Scout: A framework for Querying Networks
Speaker
Andrew Curtis-Black
Institute
PhD Student, Computer Science & Software Engineering, UC
Time & Place
Mon, 05 Oct 2020 15:00:34 NZDT in E16 - Engineering Core
Abstract
There are two kinds of network data: Network telemetry (e.g. packet counters) and business data (e.g. user roles). Existing approaches to querying network data keep these separate, increasing the number and complexity of queries users must write to answer questions about networks. We present Scout, a framework for creating tools which combine these two types of data. It is comprised of: An information model which can represent both network telemetry and business-domain data in use-case-specific schemas; a nascent query language for this information model; and an algorithm for executing queries on schemas. A preliminary evaluation showed that a Scout-based tool can answer questions pertaining to both network telemetry and business data, and reduces the knowledge and number of queries needed to answer realistic questions about networks.
Biography
Andrew is a PhD candidate at the University of Canterbury, from which he graduated with a bachelor of engineering specialising in software engineering in 2015. His research centres on network management technologies which support the daily workflows of network administrators, particularly domain specific languages (DSLs). He enjoys research with strong practical elements, and favour methodologies involving direct communication with interested parties (e.g. discussions with practitioners, interview studies and user trials). He is currently wrapping up a user study of the Scout language which is the topic of this presentation, and is aiming to submit his thesis at the end of this year.