Quality of service (QoS) requirements, which include availability, integrity, performance and responsiveness are increasingly needed by science and engineering applications. Rising computational demands and data mining present a new challenge in the IT world. As our needs for more processing, research and analysis increase, performance and reliability degrade exponentially. In this paper we present a software system that manages quality of service for Unix based distributed application clusters. Our approach is synthetic and involves intelligent agents that make use of static and dynamic ontologies to monitor, diagnose and correct faults at run time, over a private network. Finally, we provide experimental results from our pilot implementation in a production environment.