UC Santa Barbara

Cloud computing has revolutionized the way developers implement and deploy applications. By running applications on large-scale compute infrastructures and programming platforms that are remotely accessible as utility services, cloud computing provides scalability, high availability, and increased user productivity.

Despite the advantages inherent to the cloud computing model, it has also given rise to several software management and maintenance issues. Specifically, cloud platforms do not enforce developer best practices, and other administrative requirements when deploying applications. Cloud platforms also do not facilitate establishing service level objectives (SLOs) on application performance, which are necessary to ensure reliable and consistent operation of applications. Moreover, cloud platforms do not provide adequate support to monitor the performance of deployed applications, and conduct root cause analysis when an application exhibits a performance anomaly.

We employ governance as a methodology to address the above mentioned issues prevalent in cloud platforms. We devise novel governance solutions that achieve administrative conformance, developer best practices, and performance SLOs in the cloud via policy enforcement, SLO prediction, performance anomaly detection and root cause analysis. The proposed solutions are fully automated, and built into the cloud platforms as cloud-native features thereby precluding the application developers from having to implement similar features by themselves. We evaluate our methodology using real world cloud platforms, and show that our solutions are highly effective and efficient.