International Conference on Sustainable Technologies for

Industry 4.0 (STI), 24-25 December, Dhaka

Enhancing Quality of Service in SDN based on

Multi-path Routing Optimization with DFS
Md. Sajid Hossen˚ , Md. Habibur Rahman˚: , Md. Al - Mustanjid˚ , Md. Arif Shakil Nobin˚ , Md. Ahsan Habib:
˚ Department
of Software Engineering
Daffodil International University, Bangladesh
: Department of Information and Communication Technology

Mawlana Bhashani Science and Technology University, Bangladesh

mdsajidsetu@[Link], [Link]@[Link], {[Link], arifshakil28}@[Link], [Link]@[Link]

Abstract—Remarkable growth of network traffic has been (data and control) as per the SDN unplugging perception [5].
noticed recently due to the use of online applications and With Openflow protocol and Ryu controller, we implemented
cloud services. Thus, multiple routing problems need to solve two multipath algorithms BFS and DFS in SDN.
without increasing latency to optimize the traffic. Based on the
responses from industry, Software Defined Networking (SDN) Both types of network applications and their usage are
has been developed to meet this challenge. SDN allows the rising due to extensive internet expansion [6]. The particular
programmable features of network and responds dynamically. overview of a network connection’s efficiency is called QoS.
In this paper, multipath routing problem in SDN is addressed QoS can be assessed quantitatively within two choices, one
and resolved under OpenFlow protocol where Ryu controllers
are used to implement SDN using the Mininet WiFi emulator. is the user connection’s performance and another one is a
Besides, in addition to a graphical user interface, a module has collection of quantitative target parameters that classify the
been developed using python programming along with OpenFlow one in the network [7]. QoS is used in SDN to monitor
based Ryu controller. In the simulation results, depth first search network bandwidth, latency and transmission [8]. Single path
(DFS) algorithm outperforms over breadth first search (BFS) in routing produces congestion in the network as it forwards
terms of round trip time (RTT), delay and throughput. Hence,
this article enhanced the quality of service (QoS) in SDN. network traffic using single best-path. Multipath routing is
Index Terms—SDN, Multipath routing, BFS, DFS, Mininet a replacement way to distribute traffic between a network’s
WiFi, QoS, 4G, 5G, Optimization paths. Multipath routing allows traffic from one source to
another with some possible best-paths in a network, which
I. I NTRODUCTION eliminates network disturbances [9].
Multipath networks have been researched widely to en-
SDN is the fastest-growing network infrastructure, which hance throughput and reduce traffic in network. It is very
is highly efficient, scalable, and service-worthy. Up to now, hard to acquire equal-distance multipath networks because of
3G communication is well achieved in traditional network. scheduled cross layer network assets and complex network
In contrast, the technologies of 4G and 5G communications status monitoring scheduling processes such as RTT, band-
are SDN, network function virtualization (NFV). Meanwhile, widths [10]. Several shortest path algorithms like Dijkstra’s,
sustainable industry 4.0 is a portion of the fourth industrial Bellman-Ford, A* search, Floyd–Warshall, BFS and DFS are
transformation that becomes a disquiet for the industry. In implemented for detecting the shortest path in the traditional
order to create such a contribution in industry 4.0, we focus network system. To enrich the QoS in SDN, the shortest
to enhance the QoS in SDN with multi-path routing, which is path algorithm has been implemented in [11]. Dijkstra’s has
still a challenging issue in an era of industry 4.0, Internet of been already implemented in the SDN [12]. Further, the best
Things (IoT) etc. [1] [2]. The conventional network consists shortest path in multipath SDN networks will ultimately attain
of two stages in network devices called control and data or the optimization [13] [14]. However, DFS is not implemented
forwarding stage [3]. yet in SDN to find out the best fitted shortest path. Hence, we
SDN differs from traditional network by unplugging the implemented the DFS algorithm to improve the packet transfer
control stage from the data transmission stage. In general, rate.
the SDN network consists of three phases, namely application
Thus, in this article, the major contributions are summarized
phase, controller phase and phase of network or infrastructure,
as follows:
as shown in Fig. 1. SDN enables the network phase to be
programmed. Centralizing devices actually referred to as con- 1) Implementation of DFS attain the less RTT than BFS
trollers manage all routing activities. In SDN controller where for shortest path selection in SDN.
the southbound APIs interface is located to manage overall 2) DFS provides low jitter for optimized path.
decisions in the network, which can be incorporated via the 3) While implementing DFS the use of OpenFlow protocol
OpenFlow protocol [4]. Openflow also entitled SDN control and Ryu controller makes the SDN network faster than
protocol rationalizes the data exchange between two phases traditional network.

978-1-7281-6099-3/19/$31.00 c 2019 IEEE

 Fig. 1: The architecture of Software Defined Network

The paper is outlined as follows. Section II described the to acquire the necessary network activity information to make
related work based on SDN and multipath routing optimization optimum routing assumptions in data center. An SDN adaptive
research activities held recently. Section III discusses the worst-fit multi-path routing (AWMR) algorithm for data center
proposed QoS enabled multi-path routing. Section IV explains networks introduced in [24]. It was discovered and contained
about QoS evaluation and simulation results. Finally, Section an initial shortest length routing pathway collection from the
VI concludes the paper. origin and destination hosts. It also estimated the bandwidth
II. BACKGROUND AND RELATED WORK utilization of this kind of pathway collection. Again, it applied
an enhanced broadest disjoint path model and choose the
OpenFlow [15] and broad-ranging work of SDN [16] in- worst-fit routes based on the current-coming flow’s required
spired us before diving into design our process. In this section, bandwidth.
we discuss a subset of this relevant efforts.
From the above analysis, multipath routing optimization in
Authors explained in [17] how SDN leads to better network
SDN is merely seen in local area network. Thus, this article
management and a more stable network. While there was
enhanced the QoS in SDN in terms of optimum distance in
an explicit concentration on OpenFlow in the data center
multipath routing with DFS algorithm. Finally, compared the
[18], the attention on WAN was relatively low. Our emphasis
DFS along with BFS algorithm performed by RTT.
on the WAN is centered on the WAN’s containment and
quality together with the projected growth pace. Many works
explored the emergence of issues with OpenFlow scalability III. Q O S ENABLED MULTI - PATH ROUTING
[19]. Openflow protocol distinguishes the data process from
the control process and offers a customized flexible transfer This section describes the implementation of the multipath
phase. Moreover, the advantages of SDN architecture utilizing routing problem optimization in SDN. The OpenFlow network
Openflow were defined by researchers in [20]. emulator named Mininet and the OpenFlow controller named
Authors integrate SDN and OpenFlow to facilitate a network Ryu framework have been used during the implementation.
system that solves two problems faced by a network manager Mininet WiFi is used as a network simulator for implementing
[21]. Dutra et al. [13] proposed an alternative to achieve the SDN networks. It can provide virtual switches, controllers,
end-to-end QoS relying on Openflow queue management, access points and other networking devices. With a lightweight
which enables an operator with an SDN-enabled network virtual environment, it makes the framework comparable to
to conveniently distribute network resources to the user’s the real world. OpenFlow protocol 1.3 and the Ryu controller
requirements that reduce oversupply needs. E. Chemeritskiy are implemented by the python programming. For creating
and R. Smelansky suggested a procedure in [7] that relies on the virtual network topology, python API is used. To test
multi-path routing to preserve QoS of connections in SDN the minimum path selection, BFS and DFS algorithms are
networks. Researchers in [22] have also built an Openflow implemented in multipath routing in a small scale virtual SDN
protocol-based solution that can implement multi-path routing network.
to allow the available capacity of the network to be used and In the end, a comparative analysis was conducted to find
reap the benefits of separated services to ensure bandwidth. Jo out the best fitted shortest path. In contrast, QoS is attained
et al. [23] studied multipath routing for fat tree network which in SDN by best fitted shortest path as DFS optimizes the
is actually deployed in data center networks. Authors have de- minimum distance. Step by step implementation procedures
signed an algorithm in SDN which centers on a key controller are described in the following subsections.
 h1 to [Link]. Consecutively, [Link] is for h2 and then so
on. The controller is assigned the IP [Link] with the port
address 6633.

Fig. 2: Local Area Network architecture of SDN

Fig. 4: OpenFlow Version Selection
A. Environment Setup
An Intel Core-i5 along with 4GB RAM based workstation C. DFS Implementation
is chosen for practicing the network setup. Ubuntu 18.04 LTE
To find the shortest path selection algorithm, DFS is im-
64-bit operating system is used for implementing the local area
plemented to assist with the Ryu framework. Using iterative
network architecture in the SDN.
method, DFS can find the deepest vertex in a graph using a
The implementation and simulation have been performed
stack.
using python 2.7.15 by installing the Anaconda package. Then,
Algorithm 1: DFS based shortest path selection
rapid prototyping emulator named Mininet WiFi has been
1 DFS(Graph G, Vertex V) (V is the start vertex)
installed for creating the virtual SDN environment. To create
2 Stack St := ; (start with an empty stack)
and visualize the virtual network, graphical user interface
3 for each vertex uV, set visited[uV] := false;
based editor namely MiniEdit has been installed. Finally,
4 push St, V;
enabling the OpenFlow controller, Ryu framework has been
5 while (St is not empty) do
installed and performed the simulation.
6 uV := pop St;
B. Network Configuration 7 if (not visited[uV]) then
A virtual local area network topology depicts in Fig. 2 has 8 visited[uV] := true;
been created using MiniEdit for SDN network. The created 9 for each unvisited neighbor wV of uV
topology contains 5 hosts, 8 switches and 1 controller. The 10 push St, wV;
controller inside the topology has been selected as a remote 11 end if
controller (Fig. 3). Alike TCP protocol, OpenFlow supported 12 end while
OpenVswitch version 1.3 has been selected (Fig. 4). For the 13 END DFS()
distance calculation, edge weight has been assigned 1.
For multipath routing, DFS stores the depth in the stack
following descending order. From that stack, the shortest path
is placed in the last position. Because of this special feature,
DFS has been used for finding the shortest path by exploring
all possible vertices with minimum distance. For example,
in Fig. 2 host h1 wants to communicate with host h5, it
tries to find out the deepest path and put the value in the
stack. It repeats the procedure for all the possible connected
paths. If, such path does not exist in the stack it predicts the
destination is unreachable. From the stack, the shortest path
is selected for the time optimization. The DFS pseudo code
is implemented in [25] as follows.
Fig. 3: Ryu Controller Configuration To explore the shortest path BFS is implemented as [26].
As example in Fig. 2 BFS finds a shortest path from host h2
The controller is defined as c0 , while switches and hosts are to host h4, it checks if the weights of all edges in a first level
named as si and hi respectively. Where i is the integer variable on the path are greater than defined value. If satisfied the
defined as the number starting from 1. Hosts are allocated for condition it returns this path as the shortest path. Otherwise,
 TABLE I: Shortest path selection with minimum distance
Source Destination All-Possible Path (Switches) Shortest-Path Minimum
(Host) (Host) (Switches) Distance

1 2 [1,6,7,5], [1,2,3,4,5], [1,8,5] [1,8,5] 2

1 3 [1,8],[1,6,7,5,8], [1,2,3,4,5,8] [1,8] 1
1 4 [1,2],[1,6,7,5,4,3,2], [1,8,5,4,3,2] [1,2] 1
1 5 [1,6,7,5,4],[1,2,3,4], [1,8,5,4] [1,2,3,4] 3
2 3 [5,8],[5,7,6,1,8], [5,4,3,2,1,8] [5,8] 1
2 4 [5,4,3,2],[6,7,6,1,2], [5,8,1,2] [5,4,3,2] 3
2 5 [5,4], [5,7,6,1,2,3,4], [5,8,1,2,3,4] [5,4] 1
3 4 [8,5,7,6,1,2],[8,5,4,3,2], [8,1,2], [8,1,6,7,5,4,3,2] [8,1,2] 2
3 5 [8,5,4], [8,5,7,6,1,2,3,4], [8,1,6,7,5,4], [8,1,2,3,4] [8,5,4] 2
4 5 [2,3,4], [2,1,6,7,5,4], [2,1,8,5,4] [2,3,4] 2

it continues the calculations until it finds another shortest B. Minimum Distance Calculation
path. If, such path does not exist the host h2 can not get the The DFS algorithm returns the list of paths that are un-
connected path from host h4. The BFS pseudo code has been weighted. So, for calculating the minimum distance of a path, a
implemented as follows. bucket weights formula has been used. This equation has been
described in [27]. Open shortest path first (OSPF) distance
Algorithm 2: BFS based shortest path selection estimation technique is used to estimate the minimum distance.
1 BFS (Graph G, Source Node S) The calculation of the total distance from all the connected
2 let Q be queue. paths are shown in Equation (2) [25]. Here, weight of the
3 [Link](S) bucket (WoB) for a path p is assumed as
4 mark S as visited.
5 while (Q is not empty) 0 ď W oBppq ă 10 (1)
6 V = [Link]( )
and weight of the path (WoP) is as OSPF distance style and
7 for all neighbours W of V in Graph G
N is the total number of available paths.
8 if W is not visited ÿ
9 [Link](W) W oBppq “ p1 ´ pW oP { W oP ppqrN sqq ˆ 10 (2)
10 mark W as visited.
11 END BFS() C. Simulation Results and Discussion
To evaluate the shortest path for every pair of source and
IV. S IMULATION AND PERFORMANCE ANALYSIS destination, the simulation has been executed several times.
To get the result from the simulated workstation, we have Based on the simulation results the detailed traversing infor-
done ping request within all possible hosts. Available paths are mation including the minimum distance are tabulated in table
shown in terminal from every source to destination pair and I. RTT, throughput, and delay are the important parameters
vice versa. Several times simulations are tracked to analyze of QoS. Among all the QoS parameters, our implemented
the outcome behaviors. algorithm has been achieved the minimum RTT hence the
lower delay and the higher throughput. RTT, throughput, and
A. Path Selection delay has been calculated following the equation (3), (4) and
Based on the DFS algorithm the path has been pushed on (5) [28].
the stack. Then the paths are shown in the terminal. While
RT T “ p1 ´ αq ˆ P reviousRT T ` α ˆ P resentRT T (3)
pinging from h1 to h2 for a cross check validation, the list
of all possible available paths are shown in Fig. 5. We have Where, α = Constant weighting factor ( 0 ď α ă 1)
P acket Size
T hroughput “ (4)
RT T
Round T rip T ime
Delay “ (5)
Bnadwidth
Fig. 5: Path Selection among h1 and h2 (h1 is connected in The shortest path selection among multiple paths is shown
s1 and h2 is connected in s5) in Table I. The source and destination path is indicated by the
switches. From the above table, the optimum path is listed in
calculated the bucket weight and distance from h1 to h2 and the rightmost column.
vice versa. We have found that the same path has been selected In the Fig. 6, the X axis indicates the pairwise pinging
in both cases but in reverse order. The host h1 starts with test and the Y axis indicates the RTT in millisecond. From
switch number 1 while h2 linked with switch number 5. the above analysis, it has been observed that among all the
 [8] A. O. Adedayo and B. Twala, “Qos functionality in software defined
network,” in 2017 International Conference on Information and Com-
munication Technology Convergence (ICTC). IEEE, 2017, pp. 693–699.
[9] M. F. Ramdhani, S. N. Hertiana, and B. Dirgantara, “Multipath routing
with load balancing and admission control in software-defined network-
ing (sdn),” in 2016 4th International Conference on Information and
Communication Technology (ICoICT). IEEE, 2016, pp. 1–6.
[10] Q. Wang, G. Shou, Y. Liu, Y. Hu, Z. Guo, and W. Chang, “Implementa-
tion of multipath network virtualization with sdn and nfv,” IEEE Access,
vol. 6, pp. 32 460–32 470, 2018.
[11] S. Tomovic, N. Prasad, and I. Radusinovic, “Sdn control framework
for qos provisioning,” in 2014 22nd Telecommunications Forum Telfor
(TELFOR). IEEE, 2014, pp. 111–114.
[12] J.-R. Jiang, H.-W. Huang, J.-H. Liao, and S.-Y. Chen, “Extending
dijkstra’s shortest path algorithm for software defined networking,” in
The 16th Asia-Pacific Network Operations and Management Symposium.
IEEE, 2014, pp. 1–4.
[13] D. L. C. Dutra, M. Bagaa, T. Taleb, and K. Samdanis, “Ensuring
end-to-end qos based on multi-paths routing using sdn technology,”
in GLOBECOM 2017-2017 IEEE Global Communications Conference.
IEEE, 2017, pp. 1–6.
[14] S. U. Baek, H. H. Shin, and D.-R. Shin, “Consideration of multipath
Fig. 6: Comparison of round trip time between DFS and BFS routing over sdn-enabled network,” in the International Conference
on Innovations in Intelligent Systems and Computing Technologies
in SDN based on pairwise pinging test (ICIISCT2016), 2016, pp. 34–38.
[15] N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, L. Peterson,
J. Rexford, S. Shenker, and J. Turner, “Openflow: enabling innovation in
pairwise pinging test the RTT of DFS attain less than the campus networks,” ACM SIGCOMM Computer Communication Review,
vol. 38, no. 2, pp. 69–74, 2008.
BFS. Suchwise, using equation (3), (4) and (5) three of the [16] N. Gude, T. Koponen, J. Pettit, B. Pfaff, M. Casado, N. McKeown, and
QoS parameters has been enhanced. S. Shenker, “Nox: towards an operating system for networks,” ACM
SIGCOMM Computer Communication Review, vol. 38, no. 3, pp. 105–
V. C ONCLUSION 110, 2008.
[17] H. Kim and N. Feamster, “Improving network management with soft-
In this paper, we have implemented DFS algorithm in SDN ware defined networking,” IEEE Communications Magazine, vol. 51,
to optimize the routing distance in multipath routing compared no. 2, pp. 114–119, 2013.
with BFS. Based on the shortest path, the packet transmitted [18] M. Al-Fares, A. Loukissas, and A. Vahdat, “A scalable, commodity data
center network architecture,” in ACM SIGCOMM Computer Communi-
from one node to another node within smallest RTT. Thus, the cation Review, vol. 38, no. 4. ACM, 2008, pp. 63–74.
RTT and delay were minimized and throughput is enhanced [19] A. R. Curtis, J. C. Mogul, J. Tourrilhes, P. Yalagandula, P. Sharma, and
using DFS rather than BFS. OpenFlow as a protocol and Ryu S. Banerjee, “Devoflow: Scaling flow management for high-performance
networks,” in ACM SIGCOMM Computer Communication Review,
as a controller are used to attain the goal. Our simulated results vol. 41, no. 4. ACM, 2011, pp. 254–265.
validate the QoS enhancement for the SDN architecture at the [20] W. Braun and M. Menth, “Software-defined networking using openflow:
local area network. Protocols, applications and architectural design choices,” Future Inter-
net, vol. 6, no. 2, pp. 302–336, 2014.
R EFERENCES [21] A. Lara, A. Kolasani, and B. Ramamurthy, “Simplifying network man-
agement using software defined networking and openflow,” in 2012 IEEE
[1] L. D. Xu, E. L. Xu, and L. Li, “Industry 4.0: state of the art and future International Conference on Advanced Networks and Telecommuncia-
trends,” International Journal of Production Research, vol. 56, no. 8, tions Systems (ANTS). IEEE, 2012, pp. 24–29.
pp. 2941–2962, 2018. [22] J. Yan, H. Zhang, Q. Shuai, B. Liu, and X. Guo, “Hiqos: An sdn-based
[2] M. A. Habib, S. Saha, M. A. Razzaque, M. Mamun-or Rashid, multipath qos solution,” China Communications, vol. 12, no. 5, pp. 123–
G. Fortino, and M. M. Hassan, “Starfish routing for sensor networks 133, 2015.
with mobile sink,” Journal of Network and Computer Applications, vol. [23] E. Jo, D. Pan, J. Liu, and L. Butler, “A simulation and emulation study
123, pp. 11–22, 2018. of sdn-based multipath routing for fat-tree data center networks,” in
[3] C.-C. Lo, P.-Y. Wu, and Y.-H. Kuo, “Flow entry conflict detection Proceedings of the Winter Simulation Conference 2014. IEEE, 2014,
scheme for software-defined network,” in 2015 International Telecom- pp. 3072–3083.
munication Networks and Applications Conference (ITNAC). IEEE, [24] Y.-C. Lei, K. Wang, and Y.-H. Hsu, “Multipath routing in sdn-based
2015, pp. 220–225. data center networks,” in 2015 European Conference on Networks and
[4] C. N. Shivayogimath and N. U. Reddy, “Performance analysis of a Communications (EuCNC). IEEE, 2015, pp. 365–369.
software defined network using mininet,” in Artificial Intelligence and [25] M. Hua and J. Pei, “Probabilistic path queries in road networks:
Evolutionary Computations in Engineering Systems. Springer, 2016, traffic uncertainty aware path selection,” in Proceedings of the 13th
pp. 391–398. International Conference on Extending Database Technology. ACM,
[5] B. A. A. Nunes, M. Mendonca, X.-N. Nguyen, K. Obraczka, and 2010, pp. 347–358.
T. Turletti, “A survey of software-defined networking: Past, present, and [26] Y. Chen and N. Nasser, “Energy-balancing multipath routing protocol
future of programmable networks,” IEEE Communications Surveys & for wireless sensor networks,” in Proceedings of the 3rd international
Tutorials, vol. 16, no. 3, pp. 1617–1634, 2014. conference on Quality of service in heterogeneous wired/wireless net-
[6] M. Karakus and A. Durresi, “Quality of service (qos) in software works. ACM, 2006, p. 21.
defined networking (sdn): A survey,” Journal of Network and Computer [27] T. Cheocherngngarn, H. Jin, J. Andrian, D. Pan, and J. Liu, “Depth-first
Applications, vol. 80, pp. 200–218, 2017. worst-fit search based multipath routing for data center networks,” in
[7] E. Chemeritskiy and R. Smelansky, “On qos management in sdn by 2012 IEEE Global Communications Conference (GLOBECOM). IEEE,
multipath routing,” in 2014 International Science and Technology Con- 2012, pp. 2821–2826.
ference (Modern Networking Technologies)(MoNeTeC). IEEE, 2014, [28] D. Medhi and K. Ramasamy, Network routing: algorithms, protocols,
pp. 1–6. and architectures. Morgan Kaufmann, 2017.