Peter Kunszt, ETH Zürich - SystemsX.chEnabling eScience
Mon, Aug. 26, 14:30 - 15:30
Aula, KIT Campus North, FTU
By today, especially in the natural sciences, computers have become indispensible tools and instruments for research. Recently, due to progress in digital measurement technology, researchers acquire vast amounts of data in ALL domains of science. Not only the amount of data, but also its complexity is continuously increasing. And to top it off, the data needs to be shared within large scientific collaborations among many people located at various research institutions all across the globe. The term eScience has been coined in the early 2000s to describe research that heavily relies on computational technologies - today it is applicable to almost all sciences.
However, the new organisational and engineering tasks to deal with the new complexity cannot be expected to be managed by the researchers alone. There is the need for a new kind of IT, engineering and research support that focuses on the methods and technologies enabling eScience. This includes large-scale data management, data storage, data policies and data lifecycle considerations, as well as automated validation, processing and analysis of data. It also includes the creation and usage of large-scale infrastructures for eScience, like distributed Grid or Cloud infrastructures. And it requires domain-specific know-how, for example in bioinformatics, chemoinformatics, medical informatics, etc. Ultimately is all about integration of several layers of infrastructure and software so that the scientists can extract the necessary information from large amounts of raw data, to be used by to propose new insights and theories, and to plan the next round of experimentation and observation.
I will show based on the example of the SystemsX.ch SyBIT project and the Sloan Digital Sky Survey how modern eScience operates, what kinds of problems already have reasonable solutions and where the current challenges are.
Peter Kunszt is leading the SystemsX.ch SyBIT project in Switzerland, providing support to the Systems Biology research community, counting well over 1000 scientists. SyBIT builds sustainable tools and services for the life sciences in close collaboration with the researchers. Previously, Peter Kunszt has been in charge of setting up the LHC Tier2 center at the Swiss Supercomputing Center CSCS in Lugano and has helped establish the Swiss National Grid Initiative. Before that, ha has led the development efforts of the data management components in the initial EU Grid projects at CERN, heading the corresponding work packages in the EU DataGrid and the EGEE projects. Peter Kunszt also has a builder status in the Sloan Digital Sky Survey, where he has spent three years developing the SDSS Science Archive, still in use today to deliver astronomical data to researchers through the virtual observatory.
Bob Jones (CERN) on behalf of the Helix Nebula initiative
Fri, Aug. 30, 10:50 - 11:50
Aula, KIT Campus North, FTU
Helix Nebula – the Science Cloud: a public-private partnership building a multidisciplinary cloud platform for data intensive science
The feasibility of using commercial cloud services for scientific research is of great interest to research organisations such as CERN, ESA and EMBL, to the suppliers of cloud-based services and to the national and European funding agencies. Through the Helix Nebula - the Science Cloud  initiative and with the support of the European Commission, these stakeholders are driving a two year pilot-phase during which procurement processes and governance issues for a framework of public/private partnership will be appraised. Three initial flagship use cases from high energy physics, molecular biology and earth-observation are being used to validate the approach, enable a cost-benefit analysis to be undertaken and prepare the next stage of the Science Cloud Strategic Plan  to be developed and approved.
The power of Helix Nebula lies in a shared set of services for initially 3 very different sciences each supporting a global community and thus building a common e-Science platform. CERN is exploring how commercial cloud services could serve its high energy physics experiments  while EMBL is developing a portal for cloud-supported analysis of large and complex genomes. This will facilitate genomic assembly and annotation, allowing a deeper insight into evolution and biodiversity across a range of organisms . ESA is developing the Geohazard Supersites project to advance scientific understanding of the physical processes which control earthquakes and volcanic eruptions as well as those driving tectonics and Earth surface dynamics..
The work of Helix Nebula and its recent architecture model  has shown that is it technically feasible to allow publicly funded infrastructures, such as EGI  and GEANT , to interoperate with commercial cloud services. Such hybrid systems are in the interest of the existing users of publicly funded infrastructures and funding agencies because they will provide “freedom of choice” over the type of computing resources to be consumed and the manner in which they can be obtained.
But to offer such freedom-of choice across a spectrum of suppliers, various issues such as intellectual property, legal responsibility, service quality agreements and related issues need to be addressed. Investigating these issues is one of the goals of the Helix Nebula initiative.
The next generation of researchers will put aside the historical categorisation of research as a neatly defined set of disciplines and integrate the data from different sources and instruments into complex models that are as applicable to earth observation or biomedicine as they are to high-energy physics. This aggregation of datasets and development of new models will accelerate scientific development but will only be possible if the issues of data intensive science described above are addressed. The culture of science has the possibility to develop with the availability of Helix Nebula as a “Science Cloud” because:
- Large scale datasets from many disciplines will be accessible
- Scientists and others will be able to develop and contribute open source tools to expand the set of services available
- Collaboration of scientists will take place around the on-demand availability of data, tools and services
- Cross-domain research will advance at a faster pace due to the availability of a common platform
Bob Jones is head of the CERN openlab project (openlab.cern.ch) which facilitates collaboration between CERN and its industrial partners to study and develop data-intensive solutions for scientists working at the next-generation Large Hadron Collider (LHC). Bob is a leader of the Helix Nebula – the Science Cloud initiative (http://www.helix-nebula.eu/), a public private partnership to explore the use of commercial cloud services for science applications supported by the EC under grant 312301. His experience in the distributed computing arena includes mandates as the technical director and then project director of the European Commission co-financed EGEE projects (2004-2010 http://www.eu-egee.org), which established and operated a production grid facility for e-Science spanning 300 sites across 48 countries for more than 12,000 researchers.
Marek Bundzel, Department of Cybernetics and Artificial Intelligence, Technical University Kosice
Evening Lecture: Nature Inspired Computing
Tue, Aug. 27, 19:00 - 20:30
Campus North, FTU, Aula
Computers - the high point of technology. Our omnipresent slaves and sometimes masters. But thousands of years before the first vacuum tube lit up biological computing machines existed that would outmatch our contemporary silicon companions in nearly every aspect. If in doubt just try to build a machine doing what a simple ordinary house fly does. Soon you will realize, that this simple creature processes and integrates large amounts of various sensory data, infers decisions to sustain its life and adapts to the environment. How?
In the talk we will go through some theory and practice of computing methods inspired by the nature.Despite we are slowly becoming capable to build the computing hardware of the desired complexity we often fail to program that hardware to our liking. Partially, that's why we want the hardware to learn by itself. We will see how artificial neural networks are build and used. We will also see how it is possible to find solutions to complex problems by means of simulated evolutionary optimization. And finally, the memory-prediction framework a progressive new theory trying to explain how the mammalian brain works will be presented. The practical examples to be shown include land use categorization using artificial neural networks, evolutionary optimization of a mechanical structure, object identification in sequences of images using a method based on memory-prediction framework and some more.
Marek Bundzel, PhD. is the head of the Center for Applied Cybernetics at the Department of Cybernetics and Artificial Intelligence, Technical University Kosice, Slovakia. His research is oriented on the methods of computational intelligence: artificial neural networks, support vector machines, evolutionary optimization and robotics. Marek Bundzel has spent two years at Waseda University, Tokyo, where he was researching the potential of modeling the brain for the purposes of visual object identification performed on a mobile robot. He enjoys building stuff and dreams about getting funding to build a fleet of autonomous gliders recharding by regenerative soaring.