A Density Based Performance Prediction Model for Cloud Services

Performance prediction for cloud services, as the fundamental of effective resource provision and throughput, is known to be very challenging. In practice, the lack of source code and fast response requirement invalid most existing approaches. A few studies targeting at cloud services indicate a rising trend of utilizing powerful machine learning techniques such as regression and neural networks into cloud service performance prediction. In this work, we propose a density based performance prediction model specially tailored for cloud services. This model makes full use of the advantage of machine learning techniques and provides a method to predict the performance of cloud services using historical information instead of source code. Experiments verify the feasibility of our proposed method in terms of both accuracy and efficiency.