Fakultät für Informatik TU München - Fakultät für Informatik
Lehrstuhl III: Datenbanksysteme
Technische Universität München
Home  |  Personen  |  Forschung  |  Lehre  |  Sonstiges  | 

The AutoGlobe Project

Motivation

Complexity and consequently administration costs of IT-infrastructures are ever increasing. To overcome this, adaptive computing infrastructure with fully automatic server and service allocation are needed. The long-term goal is a fully automatic self-management of hardware and software. Comprehensive self-management capabilities for systems include self-configuration, self-optimization, self-healing, and self-protection.
We developed the AutoGlobe research prototype. It virtualizes, pools, and monitors hardware to provide an adaptive and dynamic computing infrastructure which is supervised by a fuzzy logic based controller. This controller, e.g., automatically restarts failed services and detects an overloaded service instance. This way, it remedies the overload situation by starting a new service instance or by moving the service to a more powerful server. Available resources are shared between all services as appropriate for a particular situation. Thus, by dynamically allocating the services, we improve the average utilization of the available hardware and minimize idle times. Thereby, total cost of ownership (TCO) is reduced, because more users can be handled using the existing hardware (or less hardware is required, to begin with).

Research topics

Current status

The AutoGlobe system is based on the ServiceGlobe system and serves as the research platform for future and current research directions. We already have implemented a fuzzy logic based controller and are currently realizing an automatic allocation manager component.

Documents


Lehrstuhl für Datenbanksysteme
Letzte Änderung: 27.08.2007 um 10:10:25