Diagnostic Framework for Distributed Application Performance Anomaly Based on Adaptive Instrumentation

Abstract

Instrumentation technology can obtain the status information of the distributed system when it is running. It is the core part of software performance management tools. But the use of instrumentation technology is often accompanied by a waste of resources. In this paper, we designed an adaptive instrumentation mechanism to balance the monitoring needs and resource load. Using analytical models based on linear regression and K-Means based on density algorithm to analyze performance data, determine the actual operating conditions and monitoring requirements of the monitored system, dynamically change the insertion point, and reduce resource consumption. Experiments show that compared with traditional tools, using the method of this article for monitoring, when the user clicks a lot, the system throughput is lower, the resource load is smaller, the average response time of the tested web page is reduced by 3.37%, and the target program is Interference is smaller.