e-science experiences: software engineering practice and the eu datagrid lee momtahan and andrew...
TRANSCRIPT
e-Science Experiences:Software Engineering Practice
and the EU DataGrid
Lee Momtahan and Andrew MartinOxford University Software Engineering Centre
Contents
• EU-DataGrid
• Challenges
• Comparisons
• Conclusions
EU-DataGrid
• 9.8m Euro project over 3 years;• 21 partners in 15 countries;• application in particle physics (and
bioinformatics, and earth sciences);• PetaBytes of data: datasets to be catalogued,
replicated where necessary;• seamless delivery of computing resources• 200 staff, meeting infrequently (60 FTE)
Project Goals
• build application frameworks potentially involving huge amounts of data, compute power and distribution
• provide secure, managed, uniform access to such resource
• facilitate collaboration, and remote access to data and scientific instruments
• manage such facilities as a persistent service
Work Package Structure
Our role
• becoming involved after project started• funded to bring computer science/software
engineering experience to the project• intending to help by modelling aspects of design
in order that the system may be better understood, designed, built, documented
• in passing made the observations documented here• interested in the generality of these issues for
e-Science
Challenges
• Requirements Volatility– Novel paradigm; New diversity;
Volatile off-the-shelf components
• Geographical Separation– communication can easily become a
limiting factor (Brookes); Physicists are used to collaborating in experiments –but software?
• System Decomposition– Political concerns; geographic
determination
Challenges
• Project Processes and Authority– there is a quality plan… how do you get people
to follow it? is a commercially-based process appropriate? what about traditional academic means of QA?
Challenges
• Planning and tracking– exit criteria for an iteration seems to be the completion
of a document detailing the problems found in testing
Comparisons
• academic software production vs. commercial software production
• academic software production vs. other academic activity
• CMM Level for Software? For Paper/Proposal Writing?
• open source software vs. open source development• open source models vs. publicly-funded research• publication in journals vs. publication to a
repository
Conclusions
• Practice of software engineering in an e-Science context is substantially different to industrial practice
• Industrial models do not seem appropriate
• Open source models seem to fit better
• Publication and review are the key to quality and process improvement