Title: A load prediction model for SDN controllers
Authors: Kshira Sagar Sahoo; Mayank Tiwary; Bibhudatta Sahoo
Addresses: Department of Computer Science, National Institute of Technology, Rourkela, Odisha, India ' Digital Business Services, Clod Product Support, SAP Labs, Bangalore, India ' Department of Computer Science, National Institute of Technology, Rourkela, Odisha, India
Abstract: Separation of the control plane from data plane and keeping the decision making activities into a centralised system called controller is the key idea behind software defined networks. The controllers take the control decisions for OpenFlow enabled devices from a higher abstraction of the network. In this work, we propose a control layer architecture, which considers the local area network and wide area network traffic independently. The objective of this work is to predict the load for WAN controllers through the LAN controller by considering PACKET_IN as the load. After a certain threshold value, the controller's response time for processing the PACKET_INs increases substantially because of excess CPU utilisation. The delay in response from controller affects the initial packets of the flow passing through the network device and thus affects the QoS of the network. Hence, the proposed technique addresses this problem and reduces the delay in response substantially.
Keywords: software defined networks; OpenFlow devices; load prediction; controller.
International Journal of Autonomic Computing, 2017 Vol.2 No.4, pp.384 - 400
Received: 05 Dec 2016
Accepted: 20 Nov 2017
Published online: 07 Feb 2018 *