Data and Resource Aware workload Management as a Service (DREAMaaS)
In this project, we enable Software-as-a-Service (SaaS) providers to utilize multiple cloud computing environments in a seamless way at the infrastructure?level. More specific, we will study, incept, design and implement an architecture that supports static and dynamic allocation of workload and of software component instances between multiple cloud environments. The resulting software system will take into account domain?specific knowledge from the specific application type.
The main driver behind this project is that it is impossible for SaaS oriented software companies to statically choose an operating modus for their service offerings: when and where should they provision extra capacity? How important is computation time and delay vs. cost? What is the impact of storage and network communication on computation time? What performance do we require from the data store? Should one rent capacity from public IaaS cloud providers (a la Amazon) or should one activate more virtual machines in (one of) their own datacenter(s)? These are but a subset of the difficult workload allocation and related scheduling questions that many actors in the ICT market need to address.
DREAMaaS provides an architecture that enables SaaS oriented software companies in dynamically or statically – dependent on the specific context ? choosing the operating modus that suits their needs. The resulting software system architecture automatically adapts this operating modus based on domain?specific and customer specific?knowledge, on monitoring information of running software and infrastructure components and on pricing information from external cloud offerings. In addition to the obvious cost and time parameters, we have identified (1) the dependencies between computation, storage and network communication, and (2) confidentiality of data as critical parameters that need to be considered when automatically adapting the operating modus. The parameters we consider as domain?specific knowledge are acquired from studying and elaborating on the support for three application cases of the industry partners. In terms of these parameters, various levels of sophistication will be incrementally added to a core solution.
DREAMaaS adapts the operating modus of SaaS offerings by creating and removing virtual machine instances. The set of active virtual machine instances can be running in a variety of cloud computing environments: one or more private cloud environments as well as public cloud environments, provided by different IaaS providers.
The project will set the requirements of the DREAMaaS solution based on a detailed study of 3 complementary case studies of applications that will be deployed in a SaaS context. The three industrial partners, Luciad, Noesis and UnifiedPost are innovators and leaders in their respective market segments. It should be noticed that the workload management system that emerges from the DREAMaaS project can be exploited as an intelligent extension of their respective SaaS applications, albeit in different variants. Additional opportunities may emerge for other SaaS providers, potentially also for PaaS vendors. The project result can thus create additional added value and opportunities for many ICT actors in Flanders.
More information about the project can be found at http://www.iminds.be/en/projects/2014/03/06/dreamaas