Monitoring and control of distributed systems
Young-Chul Shim, C. V. Ramamoorthy · 2002
Describes the Distributed System Manager (DSM) being developed on Sun workstations connected by an Ethernet at the University of California. Effective utilization of distributed systems requires many management facilities such as resource management, performance analysis, behavior analysis, fault management, and security management. Each of the management facilities becomes a client to the DSM and registers to the DSM a request of the form (assertion, action) if it wants the action to be taken when the assertion is detected. The DSM monitors the distributed system to collect information using sampling probes and tracing probes and stores the collected information in a history database. The DSM also detects the moment when the assertion in a registered request becomes satisfied and then takes the corresponding action. An assertion can be a system property holding or an event occurring at a time point or during a time interval. An action can be answering clients' queries to system status stored in the history database, notifying the detection of assertions, or exercising direct control over the system. The overall structure of the managed system and the DSM is given. Data modeling of the managed system, the mechanisms to collect information, and the representation scheme of the information in the history database are described. The assertion specification language based on the classical temporal logic is presented.>