Engineering Enterprise Networks with SDN

Agarwal, N (2014) Engineering Enterprise Networks with SDN. Masters thesis, Indian Institute of Technology, Hyderabad.

Preview

Text
CS12M1007-M.Tech.pdf - Submitted Version
Download (1MB)

Abstract

Today’s networks are growing in terms of bandwidth, number of devices, variety of applications, and various front-end and back-end technologies. Current network architecture is not sufficient for scaling, managing and monitoring them. In this thesis, we explore SDN to address scalability and monitoring issue in growing networks such as IITH campus network. SDN architecture separates the control plane and data plane of a networking device. SDN provides a single control plane (or centralized way) to configure, manage and monitor them more effectively. Scalability of Ethernet is a known issue where communication is disturbed by a large number of nodes in a single broadcast domain. This thesis proposes Extensible Transparent Filter (ETF) for Ethernet using SDN. ETF suppresses broadcast traffic in a broadcast domain by forwarding the broadcast packet to only selected port of a switch through which the target host of that packet is reachable. ETF maintains both consistent functionality and backward compatibility with existing protocols that work with broadcast of a packet. Nowadays, flow-level details of network traffic are the major requirements of many network monitoring applications such as anomaly detection, traffic accounting etc. Packet sampling based solutions (such as NetFlow) provide flow-level details of network traffic. However, they are inad- equate for several monitoring applications. This thesis proposes Network Monitor (NetMon) for OpenFlow networks, which includes the implementation of a few flow-based metrics to determine the state of the network and a Device Logger. NetMon uses a push-based approach to achieve its goals with complete flow-level details. NetMon determines the fraction of useful flows for each host in the network. It calculates out-degree and in-degree based on the IP address, for each hosts in the network. NetMon classifies the host as a client, server or peer-to-peer node, based on the number of source ports and active flows. Device Logger records the device (MAC address and IP address) and its location (Switch DPID and Port No). Device Logger helps to identify owners (devices) of an IP address within a particular time period. This thesis also discusses the practical deployment and operation of SDN. A small SDN network has been deployed in IIT Hyderabad campus. Both, ETF and NetMon are functional in the SDN network. ETF and NetMon were developed using Floodlight which is an open source SDN controller. ETF and NetMon improve scalability and monitoring of enterprise networks as an enhancement to existing networks using SDN.

[error in script]

IITH Creators: