By Rami Bahsoon, School of Computer Science, The University Of Birmingham
Abstract: I report on the activities of the ongoing EPSRC/University of Birmingham Bridging the Gap Fellowship on Green Cloud. The initiative is multidisciplinary; it involves the School of Computer Science, The School of Electrical, Electronic and Computer Engineering, Institute for Energy Research and Policy, the Department of Economics at Birmingham University, external collaborators at MIT, and industrial partners. The initiative is aimed at a framework for dynamic self-optimization of the cloud architecture taking into account the tradeoffs involved in maintaining acceptable dependability requirements with minimal power at runtime. The research pioneers research in dynamic optimization explicating green-aware software architectures and engineering, self-management of the cloud in relation power and dependability. This initiative feeds into the long-term and green-aware vision of helping in reducing power consumption and CO2 emissions in ICT infrastructures.
4 comments
Comments feed for this article
April 18, 2010 at 4:42 pm
Justice
Green Cloud in general is a great concept, though, it should be more clarified what does this term mean in this case. It seems that you aim at building a dynamic self-optimizing architecture. It is not clear what does this optimization apply for. Is it computational power? What does this kind of power mean, is it energy power? How does it relate to supercomputation or cloud computing? Please add a clarification if possible.
It would be good to add some current statistics on the emissions as the motivation and the scale of the problem is not so easy to understand in terms of CO2 generation.
April 19, 2010 at 9:16 pm
Justyna Zander
Sorry for the name confusion. Justyna stands for Justice. Had it as default.
April 21, 2010 at 12:29 pm
Bernd
Cloud computing is certainly a great area for applying “green requirements”. I wonder if you took a look at the area of super computing or computing centers. I guess that a “QoS per power” criteria is already in common use there. Also I miss the reason, why is research is specific for clouds and cannot be applied to other areas, at least for activity 2.1 and 2.2. As the paper is the description of a large and interesting project, it would be helpful to have a table with the participants rather than mentioning them inline.
Minor comments:
- I would rather use the abstract to describe the problem more in detail than listing all project participants
- Only Reference 2 is cited in the paper
April 22, 2010 at 2:45 pm
Ian Sommerville
The general issue of trading off performance and other QoS requirements against energy consumption is a significant one BUT the problem with doing research on this in a university is that you really need access and measurements from one of the major cloud providers. So far, they have not been very open about e.g. their energy usage.
If you do this on a small scale e.g. on a local cloud, then you really will have no idea how it will scale up.
I am also guessing the providers like Amazon are investing in this area (as it will save them money) and are probably putting much more resource into it than is possible in universities.
So, good ideas in this paper but I suspect very difficult to make them work in practice.