Download PDF

IEEE/ACM International Conference on Utility and Cloud Computing (UCC) 2015, Date: 2015/12/07 - 2015/12/15, Location: Limassol, Cyprus

Publication date: 2015-12-01
Pages: 78 - 84
ISSN: 9780769556963
Publisher: IEEE

2nd IEEE/ACM International Symposium on Big Data Computing

Author:

Rafique, Ansar
Van Landuyt, Dimitri ; Lagaisse, Bert ; Joosen, Wouter ; Raicu, I ; Rana, O ; Buyya, R

Keywords:

Data management middleware, Multi-cloud storage, Policy-driven, Abstraction API for NoSQL, Multi-tenant SaaS, Science & Technology, Technology, Computer Science, Information Systems, Computer Science, Theory & Methods, Engineering, Electrical & Electronic, Computer Science, Engineering

Abstract:

Multi-tenant Software-as-a-Service (SaaS) applications are increasingly built on combinations of cloud storage technologies and providers in a so-called multi-cloud setup. One advantage is that such a setup helps satisfying the different ---sometimes even contrasting--- storage requirements of different customer organizations (tenants). In such a multi-cloud environment, the application data is distributed and replicated over multiple cloud storage systems, each differing profoundly in supported data models, development APIs, performance, scalability, availability, and durability. Despite the clear benefits, managing such a multi-cloud storage architecture in practice is non-trivial. Addressing this complexity in the application layer is far from ideal, as it inherently limits the flexibility with which continuously changing application-wide and tenant-specific storage requirements can be met. To alleviate this, we present a reusable data management middleware that (i) makes abstraction of multiple cloud storage technologies and thus also providers; (ii) follows a policy-driven approach for making data placement decisions; and (iii) provides tenant customization support, i.e. by allowing tenants to define storage configurations and data storage policies. We validate and evaluate our prototype implementation in the context of a realistic multi-cloud SaaS application. Our performance benchmark results indicate that the benefits of the proposed middleware can be achieved with acceptable overhead.