Verfahren und Anordnung zur Überwachung eines verteilten Systems

摘要

A method and apparatus for monitoring the behavior over time of a distributed system (51). Time-stamped data descriptive of events at one subsystem (53) are placed (15) into a local buffer (59). The subsystem is notified (21) of the time when a trap condition occurs at another subsystem (63). Data having a time-stamp within a certain interval of the occurrence of the trap condition are archived (23) to provide a history of the system for later analysis. Time is determined by a local clock (57, 67) in each subsystem. These clocks are synchronized (25) to ensure accurate correlation between events at different subsystems. Trap conditions are categorized (31) and data descriptive of subsystem states are classified (35) to facilitate selective notification (33) of one or more other subsystems, and selective retention (37) of the data, depending on which category of trap condition has occurred and which class of data has been collected.