section 3 discusses about the result being
accomplished while section 4 discusses about the conclusion of the paper.
2. METHOD
The prime aim of this stage of study is to develop a novel model that can perform provisioning of
resources to ensure better replica server placement in dynamic distributed cloud-based CDN system.
Adopting analytical research methodology, the proposed system initiates its design using communication
model considering multi-cloud system as shown in Figure 1. For simplifying the topology design, the
proposed system implements a tree structure that are highly clustered. According to this tree structure, each
user is served by a dedicated cloud system using surrogate server that are further incorporated with
operations associated with virtual machine, storage, and bandwidth. The next part of implementation is
associated with developing a cloud-based CDN resource management system that will be further classified
into planning stage and reallocation of user stage. The proposed system will use the cost model for assessing
the cost incurred for resource allocation in present scenario. In the next stage of user reallocation, three set of
operations are carried out toward resource provisioning considering formulation of practical constraints and
achieving highest optimality in replica server placement. Finally, an objective function will be designed
which can balance the resource provisioning demands with dynamic content delivery system for effective
replica server placement in cloud-based CDN system.
This part of implementation is anticipated to achieve optimal performance in cloud-based CDN.
A test environment will be constructed to assess the influence of various network attributes e.g., channel
capacity, cost, and uncertain traffic situation on the model. The model is also anticipated to exhibited lower
computational complexity. The next section discusses about the research methodology adopted for the
purpose of implementation briefed with respect to system design and algorithm implementation.
Communication Model
Tree Structure (clustered)
Surrogate server
Virtual Machine
Storage
Bandwidth
Cloud-based CDN Resource Management
Planning Stage
Reallocation
of user stage
local
Global
optimal
Objective function
Figure 1. Architecture of 2nd
module of implementation

 ISSN: 2088-8708
5592
2.1. System design
The core system design of the proposed scheme is based on the location and proximity of the replica
server and the proxy cache in order to offer better availability of service via cloud networks. According to
this scheme, a centralized content server (CCS) is considered which is positioned at uniform proximity of
multiple local virtual machine (LVM) which retains information about replica server and proxy cache.
The Figure 2 highlights the placement strategy of replica server where ith
number of multiple
observation zone 𝑂𝑧 exists i.e., 𝑂𝑧𝑖 = [𝑂𝑧1, 𝑂𝑧2, … 𝑂𝑧𝑖] where the placement of LVM is carried out. It will
mean that each 𝑂𝑧 consist of specific number of LVM (depending upon the geographic spread of 𝑂𝑧).
Mathematically, it can be expressed as (1).
𝐿𝑉𝑀𝑖 = [𝑃𝑖, 𝑅𝑖]𝛼𝛽
(1)
The (1) shows ith
number of LVM is positioned in 𝑂𝑧𝑖 and each LVMi consists of proxy cache data Pi
and replica server Ri considering 𝛼 as total proxy server and specific bandwidth 𝛽 associated with replica
server. Each LVM are connected to each other by one-to-many relation in unique fashion. In order to
simplify this process, the proposed scheme allocate a link coefficient 𝛾 for each of the possible connection
from each LVM as shown in (2).
𝛾 = ∑ 𝐿𝑉𝑀𝑗
𝑙
⁄
𝑘
𝑗=1 (2)
In the mathematical expression (2), the variable 𝑗 and 𝑙 will represent specific number of LVM and
remaining number of individual LVM to be connected with current LVM. It should be noted that this is only
possible when the proposed scheme is designed using tree-structure. The contribution of link coefficient is
that it assists in formulating LVM connectivity with respect to reduced cost involved in data transmission
over the assigned tree structure. One of the contributions of the proposed scheme is its placement strategy
which is circular in its orientation that targets to offer a uniform performance of the data delivery services
from the CCS. However, it should be noted that multiple CCS joins in the tree structure to formulate highly
distributed and yet well-connected to each other targeting persistent quality of service delivery. Another
contribution in this scheme is uniform utilization of bandwidth. It will eventually mean that each node of
replica server in tree structure is characterized by unique orientation in the form of in-degree and out-degree
links where each node bears information associated with service provider of CDN discretely defined for each
direction in distributed manner. Another best part is its capability to identify the presence of any form of
redundant data that may possibly reside in cache proxy.
Figure 2. Scheme of replica server placement

5593
2.2. Algorithm implementation
The proposed system introduces a very simple and yet novel algorithm which uses cost estimation in
order to perform placement of replica server. It is to be noted that while performing the placement of replica
server, the importance is given more towards LVM as it is the bridge between the replica server and the
actual content server connected in distributed manner. The steps of proposed algorithm are as:
Algorithm 1. For cost-based replica server placement
Input: no, ne, LVM, j, S
Output: Φ
Start
1. init net[no, ne(LVMj), S]
2. G→(no, ne)S
3. θ=μ(G)
4. formulate matrix G=[θ, IL, OL]
5. Extract new tree=ω(G)
6. Lt=(G, cond(prob))
7. Φ=ρ(θ)
End
The algorithm 1 takes the input of origin node no, end node ne, local virtual machine LVM, number
of LVM j, and service being hosted by the CDN S that gives the resultant outcome of cost Φ. The preliminary
set of tasks of the proposed algorithm is to initialize specific number (i.e., 𝑗) of LVM considering the
structure of its placement and connectivity with origin and end node (Line-1). The next part of
implementation is associated with construction of graphical tree 𝐺 for mapping the complete nodes and links
in the form of a network topology in distributed and large-scale order (Line-2). It is to be noted that proposed
scheme constructs distributed scheme of replicate server placement by connecting LVMs associated with
centrally located CCS (Line-3). The computation of the central location is carried out by obtaining the score
of each LVM that is divided with respect to available least quantity of nodes which has essential
characteristics of edge information linking with the respective LVM. The outcome of this operation gives the
score of importance attribute θ using a specific function μ doing the above-mentioned task by taking the input
argument of 𝐺 (Line-3). After computing the importance attribute, the algorithm formulates a two-
dimensional matrix G that retains information associated with 𝜃 importance attribute, incoming link 𝐼𝐿, and
outgoing link 𝑂𝐿 (Line-4). This mechanism is deployed in order to construct a distributed network links
connecting all the LVMs in different location. This assessment is performed for the purpose of assessing the
consistency of the proposed algorithm in order to acquire reduced delay considering multiple position of the
LVM associated with the replica server in specific 𝑂𝑧. After this task is accomplished, the algorithm deploys
another function ω that constructs a diagraph tree structure G over the existing tree (Line-5). The system than
generates a local tree Lt considering the newly constructed tree G obtained in prior step followed by selection
of the LVM with an embedded statistical condition cond (Line-6). According to this condition cond, the
construction of the Lt is formed only when its probability value prob is found to be statistically significant. In
the last step, an explicit function 𝜌 is formed that performs final estimation of the cost 𝛷 associated with the
placement of replica server at the time of data transmission (Line-7). The algorithm considers importance
attribute 𝜃 considering the LVM as well as their location-specific data during this mechanism.
The complete operation of the cost function 𝜌 is stated as:
a) The cost function 𝜌 constructs a two-dimensional matrix of (d1×d2) where each elements of the matrix
represents the cost attributes that is allocated with specific d1 number of LVM with d2 storing all the
discrete jobs that it has to process. The construction of this matrix is carried out in specific form to retain
least number of d1 and d2 in the form of rows and columns where allocation of bandwidth 𝛽 is defined as
𝛽 = 𝑚𝑖𝑛𝑖𝑚𝑢𝑚(𝑑1, 𝑑2).
b) The consecutive process is to look for least number of matrix element in complete d1 rows followed by
subtracting it from residual elements residing in the rows.
c) All the elements in the matrix are then looked for presence of any element with value zero. The system
than flags all the element which is found to be non-zero elements in the matrix. Similar process is
repeated for all the elements residing within the matrix.
d) The algorithm encapsulates all the flagged zero elements residing within the matrix in the form of
columnar position. If the system finds βtot number of columnar elements that are flagged with the zero
elements than it represents this as a total allocation of bandwidth in unique form. This is the usual
scenario of termination of total operation of an algorithm, otherwise it is resumed for next level of
processes.
e) The system identified all the non-encapsulated zero elements followed by transforming into prime
numbers. If further there is no flagged zero elements over the rows in matrix, the function proceeds to
next step otherwise it encapsulates the specific elements maintained in row while it does not encapsulate

 ISSN: 2088-8708
5594
any columnar elements with elements flagged with zero. This process is continued until the function
encapsulates all the zero elements followed by saving all the minimal value of non-encapsulated elements
in matrix.
f) The function then constructs all the primed elements with zeros as well as flagged elements with zero.
g) The outcome of this process is yielded in 5th
step and this outcome is appended complete elements in the
matrix in the form of encapsulated form maintained over rows that is subtracted from all the
non-encapsulated elements in columnar form. The function further performs operation highlighted in 5th
step above without any need of changing flagging or primes.
h) Therefore, the function yields a set of scores that is required to be assigned to the node taking into
account for location of all the elements with zero and flagging over the defined cost matrix. It will
eventually mean that if matrix of dimension d1×d2 is found to be flagged as zero than all the scores
associated with the respect to d1 row will be assigned to scores that are associated with d2 column.
Therefore, the resultant outcome of the last function ρ is considered to be an optimal result with
significantly least cost of placing replica server. The next section discusses about the accomplished
outcome.
3. RESULTS AND DISCUSSION
From the prior section, it was noted that proposed scheme introduces a novel mechanism towards
replica server placement in order to ascertain better performance delivery in cloud-based distributed CDN
system. Hence, it is essential to chalk out a definitive assessment strategy in order to ensure better data
transmission performance with respect to variable test environment in CDN. The algorithm of the proposed
system is scripted in MATLAB emphasing towards accomplishing the tree structure deployment where cost
for replica server placement is the prime analysis factor towards result analysis. Table 1 discusses about the
evaluated outcomes of cost that is linked with individual LVM residing over multiple orientation considering
central location of CCS. The assessment of proposed scheme is carried out by constructing a bipartite graph.
Table 1. Simulation parameters
Simulation parameter Values assigned
Total deployment areas of LVM 9
Total replica server 10
Bandwidth assigned for replica server 5,000 Mbps
Total proxy (cache) 25
Bandwidth assigned for proxy 3,000 Mbps
Deploying the simulation parameters and its respective values exhibited in Table 1, the proposed
system carries out cost estimation for individual LVM in 9 different circular position considering CCS at its
center. The complete analysis is carried out over variable size of random data packets of 2,500 bytes over a
simulation area of size 1,000×1,000 m2
considering presence of 500 nodes. The assessment is carried out
considering randomly selected origin node and end node. The individual outcomes of cost in 9 different
locations are shown in Table 2 assessed over 50 iterations.
Table 2. Accomplished cost (𝛷)
Origin Estimated cost for nodal location of 9 LVMs
1 2 3 4 5 6 7 8 9
LVM2 71 73 28 65 23 93 5 25 56
LVM4 65 98 14 97 68 1 19 93 44
LVM1 4 54 29 25 85 47 73 28 65
LVM3 8 34 45 69 35 43 48 78 66
LVM2→LMV1 33 12 54 30 79 47 16 20 69
LVM2→LVM3 54 62 47 68 69 78 35 30 65
LVM4→LVM1 66 79 89 71 2 33 62 10 96
LVM4→LVM3 42 43 53 8 61 79 20 59 22
CCS 83 10 95 26 40 48 75 69 72
In order to assess the sustainability of the network, the analysis is carried out by further increasing
the size of packet to 3,000 bytes where observation being carried out for 500 iterations in order to arrive at
final averaged outcome of estimated cost of replica server placement as exhibited in Table 3. Comparing the

5595
difference of trends in cost in Tables 2 and 3, it can be seen the cost values get significantly reduced over a
period of time with an inclusion of more number of traffic and more iteration compared to less number of
traffic in CDN. The prime reason behind this is topology gets more branched with increasing traffic while
allocation of resources is also distributive carried out which splits up the computational burden of each LVM
using tree structure. Hence, adoption of tree structure can be seen as one contributory points towards
minimizing the cost of replica server placement in cloud-based CDN system.
Table 3. Estimated cost
Location LVM No 𝛷
LVM2 7 4
LVM4 6 0
LVM1 3 28
LVM3 1 7
LVM2→LMV1 2 11
LVM2→LVM3 8 29
LVM4→LVM1 5 1
LVM4→LVM3 9 21
CCS 4 25
Figure 3 highlights the outcome associated with the estimated cost of replica server placement
where a gain of approximately 45.2% is accomplished in contrast to existing schemes of QoS aware [31] and
consistency-aware algorithms [31] that is frequently adopted in existing replica server placement. The
existing scheme of QoS aware approach is noted to carry out pre-defined computation where the system
ascertains its network topology in order to accomplish a known target QoS via LVM for processing generated
query. This static characteristic of QoS-aware approach is therefore witnesses with discrepancies when
exposed to increasing traffic that further increases the cost of replica server placement allocation. The
consistency-aware algorithm is found to offer slightly efficient performance in comparison to QoS-aware
approach.
Figure 3. Comparative evaluation of cost of replica server placement
However, it was noted that such approach increasingly demands more computational effort in order
to find an optimal condition which is not permitted by the nature of consistency-aware algorithm. It has to
settle with the defined network attributes and hence fails to address the dynamic demands of traffic causing
extra cost involved in replica server placement. On the other hand, proposed scheme is found to offer better
capability of reducing the operational cost in presence of increasing traffic as it can handle multiple tasks
using its symmetrical form of placement of both distributed CCS aligned with multiple LVM. Hence, the
nearly uniform computational effort is used without introducing any form of staticness in the nature of the
topology-building process. Further, the computed cost is also subjected to updating process with every event
of topological alteration, which directly assists allocating the cost for other communicating nodes. Hence,
proposed scheme offers reduced cost for replica server placement in cloud CDN environment.

 ISSN: 2088-8708
5596
Figure 4 showcase approximately 37.4% of improvement of better resource allocation compared to
existing schemes. The prime reason behind this is the operation carried out by proposed scheme toward each
incoming and outgoing traffic by maintaining equal balance between normal and priority task involved for
every serviced hosted for analysis. This argument also contributes towards the reduction of latency as well as
processing time as exhibited in Figures 5 and 6. The reduction in latency is found to be approximately 67.5%
while the reduction processing time is approximately 37.2%. The conclusive remark of these outcome can be
stated to be nearly similar pattern of outcomes associated with existing system which emphasizes more
towards solving problem space of local networks whereas proposed scheme initiated with a large
interconnected distributed network where each link is characterized by a weight while the nodes are allocated
with cost factors. This assists in better deployment of decision making while performing data delivery in
cloud-based CDN.
Figure 4. Comparative evaluation of probability of
resource allocation
Figure 5. Comparative evaluation of latency
incurred
Figure 6. Comparative evaluation of processing time
4. CONCLUSION
The prime contribution of the proposed study model is to develop a unique modelling towards cost
effective positioning of replica server considering the use case of cloud-based CDN system. The contribution
of the proposed study model are as: i) a unique mechanism of computed cost based construction of structure
of records of network towards reducing the dependencies of more number of replica server that indirectly
reduces cost of data transmission; ii) the proposed scheme implements a unique concept of local virtual
machine in order to replace the convention content server for cost reduction; this adaptation increases

5597
virtualization effectiveness creating a large chain of highly indexed structure resulting in higher availability
of data and services; iii) the proposed scheme of replica server placement offers equal importance to user’s
experience and service quality; and iv) the study outcome shows that proposed system offers much better
performance in contrast to existing scheme.
REFERENCES
[1] I. Kilanioti et al., “Towards efficient and scalable data-intensive content delivery: state-of-the-art, issues and challenges,” in
Lecture Notes in Computer Science, Springer International Publishing, 2019, pp. 88–137.
[2] B. Zolfaghari et al., “Content delivery networks,” ACM Computing Surveys, vol. 53, no. 2, pp. 1–34, Mar. 2021,
doi: 10.1145/3380613.
[3] T. Bilen and B. Canberk, “Deliver the content over multiple surrogates: A request routing model for high bandwidth requests,”
Computer Communications, vol. 146, pp. 39–47, Oct. 2019, doi: 10.1016/j.comcom.2019.07.009.
[4] S. Ul Islam et al., “Leveraging utilization as performance metric for CDN enabled energy efficient internet of things,”
Measurement, vol. 147, Dec. 2019, doi: 10.1016/j.measurement.2019.07.042.
[5] G. Eslami and A. T. Haghighat, “A new surrogate placement algorithm for cloud-based content delivery networks,” The Journal
of Supercomputing, vol. 73, no. 12, pp. 5310–5331, Dec. 2017, doi: 10.1007/s11227-017-2088-5.
[6] Q. Fan et al., “Video delivery networks: Challenges, solutions and future directions,” Computers and Electrical Engineering,
vol. 66, pp. 332–341, Feb. 2018, doi: 10.1016/j.compeleceng.2017.04.011.
[7] F. Ayankoya, O. Otushile, and B. Ohwo, “A review on content delivery networks and emerging paradigms,” International
Journal of Scientific and Engineering Research, vol. 9, no. 12, pp. 211–217, 2018.
[8] L. Yala, “Content delivery networks as a service (CDNaaS),” Universite Rennes 1, 2018.
[9] M. A. Salahuddin, J. Sahoo, R. Glitho, H. Elbiaze, and W. Ajib, “A survey on content placement algorithms for cloud-based
content delivery networks,” IEEE Access, vol. 6, pp. 91–114, 2018, doi: 10.1109/ACCESS.2017.2754419.
[10] D. Priyanka, Channakrishnaraju, and B. K. Chethan, “Insights on effectiveness towards research approaches deployed in content
delivery network,” in Software Engineering Perspectives in Systems, Springer International Publishing, 2022, pp. 224–243.
[11] V. Cohen-Addad, A. Gupta, L. Hu, H. Oh, and D. Saulpic, “An improved local search algorithm for k -median,” in Proceedings
of the 2022 Annual ACM-SIAM Symposium on Discrete Algorithms (SODA), Philadelphia, PA: Society for Industrial and Applied
Mathematics, 2022, pp. 1556–1612.
[12] A. O. Al-Abbasi, V. Aggarwal, T. Lan, Y. Xiang, M.-R. Ra, and Y.-F. Chen, “FastTrack: minimizing stalls for CDN-based over-
the-top video streaming systems,” IEEE Transactions on Cloud Computing, vol. 9, no. 4, pp. 1453–1466, Oct. 2021,
doi: 10.1109/TCC.2019.2920979.
[13] J. Chuan, B. Bai, X. Wu, and H. Zhang, “Optimizing content placement and delivery in wireless distributed cache systems
through belief propagation,” IEEE Access, vol. 8, pp. 100684–100701, 2020, doi: 10.1109/ACCESS.2020.2996222.
[14] Y. Fan, C. Wang, W. Wu, T. Znati, and D. Du, “Slow replica and shared protection: energy-efficient and reliable task assignment
in cloud data centers,” IEEE Transactions on Reliability, vol. 70, no. 3, pp. 931–943, Sep. 2021, doi: 10.1109/TR.2019.2923770.
[15] C. Guerrero, I. Lera, and C. Juiz, “Optimization policy for file replica placement in fog domains,” Concurrency and Computation:
Practice and Experience, vol. 32, no. 21, Nov. 2020, doi: 10.1002/cpe.5343.
[16] L. V. Yovita and N. R. Syambas, “Caching on named data network: a survey and future research,” International Journal of
Electrical and Computer Engineering (IJECE), vol. 8, no. 6, pp. 4456–4466, Dec. 2018, doi: 10.11591/ijece.v8i6.pp4456-4466.
[17] W. T. Kusuma, A. A. Supianto, and H. Tolle, “Vertex markers: Modification of grid methods as markers to reproduce large size
augmented reality objects to afford hands,” International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 1,
pp. 1063–1069, Feb. 2020, doi: 10.11591/ijece.v10i1.pp1063-1069.
[18] Z. Liu, W. Hua, X. Liu, D. Liang, Y. Zhao, and M. Shi, “An efficient group-based replica placement policy for large-scale
geospatial 3D raster data on Hadoop,” Sensors, vol. 21, no. 23, Dec. 2021, doi: 10.3390/s21238132.
[19] N. Mostafa, W. H. F. Aly, S. Alabed, and Z. Al-Arnaout, “Intelligent replica selection in edge and IoT environments using
artificial neural networks,” Electronics, vol. 11, no. 16, Aug. 2022, doi: 10.3390/electronics11162531.
[20] B. Nazir, F. Ishaq, S. Shamshirband, and A. Chronopoulos, “The impact of the implementation cost of replication in data grid job
scheduling,” Mathematical and Computational Applications, vol. 23, no. 2, May 2018, doi: 10.3390/mca23020028.
[21] M. Safavi, S. Bastani, and B. Landfeldt, “Online learning and placement algorithms for efficient delivery of user generated
contents in telco-CDNs,” IEEE Transactions on Network and Service Management, vol. 17, no. 1, pp. 637–651, Mar. 2020,
doi: 10.1109/TNSM.2019.2961560.
[22] L. Saino, I. Psaras, and G. Pavlou, “Framework and algorithms for operator-managed content caching,” IEEE Transactions on
Network and Service Management, vol. 17, no. 1, pp. 562–576, Mar. 2020, doi: 10.1109/TNSM.2019.2956525.
[23] D. Santos, R. Silva, D. Corujo, R. L. Aguiar, and B. Parreira, “Follow the user: a framework for dynamically placing content
using 5G-enablers,” IEEE Access, vol. 9, pp. 14688–14709, 2021, doi: 10.1109/ACCESS.2021.3051570.
[24] B. P. Shankar and S. Chitra, “Optimal data placement and replication approach for SIoT with edge,” Computer Systems Science
and Engineering, vol. 41, no. 2, pp. 661–676, 2022, doi: 10.32604/csse.2022.019507.
[25] Z.-L. Shao, C. Huang, and H. Li, “Replica selection and placement techniques on the IoT and edge computing: a deep study,”
Wireless Networks, vol. 27, no. 7, pp. 5039–5055, Oct. 2021, doi: 10.1007/s11276-021-02793-x.
[26] W. Teng, M. Sheng, K. Guo, and Z. Qiu, “Content placement and user association for delay minimization in small cell networks,”
IEEE Transactions on Vehicular Technology, vol. 68, no. 10, pp. 10201–10215, Oct. 2019, doi: 10.1109/TVT.2019.2936182.
[27] H.-A. Tran, S. Souihi, D. Tran, and A. Mellouk, “MABRESE: a new server selection method for smart SDN-based CDN
architecture,” IEEE Communications Letters, vol. 23, no. 6, pp. 1012–1015, Jun. 2019, doi: 10.1109/LCOMM.2019.2907948.
[28] L. Xiong, L. Yang, Y. Tao, J. Xu, and L. Zhao, “Replication strategy for spatiotemporal data based on distributed caching
system,” Sensors, vol. 18, no. 1, Jan. 2018, doi: 10.3390/s18010222.
[29] K. Xu, X. Li, S. K. Bose, and G. Shen, “Joint replica server placement, content caching, and request load assignment in content
delivery networks,” IEEE Access, vol. 6, pp. 17968–17981, 2018, doi: 10.1109/ACCESS.2018.2817646.
[30] B. Yu and J. Pan, “A framework of hypergraph-based data placement among geo-distributed datacenters,” IEEE Transactions on
Services Computing, vol. 13, no. 3, pp. 395–409, May 2020, doi: 10.1109/TSC.2017.2712773.
[31] J. Sahoo, M. A. Salahuddin, R. Glitho, H. Elbiaze, and W. Ajib, “A survey on replica server placement algorithms for content
delivery networks,” IEEE Communications Surveys & Tutorials, vol. 19, no. 2, pp. 1002–1026, 2017,
doi: 10.1109/COMST.2016.2626384.

 ISSN: 2088-8708
5598
BIOGRAPHIES OF AUTHORS
Priyanka Dharmapal has 10 years of teaching experience for UG and PG
courses in Computer Science and Engineering and is presently working as Assistant Professor
in the Department of Computer Science and Engineering in Sri Siddhartha Institute of
Technology, Tumkur. She obtained B.E degree from Visveswaraiah Technological University
in the year 2009 and PG in Computer Science and Engineering in the year 2012 from
Visveswaraiah Technological University. Her research interests are in the areas of content
delivery networks, network security, mobile computing and cloud computing. She is currently
pursuing doctoral degree in Sri Siddhartha Academy of Higher Education, Tumkur, under the
guidance of Dr. Channakrishnaraju, Professor, Department of Computer Science and
Engineering, Sri Siddhartha Institute of Technology, Tumkur. She can be contacted at email:
prsh2019@gmail.com.
Channakrishnaraju has 26 years of teaching experience for UG and PG courses
in computer Science and Engg, and is presently working as Professor in the Department of
computer Science and Engg, in Sri Siddhartha Institute of Technology, Tumkur. He obtained
B.E from Bangalore University in the year 1995 and PG in software systems in the year 2000
from BITS Pilani and doctoral degree in Sri Siddhartha Academy of higher Education,
Tumkur. His research interests are in the areas of wireless sensor networks, network security
and artificial intelligence. He can be contacted at email: rajuck@ssit.edu.in.
Chethan Bommalingaiahanapalya Krishnamurthy working as Associate
Professor Department of CSE (AI and ML) in Vidyavardhaka college of Engineering, Mysuru,
India. He obtained B.E degree from Visveswaraiah Technological University in the year 2005
and PG in Computer Science and Engineering in the year 2007 from Visveswaraiah
Technological University and doctoral degree in the year 2020 from Visveswaraiah
Technological University. He has around 16 years of teaching experience. His research domains
are computer network, network security and software engineering. He can be contacted at
email: chethan08@gmail.com.

A novel cost-based replica server placement for optimal service quality in cloud-based content delivery network

More Related Content

Similar to A novel cost-based replica server placement for optimal service quality in cloud-based content delivery network (20)

More from IJECEIAES (20)

Recently uploaded (20)

A novel cost-based replica server placement for optimal service quality in cloud-based content delivery network