An Approach to Deal with Processing Surges in Cloud Computing

摘要

Processing surges are fast and unexpected changes in the processing demand that commonly occur in cloud computing. The cloud elasticity enables to handle processing surges, increasing and decreasing resources as required. However, a surge can be very fast, so that the overhead to provide more resource is greater than the processing benefit. On the other hand, if the surge is slow and continuous, and the required resources are not provided, the application performance may be impaired or interrupted. This paper presents a machine learning-based approach to detect and classify processing surges, in order to improve the cloud resource management, minimizing losses for the application and cloud provider. We use a real cloud dataset to select features, to construct the classifier and to test our approach, which successfully detected and classified 99% of the processing surges.