Network Traffic Management using Dynamic Bandwidth on Demand

Abstract

Traffic Analysis and measurement in large networks is very challenging task for network managers. Bandwidth plays a vital role during network traffic analysis and management. Bandwidth allocation becomes a critical issue for effective network management. Bandwidth on demand concept gradually evolved while addressing the need of network managers for monitoring on-demand traffic. Use of efficient bandwidth allocation algorithm significantly improves network performance by assuring availability of network to all users. In this paper, we propose an optimized algorithm using the concept “rating of web pages”, which is based on users’ past accessibility. This algorithm assigns a minimum guaranteed bandwidth to each connected user, instead of equally dividing the total available bandwidth among the users. Finally, based on rating of web pages, any excess bandwidth is distributed dynamically among existing users. This significantly improves the average utilization of available bandwidth.

Full text

 Abstract— Traffic Analysis and measurement in large
networks is very challenging task for network managers.
Bandwidth plays a vital role during network traffic analysis and
management. Bandwidth allocation becomes a critical issue for
effective network management. Bandwidth on demand concept
gradually evolved while addressing the need of network
managers for monitoring on-demand traffic. Use of efficient
bandwidth allocation algorithm significantly improves network
performance by assuring availability of network to all users. In
this paper, we propose an optimized algorithm using the concept
“rating of web pages”, which is based on users’ past accessibility.
This algorithm assigns a minimum guaranteed bandwidth to
each connected user, instead of equally dividing the total
available bandwidth among the users. Finally, based on rating of
web pages, any excess bandwidth is distributed dynamically
among existing users. This significantly improves the average
utilization of available bandwidth.
Index Terms— Network Traffic classification, Software Defined
Network, BoD, SeLeCT, Load balancing, Incremental clustering.
I. INTRODUCTION
In the current scenario, almost all business
applications are being carried out over Internet. Online
businesses increasingly rely on Internet for its basic
operations. Along with increase in the complexity of Internet
services, there is drastic increase in Content Delivery
Networks (CDNs) and mobile Internet usage. With the growth
of technology along with increase in users, complexity will
continue to increase in the future. According to survey done
by CISCO in 2016, nearly 40% of the world population has
Internet connection which was less than 1% in 1995. Hence
there is a high demand for Internet traffic management.
Traffic Engineering (TE) deals with the measurement
and management of network traffic to designs optimized
network traffic for routing and improving network resources
utilization. When number of user increases, it causes
bottleneck problem in accessing the network. Passive network
is an effective solution to the bottleneck problem in accessing
P. C. Sethi is Ph.D scholar and working as Senior Research Fellow (SRF)
in Department of Computer Science and Appications, Utkal University, Vani
Vihar, Bhubaneswar, Odisha, India. The author can be reached over e-mail:
pcsethijrf14@utkaluniversity.ac.in.
P. K. Behera is working as Reader (Associate Professor) in Department of
Computer Science and Appications, Utkal University, Vani Vihar,
Bhubaneswar, Odisha, India. The author can be reached over e-mail:
p_behera@hotmail.com.
This work is supported under UGC grant RGNF-2013-14-ORI-49267.
the network. Remote device configuration, network
performance monitoring, network resource usage verification
and network fault detection are the major responsibilities of
SNMP. A third-party SNMP management software or a user
defined SNMP management software can be used for network
management.
This research paper deals with a diverse research
interests that focuses on Traffic Engineering (TE) based
Software Defined Network (SDN) for network traffic
monitoring, network traffic measurement and management for
efficient processing as compared to traditional processing.
SDN is a way to deal with network organization that permits
the network managers to manage the system based on abstract
lower-level functionality. Since the static design of traditional
network doesn't bolster the dynamic, versatile figuring and
capacity needs of more advanced processing situations, SDN
idea is utilized by the various data center. This is
accomplished by decoupling or disassociating the framework
that settles on choices about where network traffic is available.
SDN is commonly associated with the OpenFlow protocols.
To begin with, we propose a reference system for TE
in SDN based on page rating. It comprises of two sections,
such as, network traffic estimation and network traffic
administration. Network traffic was estimated by monitoring
the real network and breaking down the system into different
activities. Network traffic estimation is the prerequisite for
traffic administration. Network traffic measurement and
forecasting is the fundamental requirement in the network
traffic management. Traffic load balancing, guaranteed
scheduling of network information are the related fields of
network traffic management. Here, network traffic
management using web page rating was proposed for
improving the quality of service.
The rest of this paper is organized as follows: Section 2
presents the literature overview related to network traffic
management, section 3 describes the proposed work, section 4
contains the proposed algorithm implemented using page
rating, section 5 deals with the experimental result and section
6 provides performance of the algorithm and section 7
provides the conclusion. Section 8 contains the future scope of
the research work.
Network Traffic Management using
Dynamic Bandwidth on Demand
P. C. Sethi, P. K. Behera
P. G. Department of Computer Science, Utkal University, Vani Vihar, Bhunaneswar, Odisha, India
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 6, June 2017
369
https://sites.google.com/site/ijcsis/
ISSN 1947-5500

II. LITERATURE OVERVIEW
Network traffic may occur due to the exponential growth
of Internet user and limited availability of various Internet
resources. According to Cisco, the global smartphone traffic
will increase by ten folds by 2019 [figure-1].
[Fig-1: Increase in Traffic Chart by Smart Phone users]
Millions of users relay on various broadband connections.
The ratio of bandwidth supply by different broadband service
providers are represented in figure-2.
[Fig-2: Ration of supply bandwidth by different Broadband
Service Providers]
Traffic engineering (TE) deals with the study of network
traffic analysis and measurement. Efficient routing
mechanisms are proposed by network traffic engineers to
reduce network resource utilization, control network traffic
and enhance network quality of service (QoS). A Software
Defined Network (SDN) is a new technique of traffic
engineering which works in two layers such as forwarding and
controlling layer of network system. The administrator of
system can perform forwarding to enhance the ability of
network system by appropriate task specification. In
comparison with traditional network traffic management, SDN
has many points of interest to bolster TE for globalized, fast
programmability of network system processing.
Data layer traffic and control layer traffic are two
major categories of network traffic which affects performance
of network system. The data layer traffic uses load balancing
concept for network traffic management. Contrasted with the
customary system, the primary favorable advantage of load
balancing in SDN is that it allows a centralized; stream
oriented centralized traffic management instead of a
distributed approach. Network Organization is done for
efficient system accessibility and performance enhancement.
We had proposed a Sensible network traffic management
approach for dynamic data processing, traffic management
and load balancing to enhance QoS of network.
SDN enables network traffic engineers to select
appropriate path out of various available paths between pair of
nodes participating in network. The SDN controller keeps up
worldwide perspective of present utilization of every way in
system utilizing different network traffic parameters. We had
proposed a network traffic management algorithm for dynamic
streaming of data, traffic management, load balancing, and
efficient QoS.
Various dynamic bandwidth allocation techniques are
widely studies in literatures [1-5]. In [1], the authors proposed
a probabilistic sampling approach for efficient traffic flow
control. Flow-based analysis was applied that reduced the high
volume of network traffic by dividing it into flows generators.
Flow-based analysis detection and monitoring of traffic was
applied for distribution of traffic flow uniformly. Since flow-
based analysis provides poor performance, so monitoring
technique was applied for traffic analysis and concluded that
flow-based monitoring technique provides efficient traffic
management than traditional approach.
In [2], the author proposed a method using link
dimension for traffic monitoring. Link dimension is used to
calculate packet level measurement and deploy packet
sampling technique for traffic monitoring. Three packet
sampling techniques such as Bernoulli sampling, n-in-N
sampling and sFlow sampling was done and concluded that
packet sampling have no negative impact based on sampling
rate and packet sampling. The accuracy of system remains
unaffected even for too short timescales such as 10 ms using
large dataset around the world.
In [3], the author proposed a novel technique for precise
and competent stream oriented latency calculation using load
balancing for traffic management. The latency was measured
based on packet size according to the capacity of network
without involving any time stamping or inquiry packets. A
new approach called COLATE (Counter based Perflow
Latency Estimation scheme) was applied that adds noise for
storage space minimization. Using a statistical approach,
packets are denoise to get that actual latency. For secured
implementation, single hashing along with single memory
update was applied in COLATE for each packet. COLATE
utilized less than 0.1 bits per packet. So, the connection can
accommodate nearly million packets per second. Accordingly,
a single 1 TB drive can be used to store timestamp for more
than 6 years COLATE timestamp data connections. Three
types of network traffic traces were considered such as
backbone, enterprise, and server traffic for efficient analysis of
the proposed system.
In [4], the authors proposed an automated network
protocol identification approach for traffic classification. A
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 6, June 2017
370
https://sites.google.com/site/ijcsis/
ISSN 1947-5500

secured semantic trace based information system was applied
for traffic classification. It does not need any prior knowledge
of protocol specification rather frequency rank distribution
concept was applied for management of network traffic. It
supports both connectionless as well as connection oriented
protocols for both short and long flow of data. The average
accuracy of recall is nearly 97.4% with precision nearly
98.4%.
In [5], the authors had proposed an efficient algorithm to
improve network quality of service using dynamic bandwidth
allocation. In a network, each node is assigned with equal
bandwidth which was not utilized properly. Taking the traffic
conditions into consideration, the algorithm was proposed to
provide a guaranteed access and utilize the bandwidth
properly. The resources are allocated following load balancing
condition. In [6], the authors proposed an approach for a
dynamic environment based on clustering. The dynamic
environment was defined as a zero-configuration system i.e.
any type of device can participate in the networking system
using plug-and-play concept for improving the quality of
service. [7, 8] involves a rank based clustering. [7] used click
stream approach for ranking of the pages, accordingly the
clusters are created. The authors guaranteed 100 percent data
transmission, but the time of processing is not considered.
[8,9] provided secured and faster searching approach based
using GFGS (Generalized Frequent Common Gram) technique
by SeLeCT (Self Learning Classifier) following self-seeding
approach on that involves less on-chip memory for processing.
In [10], the authors provided a brief comparison of various
security algorithms. In [11], the authors provided a more
secured approach using RSA algorithm that involves same
processing time but with increased security of data and [12]
contains the description of SeLeCT algorithm.
III. PROPOSED WORK
The data rate reinforced by a network is called
bandwidth. Bandwidth is calculated as difference between
highest and the lowest frequency supported by a network.
Generally, bandwidth is expressed in terms of bits per second
(bps) OR bytes per second (Bps). The theoretical bandwidth
distribution and real-world bandwidth always differs. For
example, theoretically Gigabit Ethernet network supports
1,000 Mbps bandwidth, but in practical this can’t be achieved
due to the overhead of hardware and system software. Hence
bandwidth calculation becomes a challenging task for network
traffic managers. Allocation of bandwidth depends on many
parameters such as type of application running, service level
agreement, hardware performance used for implementation.
Most of the time, network managers only consider number of
user involvement as the major parameter for traffic
management, but instead of number of user involved, the
actual work done by user will affect network performance
during traffic analysis. For example, in a group of 100 users in
network, each user doesn’t utilize network equally; few user
leads to bottleneck problem to the network. So, the traditional
client server distribution of bandwidth will lead to
performance degradation.
In general, each user is assigned with equal bandwidth
irrespective of application. This leads to wastage or
insufficiency of bandwidth. Due to the above reason, a
frequency distribution mechanism is applied which divides the
available bandwidth according to the rating of web page i.e.
the web page which have more rating will be assigned with
higher bandwidth.
3. 1. Bandwidth Computation for Network
Bandwidth for a network can be calculated in two basic
steps:
1. Total available bandwidth calculation.
2. Calculation of required bandwidth for specific
application based on parameter.
If the network is Giga bit Ethernet, then it will support
125,000,000 Bps (considering, 1000 Mbps for a Gigabit
network). Based on number of user and their type of
application, bandwidth needed for each application is to be
determined. According to numbers of bytes transferred per
second, network analyzer detects the bit rate for network.
Cumulative Bytes needed are calculated by the network
analyzer, and then traffic is captured from a test workstation.
According to network traffic generated by each user,
bandwidth is assigned to each user dynamically. Number of
users and type of application will affect the aggregated
efficiency of the system.
The following research work for classification of traffic
is based on three basic fields defined as Classification of
traffic according to user rating for the movieId, SeLeCT
algorithm [12] and bandwidth allocation according to rating.
The Internet traffic clustering is done following rating of
movies of movielens dataset. A Class-Based Weighted Fair
Queuing (CBWFQ) model is considered for clustering of
dataset items as well as bandwidth allocation dynamically.
Required bandwidth differs from network to network and
application to application. A minimum bandwidth called
Minimum Guaranteed Bandwidth (MGB) is initially assigned
to each user. Based on probability distribution of web page
ratings, available bandwidth will be distributed among rest of
users. A queuing system provides a load balancing during
congestion conditions.
The implementation will be done using MovieLens
(http://movielens.org) dataset. MovieLens is an outcome of a
movie reference facility. It contains ratings of movies by
random users between 1-5 according to their preference. The
whole dataset consists of 20000263 ratings with 465564 tag
values for 27278 number of movies. The dataset was created
by collecting ratings of 138493 users between 9th January
1995 and 31st March 2015 and was modified on 17th October
2016. Clients were chosen randomly for consideration. Every
client had evaluated not less than 20 films. No statistical data
is incorporated. Every client is identified by a unique userId,
and no other data was considered.
SeLect algorithm was applied to cluster dataset. SeLeCT
stands for Self-Learning Classifier. It is one of the efficient
algorithms used for Internet Traffic examination. SeLeCT is
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 6, June 2017
371
https://sites.google.com/site/ijcsis/
ISSN 1947-5500

an unsupervised algorithm based on self-seeding approach for
automatic traffic classification. It doesn’t involve any prior
knowledge of environment or grouping of data. It provides
nearly 98% perfectness of traffic classification during network
administration. The data automatically switches between
clusters due to adoptive seeding approach. Based on type of
clusters, prediction for data was done. Development of
information and growth of group size is given in figure–3.
[Figure-3: Movements & enlargements of a window]
The whole process for dynamic assignment of bandwidth
on demand for network traffic management is represented
using following flow chart [figure-4].
[Figure-4: Flowchart DBoD for network traffic management]
IV. PROPOSED ALGORITHM
Calculation of minimum guaranteed bandwidth is too
difficult task for traffic engineers because the theoretical
bandwidth and the actual bandwidth assigned differ. Most of
the time, actual bandwidth allocated is less than theoretical
bandwidth. Considering the minimum guaranteed bandwidth
as constant, the whole research was implemented. The
minimum guaranteed bandwidth (Bmin) can be calculated as:
      
  


Where, αi - the weight factor (rating) for each web page
Tcycle - maximum transmission cycle needed for each web page
N - total number of users accessing the Internet
Tguard - guard time between two consecutive access
R - transmission rate (both upstream and downstream).
The following parameters are variable for individual
user. The algorithm for dynamic bandwidth on demand service
according to rating of web pages is:
Step-1: Calculate available bandwidth and total number of
user (N).
Step-2: Clusters are made according to the weights (ratings)
by CBWFQ following an incremental approach dynamically.
Step-3: Initialized the queue depth for storing various packets
which are to be stored in a cluster earlier to the drops out of
packet (By default, queue depth is set as 64 which is
maximum length of queue).
Step-4: If any traffic doesn’t match with any cluster, then it is
assigned to one of default cluster. When more parameters
(restrictions) are assigned, data will switch to appropriate
cluster. It is maintained using a normal Weighted Fair
Queuing.
Step-5: Find total bandwidth available. Apply Minimum
Guaranteed Bandwidth (MGB) to each user.
Step-6: Calculate excess bandwidth available.
Excess bandwidth = Total available bandwidth – N × MGB
Step-7: Excess Bandwidth to be assign = (Excess bandwidth/
N) × (Current rating/Maximum rating)
Step-8: IF (End of Useri) Then
Release the allocated resources
Goto Step-6
ELSE IF (New User) Then
Allocate Bmin to New User
Goto Step-6
EndIF
Step-9: Stop
V. EXPERIMENTAL RESULT
The algorithm was implemented using MatLab-13 in Intel core
i3 2.20GHz speed processor, 8 GB RAM. Movielens dataset was
used for implementation of proposed algorithm. The pareto
distribution chart for a standard Movielens dataset was
represented in figure-5. It provides standard distribution of
movies based on UserId and Movie rating. Implement of
proposed algorithm was done by considering first 100 users of
data set. 2230 movies information (tuples) are considered for
such implementation.
Though actual implementation was done using 2230
information but for simplicity, first tuple for each user is
represented in the table and graph showing comparison among
required bandwidth and proposed bandwidth in figure-6.
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 6, June 2017
372
https://sites.google.com/site/ijcsis/
ISSN 1947-5500

Figure-6 [Comparison Graph of Required bandwidth and
Proposed bandwidth]
The table showing the comparison among the old and
new bandwidth is assigned for each user represented
(considering the first element of result) is given in table-1.
user
Id
movie
Id
rating
old
Bandwidth
new
Bandwidth
1
2
3.5
0.448833
0.442581
2
367
3.5
0.448833
0.442581
3
480
5
0.448833
0.630139
4
490
5
0.448833
0.630139
6
489
4
0.448833
0.505100
7
494
4
0.448833
0.505100
8
480
5
0.448833
0.630139
9
500
4
0.448833
0.505100
10
356
4
0.448833
0.505100
11
356
3
0.448833
0.380061
12
500
4.5
0.448833
0.567620
13
494
3
0.448833
0.380061
14
500
5
0.448833
0.630139
15
500
2
0.448833
0.255022
16
500
3
0.448833
0.380061
17
356
4
0.448833
0.505100
18
480
3
0.448833
0.380061
19
480
1.5
0.448833
0.192502
20
494
4
0.448833
0.505100
21
474
2
0.448833
0.255022
22
474
4
0.448833
0.505100
23
494
4
0.448833
0.505100
24
493
3
0.448833
0.380061
25
500
4
0.448833
0.505100
26
500
3.5
0.448833
0.442581
27
500
4
0.448833
0.505100
28
491
3
0.448833
0.380061
29
500
3
0.448833
0.380061
30
500
3
0.448833
0.380061
31
466
3.5
0.448833
0.442581
32
364
3
0.448833
0.380061
33
497
4
0.448833
0.505100
34
480
1
0.448833
0.129982
35
500
5
0.448833
0.630139
36
500
4
0.448833
0.505100
37
485
1.5
0.448833
0.192502
38
67
2
0.448833
0.255022
40
376
4
0.448833
0.505100
41
471
1
0.448833
0.129982
42
446
3
0.448833
0.380061
43
355
1
0.448833
0.129982
44
480
2.5
0.448833
0.317541
45
349
4
0.448833
0.505100
46
353
2
0.448833
0.255022
47
500
3
0.448833
0.380061
48
480
3.5
0.448833
0.442581
49
497
4
0.448833
0.505100
50
475
4.5
0.448833
0.567620
51
497
4
0.448833
0.505100
52
463
2
0.448833
0.255022
53
480
5
0.448833
0.630139
54
488
5
0.448833
0.630139
55
494
4
0.448833
0.505100
56
480
4
0.448833
0.505100
57
497
2
0.448833
0.255022
58
480
5
0.448833
0.630139
60
471
3.5
0.448833
0.442581
61
441
4.5
0.448833
0.567620
62
500
3.5
0.448833
0.442581
63
345
4.5
0.448833
0.567620
64
389
3
0.448833
0.380061
66
500
3
0.448833
0.380061
67
480
2
0.448833
0.255022
68
380
4
0.448833
0.505100
69
468
4
0.448833
0.505100
70
500
3
0.448833
0.380061
71
494
1
0.448833
0.129982
72
356
5
0.448833
0.630139
73
500
4
0.448833
0.505100
74
457
3
0.448833
0.380061
75
337
4
0.448833
0.505100
77
500
4
0.448833
0.505100
78
410
3
0.448833
0.380061
79
480
4
0.448833
0.505100
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 6, June 2017
373
https://sites.google.com/site/ijcsis/
ISSN 1947-5500

80
497
4
0.448833
0.505100
81
376
4
0.448833
0.505100
82
480
4
0.448833
0.505100
83
318
5
0.448833
0.630139
84
342
3.5
0.448833
0.442581
85
457
5
0.448833
0.630139
86
262
4.5
0.448833
0.567620
87
318
5
0.448833
0.630139
88
457
5
0.448833
0.630139
89
500
1.5
0.448833
0.192502
90
318
5
0.448833
0.630139
91
500
4
0.448833
0.505100
92
500
4
0.448833
0.505100
93
236
3
0.448833
0.380061
94
500
5
0.448833
0.630139
95
368
3
0.448833
0.380061
96
497
3
0.448833
0.380061
97
480
3.5
0.448833
0.442581
98
500
4
0.448833
0.505100
99
356
5
0.448833
0.630139
100
480
3
0.448833
0.380061
Table-1 [Comparison table of old and new bandwidth]
The graph presented in firure-7 depicts comparison
between the earlier method and the proposed method showing
utilization of bandwidth. This graph contains 2230 tuples of
the dataset. It is evident from the graph that the proposed
method results in optimized utilization of bandwidth.
VI. PERFORMANCE OF ALGORITHM
The minimum guaranteed bandwidth is assigned to each
user who participated in the network, which is essential to
satisfy the basic bandwidth requirement. The calculated excess
bandwidth is distributed among the users having higher rating.
Considering the higher rating items as high demanded web
pages, a frequency distribution technique was applied for
bandwidth on demand (using SeLeCT). By application of this
algorithm, any wastage or insufficiency of bandwidth was
managed effectively. Hence, the efficiency of the overall
system increases significantly as compared to earlier system.
VII. CONCLUSION
Due to the exponential growth of Internet users, the
limited bandwidth has to be utilized efficiently. Growth of
Internet users highly increases traffic over the network, which
puts more responsibility on traffic engineers for controlled
network management. Since each user doesn’t need equal
bandwidth for their application, most of the time there is either
wastage of bandwidth or insufficient bandwidth for
demanding users. Allocating bandwidth on demand based on
rating of web pages provides a solution for controlling
wastage of bandwidth for naive and insufficient bandwidth for
expert users. With the use of proposed algorithm, the total
available bandwidth is distributed dynamically among
different types of Internet users based on their needs. Thus, the
optimized utilization of bandwidth was carried out efficiently
among different users.
VIII. FUTURE WORK
The above research work is implemented using rating of
web pages as parameter. Higher rating web pages are assigned
higher bandwidth. We have not taken into consideration the
priority of user or the vitality of information which can also
play an important role in decision making process about the
dynamic allocation of bandwidth for the rated pages. These
parameters can also be considered for more effective
assignment of bandwidth in a dynamic environment.
REFERENCES
[1] Zahra Jadidi, Vallipuram Muthukkumarasamy, Elankayer Sithirasenan,
and Kalvinder Singh, ―A Probabilistic Sampling Method for Efficient
Flow-based Analysis‖, JOURNAL OF COMMUNICATIONS AND
NETWORKS, VOL. 18, NO. 5, OCTOBER 2016, P-818-825.
[2] Ricardo de Oliveira Schmidt, Ramin Sadre, Anna Sperotto, Hans van
den Berg, and Aiko Pras, ―Impact of Packet Sampling on Link
Dimensioning‖, IEEE TRANSACTIONS ON NETWORK AND
SERVICE MANAGEMENT, VOL. 12, NO. 3, SEPTEMBER 2015,
p392-405
[3] Muhammad Shahzad and Alex X. Liu, ―Accurate and Efficient Per-Flow
Latency Measurement Without Probing and Time Stamping‖,
IEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 24, NO. 6,
DECEMBER 2016, P3477-3492
[4] Xiaochun Yun, Member, IEEE, Yipeng Wang, Member, IEEE,
Yongzheng Zhang, Member, IEEE, and Yu Zhou, Member, IEEE, “A
Semantics-Aware Approach to the Automated Network Protocol
Identification”, IEEE/ACM TRANSACTIONS ON NETWORKING,
VOL. 24, NO. 1, FEBRUARY 2016, P583-595
[5] P.C. Sethi, P.K. Behera, ―An efficient dynamic bandwidth allocation
algorithm for improving the quality of service of networks‖, European
Journal of Academic Essays, Special Issue (1), 2014, P31-35.
[6] P. C. Sethi, ―UPnP and Secure Group Communication Technique for
Zero–configuration Environment construction using Incremental
Clustering‖, International Journal of Engineering Research &
Technology (IJERT), Vol. 2 Issue 12, December – 2013, ISSN: 2278–
0181, pp. 2095–2101
[7] P. C. Sethi, C. Dash: ―High Impact Event Processing using Incremental
Clustering in Unsupervised Feature Space through Genetic algorithm by
Selective Repeat ARQ protocol‖, ICCCT– 2nd IEEE Conference –
2011, pp. 310–315.
[8] P. C. Sethi, P.K. Behera, ―Secure Packet Inspection using Hierarchical
Pattern matching implemented Using Incremental Clustering
Algorithm‖, December–22–24, ICHPCA–2014 (IEEE International
Conference)
[9] P. C. Sethi, P. K. Behera, ―Internet Traffic Classification for Faster and
Secured Network Service‖, International Journal of Computer
Applications (IJCA), Volume 131 – No.4, December2015, pp. 15–20
[10] P. C. Sethi, P. K. Behera, ―Methods of Network Security and Improving
the Quality of Service – A Survey‖, International Journal of Advanced
Research in Computer Science and Software Engineering (IJARCSSE)
Volume 5, Issue 7, July 2015, pp. 1098–1106
[11] P. C. Sethi, P. K. Behera, ―RSA Cryptography Algorithm Using linear
Congruence Class‖, International Journal of Advanced Research (2016),
Volume 4, Issue 5, 1335-1347
[12] Luigi Grimaudo, Marco Mellia, Elena Baralis and Ram Keralapura,
―SeLeCT: Self-Learning Classifier for Internet Traffic‖, IEEE
TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT,
VOL. 11, NO. 2, JUNE 2014 (P144 – P157)
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 6, June 2017
374
https://sites.google.com/site/ijcsis/
ISSN 1947-5500

Er. P. C. Sethi received the B. Tech and M.Tech degrees in
Information Technology Engineering and
Computer Science Engineering from College of
Engineering & Technology, Bhubaneswar. He has
qualified UGC-NET three times in Computer
Science and Applications. He is currently pursuing
PhD in P.G. Department of Computer Science at
Utkal University, Odisha, India. His current
research area of interest is Network Security and
QoS. He has published five research papers in refered international
journals and two IEEE conference paper. He is a life time member of
CSI, ISTE, IAENG, CSTA.
Dr. P. K. Behera is currently working as Reader at Department of
Computer Science, Utkal University,
Bhubaneswar, Odisha, India. He has more than
two decades of teaching experience. His area of
interest is MANET, Wireless Network,
Distributed Systems, Mobile Computing,
Network and Information Security, Software
Engineering. He has published number of
research papers in reputed International
Conferences and Journals. He is a reviewer of many national and
International referred Journals. He is the Secretary of CSI
Bhubaneswar Chapter.
Fig – 5: Pareto Distribution Chart for Movielens dataset
[Fig-7: Comparison Graph of Old and New bandwidth needed]
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 6, June 2017
375
https://sites.google.com/site/ijcsis/
ISSN 1947-5500