Event monitoring in highly complex hardware systems

This paper presents a novel approach for monitoring disjunct, concurrent op erations in heavily queued systems. A nonobtrusive activity monitor is used as an on-chip tracing unit. For each pending operation the monitor uses th e hardware implementation of an event-triggered operation graph to trace th e path of the operation through the system. In contrast to conventional tra cing units, which collect and record information from one or more functiona l units for later analysis, the presented solution directly records the pat h taken by the operation through the system, making possible an immediate a nalysis of operation inconsistencies. For each followed path a unique signa ture is generated which significantly reduces the amount of trace data to b e stored. The trace information is stored together with a time stamp for de bugging and measuring of queueing effects and timing behavior in the system . The method presented has been successfully applied to the memory-bus adap ter chips in the S/390(R) G5 and G6 systems.