By Govindaraj Rangaraj and Rami Bahsoon, The University of Birmingham, Birmingham, United Kingdom

Excerpt: Software systems architects are continually faced with the challenge of scaling up software systems architectures to support constantly growing load of users’ processing needs and data. Scaling up the architectures to meet these needs does certainly introduce additional energy cost.   For  example, to meet  the scalability requirements,  additional  hardware and  software resources  may  need  to  be  deployed.    Reducing the energy demands in  such  architectures while  meeting the  scalability requirements, are always  challenging. We explicate the attention to power  as  an  architectural  constraint/property that need  to  be analyzed in relation with scalability. Current research and practice to distributed software architecture approaches are green-unaware. They don’t provide the primitives for reasoning and managing power consumption. We  argue  that the software  engineering should  be green aware,  where  the  software  engineering and design  activities should  not only  be  judged  by  their  technical merits,  but also by their  contributions to energy  savings.  In particular, the software system architecture appears to be the appropriate level of abstraction to address green-aware concerns. Software architectures should be green-aware, providing power management mechanisms as part of the architecture primitives. Furthermore, it looks plausible to leverage on advances in self-management software architectures [2], where self-managing power  could  be separated from  the core system functionalities. We  argue  that there  is a pragmatic need  for new  software architectural layer,  which  could  be  easily  integrated with existing  styles for self-managing the trade-offs between  scalability and  power. The power consumption can be minimized by only provisioning the required amount of resource at any given point of time. For example, architecture can be scaled only when the demand for the resource increases. Classical market-based economic theory is appealing for addressing this problem effectively in the context of supply/demand.

Full paper here