By Hasan Sözer, Arjan De Roo, and Mehmet Akşit, University of Twente, Enschede, The Netherlands
Abstract: Energy consumption has become one of the important system properties that should be controlled by embedded control software. There is usually an inherent trade-off between energy consumption and several system qualities. As such, optimization techniques should be adopted for making the desired trade-off among quality attributes. Implementations of these techniques are usually ad-hoc and system-specific, leading to implicit design decisions distributed over several software components. We propose an architectural framework for custom synthesis of control software from reusable and programmable elements. The goal is to facilitate systematic reuse of knowledge in the optimization domain and explicit management of quality trade-offs for energy optimization of embedded systems.
7 comments
Comments feed for this article
April 19, 2010 at 2:31 pm
Rami Bahsoon
Interesting position – here are some thoughts…
I wonder how complex is it to incorporate such optimisations, tradeoffs analysis, and conflict resolutions in architectures for embedded systems? How can you guarantee that the payoff of such analysis will exceed the s power/energy savings? I assume that the architectural synthesis will take place at runtime as a result we need a lighter solution….
You mentioned: “we propose a flexible, architectural solution in terms of a framework that comprises reusable solutions for energy optimization in embedded systems.”
Calrify what is meant by reusbale solutions in the context of embedded systems and/or architectures for such systems?
You mentioned: “Architectural styles and a uniform notation that
enables the instantiation and specification of software
architectures”.
Do we need new styles? or shall we extend existing?
April 22, 2010 at 12:09 pm
Ian Sommerville
This really struck me as a paper about power management in embedded systems which, of course, is an issue but the focus of the paper was not about climate change. Better power management, of course, helps but embedded systems are really only minor contributors to carbon emissions. Space heating, transport and manufacturing are the big areas and these are the ones we should focus on
April 23, 2010 at 5:10 pm
Sophia Drossopoulou
I think that reduction of energy consumption is directly related to climate issues, so I thibk that the paper is very relevant.
However, I am not at all clear how approaches to multi-objective optimization can be applied to software and architectures. As far as I know, we have only one objective is software, which is functional correctness. We do sometimes measure efficiency, but I know of no approach tries to model things like the interplay of speed, space requirements and energy consumption. Is there such work?
April 24, 2010 at 10:19 pm
Mehmet Aksit
We know in practice that functional correctness is essential but there may be many different implementations of software that are correct but display different non-functional properties. In fact in industrial applications, these non-functional propoerties can be the main distinguishing characteristic. This paper defines optimization algorithms on the accurate physical models which have been tested in actual field tests by the company which manufacturs these machines. This is a control architecture, so the multi-objective optimization is controlling the machine. This is a known concept in control theory called optimal control;
The paper is relevant as this topic was identified at the OOPSLA workshop last year as one of the main areas of research; Since software is controlling many machines, energy consumption appears to be one of the non-functional properties that must be considered. For example, only the energy consumption of servers in the world is considerable. if you also add all the equipments controlled by software to this; the importance of this issue becomes obvious. However, it is always a treade-off which is difficult to predetermine at design time by hand. Therefore optimal control is needed.
This is an architectural framework bacause, the company builds different kinds of machines with possibly different characteristics. Therefore you need to adjust to different characteristics. Since this is an architectural framework, basic algorithms and models can be reused and adjusted. Otherwise you need to write different algorithms again and again. This cannot be addressed by parameterization only; you need to have a framework based approach, in the same way as we adopt application frameworks in practice.
Since the workshop demanded very short papers these details can be shown during the workshop together with current implementation details.
April 26, 2010 at 3:35 pm
Ioannis Athanasiadis
The paper considers energy consumption optimization with respect to software controllers for embedded systems, and reports authors experience from multi-objective optimization in the field of printing systems.
Authors present an abstract architecture and key requirements for a software system that will assist to the complete and systematic reuse of reusable programming artifacts, to be composed optimally wrt to energy consumption.
While authors make reasonable assumptions, and present interesting ideas, I am afraid that this paper is indirectly relevant to the workshop goals. The impact of software methods for climate change issues is very implicit.
April 26, 2010 at 4:41 pm
Mehmet Aksit
The statement that:
The impact of software methods for climate change issues is very implicit.
is NOT true.
The software architecture here is controlling a machine. These machines consume energy. Only the energy consumption of servers today is equivalent to the energy consumption of air traffic. If you add to this software controlled machines, you reach a significant percentage. Smart software can reduce energy consumption considerably.
April 28, 2010 at 5:54 pm
Ideas for the workshop program « Software Research and Climate Change
[...] role of architecture in energy management / energy efficiency (inspired by Hasan Sozer et al’s paper, and subsequent [...]