IDENTIFICATION OF SOFTWARE PHASES USING MACHINE LEARNING

摘要

Technologies for application performance analysis include a computing device capable of executing an application having a number of phases. The computing device collects performance data indicative of a number of performance attributes of the computing device during execution of the application. The computing device may merge data collected from a processor performance monitoring unit with data collected from an operating system of the computing device. The computing device partitions the performance data into a number of clusters using an unsupervised clustering algorithm such as K-means clustering, forming a cluster model. The computing device may partition the performance data into a number of cluster models, apply a classification algorithm such as regularized discriminant analysis to each cluster model, and select the cluster model with the lowest misclassification rate. The computing associates each cluster of the cluster model to a phase of the software application. Other embodiments are described and claimed.

主权项

1. A computing device for software application performance analysis, the computing device comprising:
a performance monitoring module to collect performance data indicative of a plurality of performance attributes of the computing device during execution of a software application, wherein the performance data includes a plurality of data samples, each data sample including a plurality of values corresponding to the performance attributes; a data analysis module to partition the performance data into a plurality of clusters using an unsupervised clustering algorithm, wherein each cluster includes a number of data samples of the performance data; and a phase analysis module to associate each cluster of the plurality of clusters to a phase of the software application.