ORESTES: a Scalable Database-as-a-Service Architecture for Low Latency

Abstract-Today, the applicability of database systems in cloud environments is considerably restricted because of three major problems: I) high network latencies for remote/mobile clients, II) lack of elastic horizontal scalability mechanisms, and III) missing abstraction of storage and data models. In this paper, we propose an architecture, a REST/HTTP protocol and a set of algorithms to solve these problems through a Database-as-a-Service middleware called ORESTES (Objects RESTfully Encapsulated in Standard Formats). ORESTES exposes cloud-hosted NoSQL database systems through a scalable tier of REST servers. These provide database-independent, object-oriented schema design, a client-independent REST-API for database operations, globally distributed caching, cache consistency mechanisms and optimistic ACID transactions. By comparative evaluations we offer empirical evidence that the proposed Database-as-a-Service architecture indeed solves common latency, scalability and abstraction problems encountered in modern cloud-based applications. I. INTRODUCTION The emergence of cloud computing, Database-as-a-Service (DBaaS) and "NoSQL" databases has demonstrated a clear demand for scalable database systems with cloud-capable, webbased interfaces Different studies have shown the dramatic effect of latency on user behavior. For instance, Amazon found that an additional latency of 100ms resulted in 1% less revenue and Google measured that increasing the load time of search results by 500ms decreased user traffic by 20% ORESTES (Objects RESTfully Encapsulated in Standard Formats) is our proposed BaaS/DBaaS architecture to overcome these current limitations of the Backend-as-a-Service model. ORESTES targets the read-intensive, latency-sensitiv