Implementing a Resilient Application Architecture for State Management on a PaaS Cloud

Abstract

Platform as a Service Clouds typically lack direct support for application state management, and traditional state management techniques like clustering are not applicable as PaaS platforms offer little support for changing the underlying platform configuration. In this paper we build upon our earlier work where we proposed a session-state management architecture for Cloud called ReLoC, that uses loosely-coupled services and platform agnostic scalable messaging technology to propagate and save session states. Here, we present an actual implementation of the ReLoC onto a PaaS platform and an empirical evaluation of the original hypotheses of scalability and resilience of the proposed application architecture. We also present the challenges faced in implementing ReLoC on Heroku. The results indicate that ReLoC indeed allows applications to scale well and mitigates failures in individual application instances while maintaining state and hiding such failures from the users. The results also indicate that the performance degradation due to use of ReLoC is minimal and it is thus a promising approach for resilient user state management on PaaS Clouds.