Achieving Enhanced Network Performance in UDP and TCP Traffic of Software Defined Networking by Selecting Java Based Controllers Decisively
Pinkey Chauhan1, Mithilesh Atulkar2

1Pinkey Chauhan*, Department of Computer Applications, NIT Raipur, Raipur, C.G.
2Mithilesh Atulkar, Department of Computer Applications, NIT Raipur, Raipur, C.G.
Manuscript received on April 20, 2020. | Revised Manuscript received on April 30, 2020. | Manuscript published on May 10, 2020. | PP: 268-274 | Volume-9 Issue-7, May 2020. | Retrieval Number: G4868059720/2020©BEIESP | DOI: 10.35940/ijitee.G4868.059720
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: Separating out control plane from the forwarding plane is the core idea of Software Defined Networking (SDN). Control plane has the complete decision making logic and the forwarding plane is restricted to just follow the instructions given by control panel i.e. it is concerned with just forwarding the data. This part of decision making in the control plane is looked after by the Controller, so it is the most important component of the SDN. Depending upon the main objective of the created network a wise selection of the controller is important so that it can provide maximum performance for the said objective. Depending upon the need, if a right controller is selected it will give more enhanced performance as compared to a randomly selected controlled which is not made for the said objective. In this paper four Java based controllers namely Beacon, Open Daylight, Floodlight and ONOS have been taken and their suitability has been checked for the performance measuring parameters namely throughput, jitter, packet loss and latency in TCP and UDP traffic of the network. Out of these four controllers, two controllers namely ONOS and Open Daylight are distributed controllers whereas Beacon and Floodlight are centralized controllers. All the controllers have been run in centralized manner one at a time and the performance measuring parameters have been evaluated for TCP and UDP traffic. Logic behind running distributed controllers in centralized manner is to compare their performance with pure centralized controllers if they are run in centralized manner. TCP is a connection oriented protocol and due to its property of error checking, packet retransmission, congestion control mechanism etc measurement of throughput in TCP traffic is more important than measuring it for UDP traffic so throughput has been checked in TCP traffic. UDP is connection less data transmission protocol which is basically used with the applications which send data sporadically or send stream data where small amount of data loss can be accepted. So, two parameters concerned with UPD traffic namely jitter and packet loss have been evaluated with UDP traffic of the network. It has been analysed that Floodlight is performing best for throughput in TCP traffic when the size of the network is increasing. In case of jitter and delay ODL is performing best in UDP traffic. In case of Packet Loss Beacon is best. Mininet has been used as emulator to create the tree networks of different depths. iperf has been used as TCP and UDP traffic generator. 
Keywords: SDN, Openflow, TCP & UDP, Iperf, Mininet.
Scope of the Article: Performance Evaluation of Networks