Predicting response time-related quality-of-service outages of paas cloud applications by machine learning

Abstract

For customers running their applications on Platform-as-a-Service (PaaS) cloud environments it is important to ensure the Quality-of-Service (QoS) of their applications. Knowing in advance if and when a potential problem is likely to occur allows the application owner to take appropriate countermeasures. Therefore, predictive analytics using machine learning could allow to be alerted in advance about potential upcoming QoS outages. In this context, mainly Infrastructure-as-a-Service (IaaS) or Software-as-a-Service (SaaS) have been studied in the literature so far. Studies about predicting QoS outages for the Platform-as-a-Service (PaaS) service model are sparse. Therefore, in this paper an approach for predicting response-time-related QoS outages of web services running in a PaaS cloud environment is presented. The proposed solution uses the open source Apache Spark platform in combination with MLib and binary classification by the naive Bayes algorithm. The approach is evaluated by using test data from a social app backend web service. The results indicate that it is feasible in practice.