Dr Paul Anderson – University of Edinburgh
Describing System Configurations with L3
System configuration languages are now ubiquitous. Almost all major installations and services depend on specifications written in languages such as “Puppet” to deploy and manage the underlying infrastructure. But these languages are almost always developed in an ad-hoc way, which makes them very difficult to verify or reason about, and their complexity makes them difficult to use correctly. The resulting configuration errors are a frequent cause of system failures.
I will give a little background on declarative configuration languages in general, and discuss some of the characteristics of the system configuration task which typically cause difficulties in practice. I will then give an overview of our work on configuration languages, before describing some current work on an experimental language (L3) which is intended to be a simple language with a clear semantics providing primitive operations specifically suited to configuration tasks.
Paul Anderson has a background in the practical management and configuration of large computing infrastructures, working both with industry, educational institutions, and large research installations such as CERN. He is currently a Senior Research Fellow with the School of Informatics where he is interested in applying knowledge from various areas in the School to the practical problems of large-scale system configuration. He is particularly interested in configuration languages, including provenance, security, and usability – but also in the deployment of configurations including automated planning, and agent-based approaches to distributed management
School of Computing Science & Digital Media, Robert Gordon University, Sir Ian Wood Building, Garthdee, Aberdeen, Conference Room N118, 13:00 – 14:00.