Distributed computation of the critical path from execution traces

Due to the ever-increasing number of computer nodes in distributed systems, efficient and effective tools have become crucial for their analysis. Although several methods been proposed monitor profile tracing remains most solution in-depth system Tracing is act collecting a trace, which sequence low-level events generated by kernel or userspace. After data collection, important part event The paradigm choice graphs determine ability user detect abnormal behaviors identify root cause. highly approach analyzing complex scalability current analysis limited. As consequence, often impractical large systems. This paper identifies shortcomings approaches, notably critical path computation trace file transfer between nodes. Then, this proposes new solutions these drawbacks, algorithm compute path, that does not aggregate all traces single node, an architecture perform on These are made publically available.