by Sebastian Wätzoldt, Stephan Hildebrandt, Andreas Seibel, Gregor Gabrysiak, Holger Giese
Abstract:
Developing software that really exploits the potential of multi-core and cloud computing is inherently more difficult than traditional software development as the essentially required development of parallel behavior is much more difficult than for sequential behavior. However, multi-core and cloud computing is more about scalability of service-oriented architectures than only performance as in case of parallel computing and instead of a given fixed hardware configuration the possibility to exchange the underling hardware or provider to handle even higher loads is key. We propose to approach these new challenges by a model-driven approach where the higher-level abstraction of the software description enables to derive several optimized platform-specific solutions for different as well as changing hardware settings. In order to ensure that the system operates always with a good solution, the software should be able to adapt itself such that in the spirit of autonomic computing the software takes care of the permanent self-optimization of its execution strategies to ensure scalability. To evaluate different initial static options for our related currently developed self-adaptive model-driven approach, we employed the HPI Future SOC lab as a test bed.
Reference:
Towards Scalable and Self-Optimizing Software for Multi-Core and Cloud Computing (Sebastian Wätzoldt, Stephan Hildebrandt, Andreas Seibel, Gregor Gabrysiak, Holger Giese), Technical report 42, Proceedings of the Fall 2010 Future SOC Lab Day; Universitätsverlag Potsdam, 2011.
Bibtex Entry:
@TechReport{Waetzoldt:2010ta,
AUTHOR = {Wätzoldt, Sebastian and Hildebrandt, Stephan and Seibel, Andreas and Gabrysiak, Gregor and Giese, Holger},
TITLE = {{Towards Scalable and Self-Optimizing Software for Multi-Core and Cloud Computing}},
YEAR = {2011},
MONTH = {February},
NUMBER = {42},
PAGES = {15--19},
INSTITUTION = {Proceedings of the Fall 2010 Future SOC Lab Day; Universitätsverlag Potsdam},
URL = {http://opus.kobv.de/ubp/volltexte/2011/4976/},
PDF = {uploads/pdf/WaetzoldtFutureSOC2010.pdf},
ABSTRACT = {Developing software that really exploits the potential of multi-core and cloud computing is inherently more difficult than traditional software development as the essentially required development of parallel behavior is much more difficult than for sequential behavior. However, multi-core and cloud computing is more about scalability of service-oriented architectures than only performance as in case of parallel computing and instead of a given fixed hardware configuration the possibility to exchange the underling hardware or provider to handle even higher loads is key. We propose to approach these new challenges by a model-driven approach where the higher-level abstraction of the software description enables to derive several optimized platform-specific solutions for different as well as changing hardware settings. In order to ensure that the system operates always with a good solution, the software should be able to adapt itself such that in the spirit of autonomic computing the software takes care of the permanent self-optimization of its execution strategies to ensure scalability. To evaluate different initial static options for our related currently developed self-adaptive model-driven approach, we employed the HPI Future SOC lab as a test bed.}
}