NoSQL application data stores often combine on-premise and public cloud resources. As a consequence, they may deal with heterogeneous technologies, flexible but less structured storage architectures, and dynamic changes. In this paper, the authors propose PERSIST, a middleware architecture that facilitates different storage and privacy-related requirements for federated cloud storage and associated software-as-a-service (SaaS) applications, while also supporting dynamic reconfiguration.

The paper starts with a brief introduction to NoSQL database systems and associated issues in a multi-resource, multi-tenant context. The next section describes the motivation for this work, that is, to treat multi-tenant log management-as-a-service (LMaaS) and manage federated cloud storage architectures. It identifies key challenges: (a) decoupling complexities, particularly the abstraction of federated cloud storage and externalizing the federated cloud storage logic; (b) addressing the dynamic confidentiality requirements of tenants, including data storage; (c) tenant self-service/customization; (d) back-end portability; and (e) practical feasibility. A description of PERSIST follows, including design principles, tenant configuration, managing the SaaS application, multi-tenancy, and data management. Sections 4 through 6 are on prototype implementation, evaluation, and related work, and sections 7 and 8 cover “Discussion” and “Conclusion and Future Directions.” The authors claim “the acceptability of the performance overhead.”

The paper has 50 references. Its audience includes researchers interested in federated and data architectures and the multi-cloud portability of SaaS.