Value-oriented network management

The value of a network lies in the business services that it provides. The goal of network management is ensuring the availability of these services. While the physical management task entails examination and repair of physical components, it is impossible to effectively maintain these components without explicit representation of services provided and consumed. We present a layered architecture for managing network value. The heart of this architecture is a service model based on a dependency graph whose nodes represent managed objects and edges represent service relationships. Events generated at lower layers of the architecture identify operational state changes in managed objects. These state changes may give rise to alarms that can be viewed at higher (presentation) layers. States propagate from one managed object to another where the service dependency graph identifies a behavioral dependency. Events and alarms representing impacts of a fault elsewhere correlate to the faulty component, focusing attention on root cause faults. Common semantics for expressing managed object state and service dependency allow the use of shared fault management techniques for all levels of abstraction. The presentation layer employs both linear and planar views, facilitating shifting between narrow preoccupation and contextual exploration. A rich filtering capability facilitates effective presentation to different audiences. This architecture has been partially implemented in the commercial Trinity product