Giving computational science a friendly face

Today, most researchers from any discipline will successfully use web-based e-commerce systems to book flights to attend their conferences. But when these same researchers are confronted with compute-intensive problems, they cannot expect elaborate web-based systems to enable their domain-specific tasks. Instead, they have to deal with archaic command-line tools or generic portals that mimic the technical complexity of the underlying infrastructure. These interfaces are expensive to use, require much training, and their laborious and intricate processes often lead to errors.
Of course exceptions to this situation exist. The most fortunate researchers have access to web portals or scientific gateways that were built to enable their specific tasks. But here we find a significant difference with e-commerce systems: where flights are purchased by millions of customers, these science gateways serve a specific niche inside an already small scientific community. Most of these communities cannot meet, let alone sustain, the cost of developing and maintaining a gateway specific to their requirements.
Scientific tasks are not only complex, they also change frequently to reflect progress in the scientific field. The development of scientific gateways is thus expensive since each specific task has its own specific requirements. In contrast, purchasing a flight is equivalent for almost everyone.

Cost-effective scientific gateways
Researchers at the University of Edinburgh’s National e-Science Centre have developed a cost-effective and scalable solution to delivering scientific gateways for computational science. They have recognised that compute-intensive tasks have common use-patterns, and encapsulated these patterns in a declarative language. Using this language, a portal designer can rapidly deploy several portlets within days, forming a scientific gateway. Each portlet guides researchers through their task and enables delayed collection of results. Their solution is appropriately named ‘Rapid’.

So how does Rapid work?
(1) The portal designer specifies the user interface and logic flow of a task. This specification is created in one XML file.
(2) The designer then uses the Rapid Translator, which reads this XML file and creates a portlet
(3) The portlet can be deployed into any portal container compliant with the Java Specification Request 168 Portlet Specification, such as GridSphere, WebSphere or Liferay.
(4) A domain specialist can use a web browser to log in to the portal and access the new portlet.
(5) The user can then configure and perform the task at hand, using domain-specific jargon and graphical user-interface elements such as drop-down menus, radio buttons and check boxes.
(6) The computational task manager embedded in each Rapid portlet runs the appropriate compute jobs on designated compute resources. Rapid can submit jobs to most high-performance computing vendors and grid computing infrastructures, and cloud computing resources will be added soon.
(7) The progress of all tasks can be monitored
(8) When the compute jobs finish, the results can be transferred and analysed by embedding web-based visualisation components.

Success stories and more
Rapid has been successfully applied in several specialist domains:

• Medicine: to enable wider uptake of image analysis on MRI brain scans of stroke patients (Scottish Funding Council Brain Imaging Research Centre)

• Chemistry: to create portlets for teaching and research to improve the usage of computational chemistry tools (joint Chemistry Research School of the Universities of Edinburgh and St. Andrews)

• Engineering: to run non-linear finite element analyses to develop real-time emergency response systems (University of Edinburgh)

• Seismology: to allow analysis of data available in the Orfeus Data Center, the primary EU centre for seismic data; these data no longer have to be downloaded and interpreted by individual researchers and thus can be more effectively used (Observatories and Research Facilities for European Seismology)

• Biology: to process microscopy data in conjunction with an Open Microscopy Environment, thus lowering the adoption barrier to these processing techniques (The Swedlow Lab at the University of Dundee)

Stable software releases, documentation, tutorials and examples are available at http://research.nesc.ac.uk/rapid/. Rapid is developed under an Open Source model and is available freely through a GNU General Public license. Feel free to contact us at j.vanhemert@ed.ac.uk to discuss potential applications.