Sunteți pe pagina 1din 10

Simulation of QoS Routing Algorithm Based

on SDN Framework

Zhaochenzi Zhang and Yanfang Fu(&)

Xi’an Technological University, Xi’an, China


149481698@qq.com

Abstract. Specific to the increase of Internet dataflow at the present stage and
the higher requirements for service quality (QoS), the Lagrangian relaxation
algorithm is introduced into the SDN framework. In this design, new traffic
scheduling system has been realized in the design, this system takes bandwidth
and time delay as the cost, selects the optimal path. Secondary development of
RYU controller is conducted, and Mininet simulation network topology is used
to verify the system function.

Keywords: QoS routing  Mininet simulation  SDN architecture 


MCMP algorithm

1 Introduction

SDN refers to a software-defined network, belonging to a new type of network inno-


vation architecture, and also a way to realize network architecture, as well as a kind of
implementation model of network virtualization, which adopts OpenFlow technology,
the core technology, separating network equipment control surface from data control
surface, and then the realization of the control of network traffic is able to be available,
make the network intelligent. On account of the continuous development of internet
technology, network environment is gradually developing to complexity, the trans-
mission traffic type presents diversity. Under the complex network environment, tra-
ditional process mode is unable to cope with the multiple transmission flow type, and
can make network architectures difficult to change.

2 Introduction of SDN Network Framework

2.1 SDN Network Framework and Interface

2.2 Application Plane


In this study, the main interaction between the application plane and the control plane is
through the interface, and in the application plane, multiple programming languages
can be used, at the same time, multiple programs and logic can be operated, to improve
SDN’s high adaptability and programming ability, the main way of representation is to
support the application.

© Springer Nature Switzerland AG 2020


Z. Xu et al. (Eds.): CSIA 2019, AISC 928, pp. 469–478, 2020.
https://doi.org/10.1007/978-3-030-15235-2_69
470 Z. Zhang and Y. Fu

2.2.1 Control Plane


The realization of the interaction between the application plane and the control plane is
mainly through the interface-based approach. Controller constitutes the control plane,
the controller can be one or more, and these controllers are responsible for maintaining
the network and continuously implementing the topology, and then routing algorithm is
adopted and passes the calculated data to the switch through the interface [1].

2.2.2 Data Plane


The switch is an important configuration of the data plane. The main function of the
data plane is the management and collection of network link state, which needs to be
realized through the controller and issue OpenFlow message, another function of which
is to forward the data. The switch needs to forward the message according to the
relevant rules issued in the controller, and the message control needs to be realized
through the controller, rather than control by means of network packets transmitting
(Fig. 1).

Application

North Bound
Interface
East West
Bound
Interface
Control Plane

North Bound
Interface

Data Plane

Fig. 1. Classical SDN architecture

2.3 SDN Simulation


2.3.1 Mininet Simulation System
When performing simulation, it is necessary to use Mininet for verification. Mininet is
a kind of simulation tool for SDN network topology and network equipment transfer.
Primarily due to the limited quantity of related resources in ordinary computers, the
large-scale network topology cannot be proceeded. In view of this situation, Mininet
can be used. Mininet can exist in the single host, thus simulating the topology network
type and forwarding function. In addition, Mininet also possesses higher scalability and
supports a variety of software - defined network components, which is also of certain
portability. Therefore, it is widely used in simulation [2].
Simulation of QoS Routing Algorithm Based on SDN Framework 471

2.3.2 Ryu Controller


SDN application layer is divided into three types: SDN management application
manages and controls the SDN frame through managing Restful; SDN cloud appli-
cation manages and controls the network by combining Restful API with the Neutron
component of OpenStack; SDN user application manages and controls the SDN frame
through user-defined Restful API or RPC.
Many applications developed in the Python language are built in the Ryu frame-
work layer, mainly including topology discovery, tenant isolation, firewall and other
components, whose core functions include flow table issue, topology discovery, etc.
(Fig. 2).

Ryu Ryu Ryu Ryu


...
APP1 APP2 APP3 APPn

SERVICE_BRICK:event route event


Hello SwitchFeatures Portstate Echo
Handler handler Handler Handler

event
Ofproto_v1_x_parser

OpenFlowController Response

Datapath
DP DP DP ... DP OpenFlowSwitch
new connecƟon?-> OpenFlowSwitch
new Datapath(socket)
Read&write socket
StreamServer(Gevent) SocketClient
Listen Connect

Fig. 2. Ryu internal operation mechanism

3 Research on Multi-restriction Multi-path Algorithm Based


on SDN Architecture
3.1 Description of MCMP Algorithm
MCMP (Multi Constrained Multi Path) is also known as multi-restriction multi-path
algorithm, when analyzing the MCMP algorithm, the entire network can be regarded as
directed graph, G = (V, E), among which, the letter V mainly represents various stages
in the network, while the letter E represents various links in the network. n = jVj,
m = jEj, these two formulas represent the number of different network nodes and
network links. In normal conditions, the bidirectional transmission means is adopted in
links in the network, so e = (u, v) can represent the link between node u and node v,
and each link has the following features, time delay d(e), bandwidth b(e) [3].
472 Z. Zhang and Y. Fu

The design and application principle of MCMP algorithm is to fully guarantee the
reliability of the network and minimize the impact of the switch failure on the network.
In the actual design process, appropriate path selection can be made on the basis of the
actual situation:
(1) Under the circumstances of the perfect network and the reasonable special
requirements proposed by the user, that is to say, if the network conditions permit,
the MCMP algorithm can be adopted to find the related path that fits the user’s
needs [4].
(2) Under the circumstances that the network situation is poor and there is no path to
meet the user’s requirements, comprehensive consideration is needed, and then in
multiple restricted conditions, [5] the path that best meets the user’s requirements
is selected. When choosing other paths, it is required to ensure that the chosen
path does not intersect with the switch node. The algorithm flow chart is shown
below (Fig. 3).

Fig. 3. MCMP algorithm flow chart

Finally, in order to ensure the path reliability, the improved LBA algorithm should
be used on the basis of the selected links.
Simulation of QoS Routing Algorithm Based on SDN Framework 473

3.2 Path Selection Method


3.2.1 To Select the Path that Fully Meets the Conditions
In the path selection, if the user gives two restricted conditions, maximum time delay
Dmax and minimum bandwidth B, then the path to be searched is:

L ¼ minfcðpÞ : p 2 Pðs; tÞ; dðpÞ \ ¼ Dmax; bðpÞ [ ¼ Bming ð1Þ

Among which, P(s, t) is a set of paths from source node s to destination node t, c(p)
is the total cost of the link.
The idea of Lagrangian relaxation algorithm is adopted in this research. Suppose:

ck ¼ c þ k d ð2Þ

Among which, c = Bmin − b(l), d = d(l) − Dmax, if the calculated results of c and
d of a certain link are less than 0, it indicates that this link does not meet the limit
condition requirements, so the path needs to be abandoned to ensure that both c and d
are positive values [7].

LðkÞ ¼ minfCk; ðpÞ : p 2 Pðs. tÞg  k Dmax ð3Þ

Therefore, for any lower bound that k [ ¼ 0; LðkÞ is L, it is proved that:

L(kÞ ¼ minfckðpÞ : p 2 Pðs. tÞ  k Dmax


¼ minfcðpÞ : p 2 Pðs, tÞ; dðpÞ \ ¼ Dmax, bðpÞ [ ¼ Bming þ k d  k Dmax
¼ L þ k d  k Dmax
¼ L þ k ðd  DmaxÞ
L

To get closer to L, the maximum value of LðkÞ needs to be searched,

L ¼ maxL(kÞðk [ ¼ 0Þ ð4Þ

so in this case, L* is the approximate solution to L.


Making the best of the Lagrangian relaxation algorithm, the optimal k; of the
setting source node and destination node can be found out, and then the best solution
can be obtained.
Observe and analyze whether there are paths in the network that fully meet the
conditions, and make judgment. The main means of judgment is to use only broadband
as the limited condition, and select the path PC, or use time delay as a limited con-
dition, and then choose path Pd. If the path that meets the user’s demand cannot be
selected under the single limited condition, then it is judged that the user satisfaction
path in the current network path does not exist, so the comprehensive routing algorithm
should be adopted [8].
If the only used single parameter routing exists in the network, the path can satisfy
the condition, then it indicates that there may be a path that satisfies both constraint
474 Z. Zhang and Y. Fu

conditions in the network. The Lagrangian relaxation algorithm is employed for the
calculation [9].

3.2.2 Multiple Limiting Conditions Are Integrated to Select the Path


When the network state is poor or the user limited conditions are more, there is no fully
qualified path in the network path, aiming at such circumstance, it is required to
comprehensively consider each limit parameter suggested by the user, and select the
path that best meets the conditions in terms of the condition. In the selecting course,
AHP algorithm is needed to analyze the constraint conditions proposed by users. [10]
AHP algorithm is a kind of decision analysis method, which mainly decomposes the
big target into several small ones, and divides the single level into multiple levels, then
analyzes the effect of realizing the small targets, so as to solve the problem. In the case
of complex network conditions, user demands are no longer simple, and there will
often be situations where user restricted condition and paths do not match. Therefore,
AHP algorithm can be adopted to select the most appropriate path for users.

4 Implementation and Simulation of QoS Routing Algorithm


Based on SDN Architecture

4.1 MCMP Module Design


4.1.1 Data Monitor Module
In the implementation of MCMP algorithm, broadband and delay conditions in the
network should be acquired first, while the link discovery module and topology module
are responsible for connection conditions of multiple switches, so it is unable to probe
the detailed parameter information in the link. Yet when the QoS data monitoring
module collects parameters index, it is by means of issuing OpenFlow messages.

4.1.2 Routing Calculation Module


Routing calculation module is the most core module in the whole experimental plat-
form, whose main function is to write the MCMP algorithm, and in the memory
module, it can also acquire the user’s specific limited conditions and the related
parameter information of the link, and it can also get the network topology in the
topology management module. After obtaining certain data, it will calculate according
to the user’s limited conditions and combined with the topology situation, and then the
path that meets the user’s needs can be selected. Moreover, the routing calculation
module can also launch all the calculation results into the switch [11].

4.1.3 Memory Module


The memory module mainly has two functions, one is to store user restrictions, the
other is to store various information detected in the data monitoring module. The
memory module mainly uses MySQL database to store existing data. In addition, the
routing module in the Floodlight controller is used to obtain information of the user
restrictions, and then link resources. After obtaining the corresponding resources, the
network status information will be read and various data will be calculated [3] (Fig. 4).
Simulation of QoS Routing Algorithm Based on SDN Framework 475

Fig. 4. Schematic diagram of MCMP architecture

5 Experimental Platform Building

When building the experimental platform, the custom topology in the Mininet simu-
lation platform we shall be first built. After the completion of the construction, users
only need to script on the basis of the provided network topology requirements.
As can be seen from Fig. 5, the custom topology generates 9 switches, 8 virtual
hosts, virtual hosts h1, h2, h3 connect with virtual switch s1, virtual host h4 connects
with virtual switch s2, virtual host h5, h6, h7 connect with virtual switch s6, and virtual
host h8 connects with virtual switch s7.
After writing, run the Ryu controller, the current network simulation topological
structure on the Ryu network visualization platform cam be seem, as shown in Fig. 5:
Next, the Ryu controller on the basis of building an automatic network topology is
needed to run. First, the data monitoring module needs to be used to obtain the
corresponding data, and to obtain the data accuracy. Then, the data monitoring module
is used to monitor broadband data and delay data, and the corresponding data is
calculated by issuing the following message. Through experimental analysis, it can be
found that the data monitoring module has a high accuracy of information retrieval, and
the simulation and accuracy of the experiment are obtained on the new basis of data
accuracy.
Finally, the flow table information is required to be obtained. It is necessary to
judge the routing effect of MCMP, set certain parameters for the path that meets the
constraint condition and the path that does not meet the constraint condition simulation
analysis
The parameters shown in Tables 1 and 2 are obtained through experiments:
476 Z. Zhang and Y. Fu

Fig. 5. Mininet network simulation topological graph

Table 1. Original link bandwidth


Bandwidth S1 S2 S3 S4 S5 S6 S7
(Mbps)
S1 – 500 – 400 500 – –
S2 500 – 600 – – – –
S3 – 600 – 500 – – 600
S4 400 – 500 – 500 600 400
S5 500 – – 500 – 400 –
S6 – – – 600 400 – 600
S7 – – 600 400 – 600 –

Table 2. The link bandwidth measured by the data monitoring module


Bandwidth S1 S2 S3 S4 S5 S6 S7
(Mbps)
S1 – 493 – 496 488 – –
S2 493 – 598 – – – –
S3 – 598 – 490 – – 597
S4 406 – 490 – 496 590 396
S5 488 – – 496 – 399 –
S6 – – – 590 399 – 594
S7 – – 597 396 – 594 –
Simulation of QoS Routing Algorithm Based on SDN Framework 477

Comparing the two sets of data, it can be found that the data obtained by the data
monitoring module is slightly smaller than the actual bandwidth of the link. The reason
for this phenomenon is that the maximum bandwidth of the link is adopted in the
process of setting link bandwidth, while the network transmission of data packets in the
actual network topology will occupy a certain network bandwidth, so the measured
value is less than the maximum value.
In accordance with the simulation data, the link state information of the data
monitoring module is basically correct.
According to the experimental platform to obtain the policy data, the algorithm
simulation is strictly conducted based on the path selection method mentioned above.
The analysis to the simulation results is mainly to compare the difference between
MCMP algorithm and the other two algorithms, and the comparison content is the
accuracy of selection path and link recovery time respectively.
(1) Comparison of path selection accuracy: The MCMP algorithm can be used to
select the path that meets the constraint conditions for users who meet the
restricted conditions in the network, while the application of AHP and GA will
lead to instability. The main reason for this situation is that the main goal of AHP
and GA is to select a relatively good path under multiple conditions of compre-
hensive link, but the restriction conditions are neglected, so it is impossible to
select a path that meets users’ needs.
(2) Link recovery time comparison: When the switch breaks down, the effect of the
link disjoint algorithm is the same as that of the conventional passive routing, the
main reason is that the link disjunction algorithm only guarantees the disjoint
between the backup path and transmission path, but does not guarantee the node-
disjoint between the backup path and the main transmission path. While the
MCMP algorithm can ensure that the selected switch node does not intersect the
backup path, so the recovery time is very stable.
This shows that the MCMP algorithm can satisfy various requirements and
restricted conditions of users, and the algorithm path that meets the constraints of users
can be accurately selected for the first time, both in terms of time delay and broadband.
The MCMP algorithm can also solve the problems such as slow link recovery and so
on.

6 Conclusion

Along with the continuous development and progress of SDN network, users’
requirements for the network are gradually increasing. Users hope that the relevant data
can be propagated in the path under certain restrictions. While the SDN network is
characterized by centralized control and flexible management etc., therefore, specific to
such kind of SDN network, a multi-restricted multi-path QoS algorithm, MCMP, is
proposed, and the main objective of this algorithm is to select a satisfactory path for the
user based on the user constraints and network conditions.
478 Z. Zhang and Y. Fu

Acknowledgments. This paper is supported by the industrial research project of Shaanxi sci-
ence and technology department (No. 2016GY-088), the fund of school of computer science and
engineering of Xi’an Technological University (No. GSYSI2016012) and the educational reform
fund of Xi’an Technological University (No. 16JGY25).

References
1. Chen Z (2017) Research and implementation of QoS control technology in SDN network.
University of Electronic Science and Technology of China
2. Gao L (2017) Extensibility study on spatial information network control plane based on
SDN. Beijing University of Posts and Telecommunications
3. Zhan, ZH (2016) Research on data center routing strategy and security authentication based
on SDN. Anhui University
4. Lawson S (2013) Network heavy hitters to pool SDN efforts in OpenDaylight project.
Network World, 8 April 2013. www.networkworld.com/news
5. Open_vSwitch, Open_vSwitch database schema, DB Schema 7. 6. 0
6. Xu K, Liu H, Liu J, Shen M (2011) One more weight is enough: toward the optimal traffic
engineering with OSPF. In: 2011 31st international conference on distributed computing
systems (ICDCS). IEEE, pp 836–846
7. Zhang M, Liu B, Zhang B (2009) Multi-commodity flow traffic engineering with hybrid
MPLS/OSPF routing. In: 2009 IEEE global telecommunications conference, GLOBECOM
2009. IEEE, pp 1–6
8. McKeown N, Anderson T, Balakrishnan H et al (2008) OpenFlow: enabling innovation in
campus networks. ACM SIGCOMM Comput Commun Rev 38(2):69–74
9. Solberg M, Silverman B (2017) OpenStack for architects. Packt Publishing Ltd.
10. Chirammal HD, Mukhedkar P, Vettathu A (2016) Mastering KVM virtualization. Packt
Publishing Ltd.
11. Ivanov K (2017) KVM virtualization cookbook. Packt Publishing Ltd.
12. Subramanian S, Chowdhury CD (2015) OpenStack networking cookbook. Packt Publishing
Ltd.

S-ar putea să vă placă și