Developing a Pattern Discovery Model for Host Load Data

Investigating the pattern of host load in computing systems is very useful for discovering the data features and predicting the host load in the future. Since the host load can be regarded as the time series data, this paper proposes a pattern discovery framework for host load data by applying time series analysis methods. In the proposed framework, the effective data representation, data segmentation and feature extraction methods are designed based on the characteristics of the host load data. The DBSCAN clustering algorithm is then adopted in the pattern discovery framework to find the patterns in the host load. The extensive experiments have been conducted in this paper to verify the effectiveness of the proposed framework.