PInfer: Learning to Infer Concurrent Request Paths from System Kernel Events

摘要: Operating system kernel-level tracers are popularly used in the post-development stage by black-box approaches. By inferring service request processing paths from kernel events, these approaches enabled diagnosis and performance management that application-logic aware. However, asynchronous communications multi-threading behaviors make path patterns dynamic on event level, this causes previous methods to focus either software instrumentation techniques or better statistical inference models. In paper, we propose a novel learning based approach called PInfer infers automatically with high precision. first learns of inter-thread intra-thread training data sequential requests. On testing containing concurrent requests, individual effectively solving graph matching problem generalized assignment learned patterns. We have implemented our proprietary tool, present results 40 sets traces. achieves average 65% precision 85% recall for profiling paths.