Cloudy transactions: cooperative XML authoring on Amazon S3

Abstract

Over the last few years cloud computing has received great attention in the research community. Customers are entitled to rent infrastructure, storage, and even software in form of services. This way they just have to pay for the actual use of these components or services. Cloud computing also comes with great opportunities for distributed design applications, which often require multiple users to work cooperatively on shared data. In order to enable cooperation, strict consistency is necessary. However, cloud storage services often provide only eventual consistency. In this paper, we propose a system that allows for strict consistent and cooperative XML authoring in distributed environments based on Amazon S3. Our solution makes use of local and distributed transactions, which are synchronized in an optimistic fashion, in order to ensure correctness. An important contribution of this paper is the evaluation of our system in a real deployment scenario upon Amazon S3. We show the strong impact of write operations to S3 on the transaction throughput. Furthermore, we show that fragmenting data increases write performance and reduces storage costs.