A Monitoring and Analysis Framework to Support Self-management in Cloud Application Platforms

Abstract

Cloud computing in general and its Platform-as-a-Service segment in particular are becoming the de facto way of running enterprise-level software. By providing hardware infrastructure and support for software development, cloud application platforms exempt companies from major up-front investments, thus saving time and money. Additionally, they offer developers a range of generic reusable services, ready to be integrated into users' applications. However, to support such a flexible software development model, cloud providers have to exercise constant control over all critical activities taking place on the platform so as to prevent millions of deployed applications coupled with hundreds of add-on services from quickly dissolving into tangled and unreliable environments. To address this challenge, our approach combines techniques from the domains of Semantic Sensor Web, Stream Processing and Big Data analytics to create a monitoring and analysis framework and support autonomy in cloud application platforms. The approach relies on annotating monitored heterogeneous values with uniform semantic descriptions and applying run-time formal reasoning to these data streams so as to detect and diagnose critical situations.