A Green Programming Model for Cloud Software Efficiency

Cloud computing is emerging as a methodology for delivering more energy efficient computing provision. The potential advantages are well-known, and are primarily based on the opportunities to achieve economies of scale through resource sharing: in particular, by concentrating data storage and processing within data centers, where energy efficiency and measurement are well established activities. However, this addresses only a part of the overall energy cost of the totality of the cloud: energy is also required to power the networking connections and the end user systems through which access to the data center is provided, and researchers are beginning to recognize this. One further aspect of cloud provision is less well understood: the impact of application software behavior on the overall system’s energy use. This is of particular concern when one considers the current trend towards “off the shelf” applications accessed from application stores. This mass market for complete applications, or code segments which are included within other applications, creates a very real need for that code to be as efficient as possible, since even small inefficiencies when massively duplicated will result in significant energy loss. This position paper identifies this problem in detail, and proposes a support tool which will indicate to software developers the energy efficiency of their software as it is developed. Fundamental to the delivery of any workable solution is the measurement and selection of suitable metrics, we propose appropriate metrics and indicate how they may be derived and applied within our proposed system. Addressing the potential cost of application development is fundamental to achieving energy saving within the cloud – particularly as the application store model gains acceptance.

http://saiconference.com/Conferences/FTC2017