CSR--PDOS: Autonomic Systems: Integrating Machine Learning with Computer Systems

University Of Texas At Austin, Austin TX

Investigators

Abstract

Computer systems are becoming victims of their own success. Deployed software has orders of magnitude more users, software environments, and usage patterns than it has in testing labs, which leads to two fundamental problems. First, software complexity accounts for a large fraction of system cost---the cost of administering, configuring, updating, performance tuning, and supporting systems is surpassing the cost of the hardware. Second, software complexity becomes a first-order limit on key system properties like reliability, performance, and usability. The goals of this work are to build prototype software systems that integrate machine learning to simplify operation, avoid performance problems, and make efficient use of computer resources. For distributed systems like web services the key goals are avoiding performance pitfalls and making efficient use of computing resources. Software support addresses the problem of configuring software and fixing it or the environment when something goes wrong. Autonomic network diagnosis requires the system to identify and correlate events to reach a diagnosis for the state of the network. The work will address the security problems of information leakage that arise when mutually distrustful parties share machine learning models, as they might for software support (software vendor and user) and network diagnosis (competing network providers).