Always On: Architecture for High Availability Cloud Applications

With the shift towards cloud computing, application availability becomes a valid concern. Cloud applications have to be taken down to address various hardware and software problems. Even thoug redundancy insulates applications from hardware errors, the software updates and patches still require the service to be taken down for the period of the update. Organizations generally use lowactivity hours to do software updates. This paper proposes architecture for constructing high-availability cloud applications that can be updated without shutting down the service. The paper discusses how the various components of application can be structured into request receiver nodes and request processor nodes. Request receivers receive the request and request processors act on the request. We provide a mechanism of detecting whether any worker or responder node is not working optimally and allow user to save the state and restart this particular node. Finally we present scenarios wherein these nodes can be upgraded either individually or in batches. This reduces dependency on administrators and operations engineers in data center thereby reducing the cost. The reduced cost can directly benefit cost sensitive emerging markets. Furthermore this improves the service level agreement (SLA) since the application will not have to be taken down for updates.