摘要:
Approaches are described for capturing illegal and undesired behavior for network components and for interactions between components. One approach comprise specifying one or more states and state transitions for one or more components or interactions between components, wherein at least one composite state transition is specified, and generating a notification corresponding to a specified state or state transition when the state or state transition occurs. These approaches provide a holistic view of the entire state space of a network or component or component interactions, and allow undesired or illegal states to be captured in a timely manner.
摘要:
An approach is provided for characterizing the dynamic availability behavior of network elements using metrics that indicate patterns of availability of a network element over time. The metrics provide an indication of stability and instability of a network element and include one stability metric and three instability metrics. The stability metric is a stability order that indicates how long an operational state holds with respect to a prior operational state. The instability metrics include an instability order, a repeatability order and a multiplicity order. The instability order indicates how quickly a network element changes operational states. The repeatability order indicates short-term instability within a vicinity of a change in operational state. The multiplicity order indicates long-term instability. The metrics may be used alone, or in various combinations, by network management applications to better manage network elements.
摘要:
A method of managing performance data and accounting data that are generated in a computer network comprises collecting performance data from the network using one or more collector blades that are installed into a data collection manager at a time when the data collection manager is executed; determining whether the performance data satisfies one or more thresholding rules, and if so, generating and sending one or more performance messages; filtering the performance data using one or more filters; aggregating the performance data; and correlating the performance data with other data received from one or more network management subsystems.
摘要:
A uniquely configured data structure is used to store event information for each network entity, where logical and physical dependency relationships among entities are captured in the data structure. For each entity, the data structure is configured to store (a) a “genuine event set”, which includes a list of events that originated in association with the entity; and (b) a “derived event set”, which includes a list of records in which each record is associated with an event that originated in association with an entity that has a dependency relationship (e.g., layering or topological) with the entity. The derived event set may simply comprise references to the genuine event sets for entities that have a dependency relationship with the entity.
摘要:
A method for correlating output of distributed processes includes receiving certain information at a first process of multiple distributed processes that perform one computer application. The information includes a unique identifier for a particular instance of the application that is different from any other identifier for any other instance of the application. The unique identifier is included in data written by the first process to a first data structure. Based on the unique identifier, the data in the first data structure is correlated with data in a different, second data structure written by a different, second process.