A Generic Architecture for Scalable and Highly Available Content Serving Applications in the Cloud

The cloud computing paradigm allows service providers to offer scalable and highly available applications to their end users. Typical cases where this is required are content serving applications, where a large number of connected users manage arbitrary data amounts. In the Big Data era, where the amount of information that is being produced and consumed grows exponentially, centralized legacy approaches are inefficient, as they cannot adequately scale according to the number of connected users or the dataset sizes. In these cases, an efficient cloudification of content serving applications is required in order to benefit from the cloud's offerings. In this work, we present a generic architecture that can be used by almost any content serving application in order to offer scalable and highly available data management operations to their users by employing cloud management techniques. We describe the architectural blocks of our approach along with how they can be efficiently deployed in a cloud environment. We document our experiences with an actual deployment of a typical content serving application over ~okeanos, an Openstack compatible public cloud service. We describe the open source frameworks that we have selected from a plethora of existing tools, we justify our choices and we describe our initial observations during their operation. We give a detailed overview of how we installed and configured these systems to achieve high availability and scalability in a public cloud setting. Finally, we document our initial performance evaluation where we showcase the system's ability to handle increasing workloads by elastically scaling its resources.