RapidLearn: A General Purpose Toolkit for Autonomic Networking

TL;DR

RapidLearn introduces a versatile toolkit that simplifies the deployment of distributed machine learning applications in SDN environments, enabling local decision-making and global coordination for network management tasks.

Contribution

It provides a generic, user-friendly framework for creating and deploying distributed ML applications in SDN, abstracting complex algorithms and structures.

Findings

Effective DDoS detection demonstrated

Framework simplifies ML deployment in SDN

Local and global decision coordination achieved

Abstract

Software Defined Networking has unfolded a new area of opportunity in distributed networking and intelligent networks. There has been a great interest in performing machine learning in distributed setting, exploiting the abstraction of SDN which makes it easier to write complex ML queries on standard control plane. However, most of the research has been made towards specialized problems (security, performance improvement, middlebox management etc) and not towards a generic framework. Also, existing tools and software require specialized knowledge of the algorithm/network to operate or monitor these systems. We built a generic toolkit which abstracts out the underlying structure, algorithms and other intricacies and gives an intuitive way for a common user to create and deploy distributed machine learning network applications. Decisions are made at local level by the switches and communicated to other switches to improve upon these decisions. Finally, a global decision is taken by controller based on another algorithm (in our case voting). We demonstrate efficacy of the framework through a simple DDoS detection algorithm.