Section: New Results

Green Middleware

The energy consumption of ICT is widely acknowledged as continuously growing over years and its carbon footprint can now be compared to the avionics domain. While green computing has emerged as a new research area concerned with the optimization of the energy consumption of large-scale systems, such as datacenters, our project-team investigates the analysis of the energy consumption from a software engineering point of view. In particular, we developed e-Surgeon, a middleware framework to estimate the power consumption of legacy software at various levels of granularity. With respect to this objective, the first result we obtained in [52] relates to an evaluation of the impact of programming languages and programming styles on the energy consumption of applications. While the current trend in application servers is to adopt interpreted languages (e.g., JavaScript, Python) on the server side, our preliminary results highlight that these languages impose a large overhead to the energy consumption of the resulting system. In [53] , this preliminary result is further investigated by identifying energy hotspots within legacy application servers. To do so, we automatically instrument the code of the application server to analyze how the energy is consumed by the application server under various stress scenarios. Our results show that the energy is mostly consumed by a restricted number of classes and methods of these application servers, thus giving hints to software developers on candidate snippets for optimization.