International Journal of Electrical and Computer Engineering (IJECE)

Vol. 7, No. 5, October 2017, pp. 2706~2712

ISSN: 2088-8708, DOI: 10.11591/ijece.v7i5.pp2706-2712  2706

Link Aggregation Control Protocol on Software Defined

Network

I. D. Irawati1, Y. Sun Hariyani2, S. Hadiyoso3

Telkom Applied Science School, Telkom University, Indonesia

Article Info ABSTRACT

Article history: A physical connection of computer network must be made reliably. Breaking
connection will cause communication between nodes (for example routers,
Received Mar 29, 2017 switches, hosts) can be disconnected. One of the solutions is implemention of
Revised May 31, 2017 link aggregation (LA). LA integrates several of physical ports together to
Accepted Aug 11, 2017 make a single logical communication link. Accordingly, there is load sharing
traffic among the member port of the group, high-throughput increasing via a
single link, and redundancy providing for broken links. We present the
Keywords: implementation of link aggregation using Ryu controller on Software
Defined Network (SDN) topology. The results show that the implementation
Link Aggregation, of SDN with OpenvSwitch and Ryu controller can successfully run link
Redudancy, aggregation function to solve the problem of link failure.
Ryu controller,
SDN,
Traffic Copyright © 2017 Institute of Advanced Engineering and Science.
All rights reserved.

Corresponding Author:
I. D. Irawati,
Telkom Applied Science School,
Telkom University,
Telekomunikasi Rd. Terusan Buah Batu,
Bandung 40257 Indonesia.
Email: indrarini@telkomuniversity.ac.id

1. INTRODUCTION
Switch or router over the network [4]. OpenFlow is developed by Open Networking Foundation
(ONF) Today, the performance of computer networks is a great concern by the customers. Network
performance is influenced by the reliability of the physical network topology. When the physical connection
between two nodes is failed, then the packet can not be sent through the link. Otherwise, if a physical
connection is a good condition with limited bandwidth, it can cause congestion since several nodes using the
link for delivery. This problem can be solved by Link Aggregation (LA). LA is a method that aggregates
multiple network connections acts as a single logical interface using software. LA provides redundancy for
link failure, for example: if one or more physical interface on a logical aggregate loss, the other physical
connection still up and operate. Moreover, LA increases throughput passing through a single link which can
reduce bottleneck by allowing packets to be traversed over multiple interfaces. LA is defined in IEEE
802.3ad that allows to group Ethernet interface at the physical layer to form a single link interface, namely
Link Aggregation Group (LAG). Link Aggregation Control Protocol (LACP) is a mechanism for changing
the port to preserve LAG bundles. LACP provides a dynamic configuration that means the other end can
handle link aggregation. In addition, it handles failover automatically when a link failed [1].
Software Defined Network (SDN) is a new architecture for a computer network that separates
network control and forwarding functions, which allowing the controller can connect directly programmable.
SDN’s characteristics are manageable, adaptable, scalable, appropriate for high bandwidth, and dynamic
features of recent applications compared to traditional network architecture [2]. Hence, users can assign the
logical network topology using software, regardless of underlying network structure [3]. In general, SDN

Journal homepage: http://iaesjournal.com/online/index.php/IJECE

IJECE ISSN: 2088-8708  2707

associates with OpenFlow protocol as a communication protocol that remote communication to the
forwarding plane of a network as an organization that promotes networking through SDN and standardizing
the OpenFlow protocol. The OpenFlow architecture consists of OpenFlow controller and OpenFlow Switch.
The function of OpenFlow controller manages OpenFlow switch.
Studies on link redundancy on SDN had been done [5]-[7]. For example, M. Steinbacher et al.
presented LACP implementation using Floodlight OpenFlow controller [5] and P. Skoldstrom et al. proposed
virtual aggregation forwarding model on backbone topology [7]. While I.D. Irawati et al. designed link
redundancy for high availability network based on OpenFlow fast-failover for cascade SDN topology.
Other related research in terms of maintaining quality and network performance presented
in [8]-[9]. Yahya et.al present Extended Dijkstra's with (edges and nodes weights) parameters to find the
shortest server [8]. This algorithm is used for SDN applications. The use of a genetic algorithm load balancer
for service recovery on the intercloud network is done by Jena et al. [9].
In this paper, we present the implementation of IEEE 802.3ad dynamic LACP configuration on
SDN topology. We apply the LACP between a server and a switch using OpenFlow, i.e., Ryu controller.
Ryu manages LACP function to perform failover functionality.

2. BASIC THEORY
2.1. SDN Architecture
Software Defined Network (SDN) is an approach to apply open protocols, such as OpenFlow to
implement software control by the edge of the network to access the entire network devices. The architecture
of SDN is as shown in Figure 1. It composes by three layers, such as infrastructure layer, control layer, and
application layer. An Infrastructure layer represents network device as a switch. Control layer manages both
Southern interface and northern interface. The Southern interface is used to control the state of the network.
While northern interface provides program applications to configure, supervise, and optimize the service
applications [10].

Figure 1. SDN architecture [8]

2.2. Link Aggregation Control Protocol (LACP)

Link aggregation is a technique noted in IEEE 802.1ax-2008 by David Law for local and
metropolitan area network link aggregation standard [11]. And an automatic configuration feature that known
as link aggregation control protocol (LACP) is determined in IEEE 802.3ad [1]. LA combines several
physical interfaces to be as a logical link that useful for increasing bandwidth between network devices and
handling fault tolerance by link redundancy. LA functions are shown in Figure 2 [11].
There are several features of LACP. The maximum quantity of integrated port appropriates with
the channels of the device, usually 1 to 8 channels. LACP packets use multicast MAC address
01:80:c2:00:00:02 for sending packets. LACP provides a load-balance mode that identifies the member of
links which serve load balancing. LACP mode has two states, i.e., active and passive. Active mode enables
LACP unconditionally, while passive mode allows LACP only when an LACP device is detected [12].

Link Aggregation Control Protocol on Software Defined Network (I. D. Irawati)

2708  ISSN: 2088-8708

Increasing Bandwidth Fault Tolerance

200Mbps Link used
100Mbps 100Mbps

100Mbps 100Mbps

(a). LA for increasing bandwidth (b). LA for fault tolerance

Figure 2. Link aggregation function, (a). Increasing bandwidth, (b). Fault tolerance [12]

3. DESIGN AND IMPLEMENTATION

We design and implement LACP according to the Ryu controller using Python programming
language. The topology consists of Controller (C0), Switch (S1), Host 1 (H1) as a server, and Host 2 (H2) as
a client that is shown in Figure 3. Host1 configuration is intended to create a server that supports link
aggregation capabilities. It is configured by using modprobe bonding mode four that indicates the dynamic
LA is performed using LACP. There is two physical link connection between S1 and server H1 via eth0 and
eth1 that run LACP function.

Figure 3. Implementation topology

4. RESULTS AND ANALYSIS

The configuration of link aggregation on server H1 is shown in Figure 4. The LACP setup is
successful. LACP rate is slow with 30-second intervals.

Figure 4. LA configuration on server H1

IJECE Vol. 7, No. 5, October 2017 : 2706 – 2712

IJECE ISSN: 2088-8708  2709

H1-eth0 and H2-eth1 act as link aggregation with up status and each link have 10000 Mbps rate.
Ryu controller runs LACP function to perform fault tolerance functionality. SDN Apps operates
simple_switch_lacp_13 that running on OpenFlow 1.3. When server H1 transmits LACP data unit every 30
seconds, the Switch S1 takes in the LACP data unit from server H1. We can see the process that occurs on
the controller C0 as shown in Figure 5. LACP received status indicates an LACP data unit was received by
the port and LACP sent satus means an LACP data unit was sent by the port. Whereas the slave i/f has just
working state represents a change in state from disable to enable. The state of the LACP data unit time has
changed denotes that the time for communication monitoring was changed. In this event, the standard state is
converted into long timeout periode (0 to 90 seconds).

Figure 5. Ryu controller runs LACP

The information of flow entry on S1 is shown in Figure 6. LACP data unit from server H1 is
received by the Switch S1. Switch S1 responds by sending LACP data unit. The Switch S1 replays by
Packet-In message in case LACP data unit with ethernet type 0x8809 is transmitted from H1-eth0 using the
s1-eth1 input port and the MAC address 00:00:00:00:00:11. Moreover, the Switch S1 replays by the Packet-
In message in case LACP data unit with ethernet type 0x8809 is transmitted from H1-eth1 using the s1-eth2
input port and the MAC address 00:00:00:00:00:21. The switch ports have been able to perform the LACP
functions.

Figure 6. Switch LACP

Link Aggregation Control Protocol on Software Defined Network (I. D. Irawati)

2710  ISSN: 2088-8708

In Figure 7, we test the LACP fault tolerance by breaking the link of H1-eth0. We run the
command ip link set h1-eth0 nomaster. The state of controller C0 is shown in Figure 8. The state of LACP
exchange timeout has occured denotes that there is no communication controlling until elapsed time. In
addition, the satus of slave state changed port: 1 enable: False represents that there is a change in handling of
data transmission from port 2 to port 1. Then, all listed that contained previous MAC addresses and flow
entries will deleted automatically. If the new communication will be recreated, the new flow entries and new
MAC addresses are registered again.

Figure 7. Link broken scenario

The port state changed

into false

Figure 8. Ryu controller is shown link break

In Figure 9 is shown the ping state of H1 to S1 connection. Because of broken link on H1-eth0 to
S1, ping can not be sent from H1 to S1. The time required for recovery is about 90 seconds. The throughput
result is shown in figure 10. The observation is done in 600 seconds. When the event of link failure, the
throughput will decrease until 0 Mbps over 90 seconds. The average of throughput in normal condition is
about 7413.33 Mbps.

IJECE Vol. 7, No. 5, October 2017 : 2706 – 2712

IJECE ISSN: 2088-8708  2711

State before link

broken

recovery
time

The process is
restarted

Figure 9. Ping state observation using Wireshark

Figure 10. Throughput observation in link failure period

5. CONCLUSION
We implement link aggregation successfully using Ryu controller on SDN topology. LACP
provides fault tolerance in the communication monitoring between nodes in the network. The recovery time
is about 90 seconds. This procedure can apply for all connection entire the network. For improving the
performance, we can set the input parameter of LACP according to the network requirements.

Link Aggregation Control Protocol on Software Defined Network (I. D. Irawati)

2712  ISSN: 2088-8708

REFERENCES
[1] IEEE Standard for Ethernet Link Aggregation: IEEE802.3ad, available: http://www.ieee802.org/3/ad/, accessed
at: June 13, 2016, 2016-01-10.
[2] L. Junyi, L. Lin, Z. Yongxin, "Direct Radio Frequency Sampling System on Softwaredefined Radio,”
TELKOMNIKA Indonesian Journal of Electrical Engineering., vol. 12, no. 11, pp. 7824-7831, 2014.
[3] K. Greene. TR10: Software-Defined Networking. MIT Technology Review. 2009; March–April.
[4] N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, L. Peterson, J. Rexford, S. Shenker and J. Turner.
OpenFlow: Enabling Innovation in Campus Networks. ACM SIGCOMM Computer Communication Review.
2008; 38(2): 69-74.
[5] M. Steinbacher and M. Bredel. LACP Meets OpenFlow – Seamless Link Aggregation to OpenFlow Networks.
Available: https://tnc15.terena.org/getfile/1867. Accessed at: January 13, 2016.
[6] I.D. Irawati, S. Hadiyoso and Y. S. Hariyani. Link Redudancy for High Availability Network based on Software
Defined Network. Pertanika Journal of Science & Technology. (under review)
[7] P. Skoldstrom and B. C. Sanchez. Virtual Aggregation using SDN. IEEE Proceeding: European Workshop on
Software Defined Networks. 2013.
[8] W. Yahya, A. Basuki and J.R. Jiang. The Extended Dijkstra’s-based Load Balancing for OpenFlow Network.
International Journal of Electrical and Computer Engineering (IJECE). 2015; 5(2): 289-296.
[9] T. Jena and J. R. Mohanty, Disaster Recovery Services in Intercloud Using Genetic Algorithm Load Balancer.
International Journal of Electrical and Computer Engineering (IJECE). 2016; 6(4); 1828-1838.
[10] Y. Zhou, L. Ruan, L. Xiao and R. Liu. A Method for Load Balancing based on Software-Defined Network.
Advanced Science and Technology Letters. 2014; 45: 43-48.
[11] D. Law, IEEE Standard for Local and Metropolitan Area Network-Link Aggregation. Available:
http://standards.ieee.org/findstds/standard/802.1AX-2008.html. Accessed at: February 10, 2016.
[12] R. P. Team. RYU SDN Framework. Available: https://osrg.github.io/ryu-book/en/Ryubook.pdf. Accessed at:
January 4, 2016.

IJECE Vol. 7, No. 5, October 2017 : 2706 – 2712