NoSQL Cloud data services provide scalability and high availability properties for web applications but at the same time they sacrifice data consistency. However, many applications cannot afford any data inconsistency. CloudTPS is a scalable transaction manager to allow cloud database services to execute the ACID transactions of web applications, even in the presence of server failures and network partitions. We implement this approach on top of the two main families of scalable data layers: Bigtable and SimpleDB. Performance evaluation on top of HBase (an open-source version of Bigtable) in our local cluster and Amazon SimpleDB in the Amazon cloud shows that our system scales linearly at least up to 40 nodes in our local cluster and 80 nodes in the

Abstract. Cloud computing platforms provide scalability and high availability properties for web applications but they sacrifice data consistency at the same time. However, many applications cannot afford any data inconsistency. We present a scalable transaction manager for cloud database services to execute ACID transactions of web applications, even in the presence of server failures. We demonstrate the scalability of our system using a prototype implementation, and show that it scales linearly to at least 40 nodes sustaining a maximum throughput of 7286 transactions per second. 1

Abstract—The proliferation of Data as a Service (DaaS) available on the Internet and offered by cloud service providers indicates an increasing trend in providing data under Web services in e-science and business domains. While data usage and selectionaredependentondifferentconstraintsestablishedonthe basis of several data concerns, for example, quality of data and data privacy, existing data service engineering approaches lack techniques to allow the evaluation, association and publishing of such concerns with data provided via DaaS. Furthermore, data sources behind DaaSs are not static but dynamically changing, thus requiring the evaluation and publishing of data concerns to be dynamic and on-the-fly as well. In this paper, we present a novel data concern-aware service engineering process for evaluating and publishing data concerns inside DaaS that covers different evaluation and publishing scopes, modes, and integration models. Based on our process, we present a framework and its implementation for the evaluation and publishing of quality of data metrics associated with data provided by DaaSs. In this paper, we also perform several experiments to demonstrate the usefulness of our framework. I.

To my wife Jill, without whose unending support I most certainly would not be here today. Acknowledgements I would like to acknowledge the generous support of my many professors at the University of Texas who have given graciously of their time in support of my education and research over the past two years; especially, Professors Daniel Miranker and Adnan Aziz, who advised on this project; and Professors Christine Julien and Joydeep Ghosh, whose courses and research heavily informed the work herein.

Abstract—NoSQL Cloud data stores provide scalability and high availability properties for web applications, but at the same time they sacrifice data consistency. However, many applications cannot afford any data inconsistency. CloudTPS is a scalable transaction manager which guarantees full ACID properties for multi-item transactions issued by Web applications, even in the presence of server failures and network partitions. We implement this approach on top of the two main families of scalable data layers: Bigtable and SimpleDB. Performance evaluation on top of HBase (an open-source version of Bigtable) in our local cluster and Amazon SimpleDB in the Amazon cloud shows that our system scales linearly at least up to 40 nodes in our local cluster and 80 nodes in the Amazon cloud.

acceptée sur proposition du jury: Prof. B. Faltings, président du jury Prof. K. Aberer, directeur de thèse