Focused Effort Title: Collaborative Computational Science Portals: Convenient Portal Specification and Construction Organization: Florida State University Thematic Area(s): Scalable Computing Migration HPC Training and DoD User Productivity Lead: Geoffrey Fox/David Bernholdt Email Address: gcf@cs.fsu.edu/bernhold@npac.syr.edu Telephone: 315 443 3857 Fax: 315 443 1973 Statement of Work: In a set of synergistic efforts at ERDC and other NSF and DoD activities, we have been one of the leaders in developing web-based interfaces for both education and computational science. These are often termed portals in analogy to very successful Internet Sites like Yahoo and Netcenter, which provide one stop access to general information and tools. In a computational science portal, we aim to provide an interface with all the capabilities needed for a particular possibly multi-disciplinary application domain -- this is also called a problem solving environment (PSE) where even the early pioneers like Purdue are switching to the web approach in systems like WebPDELab. In a training portal, we similarly provide a convenient access to all the curricula and tools needed for education and training. We have set up a national study group (see http://www.computingportals.org) with aim of designing an architecture and implementation plan for the portals of interest to ERDC. The architecture is designed to include lessons from our current work and decrease risk to ERDC by agreement on interfaces and re-usable components so that we can easily evolve systems to use the best available sub-systems from academic or commercial sources. We have described the basic portal architecture in http://www.new- npac.org/users/fox/documents/pajavaapril00/ and will accumulate useful links in http://www.new-npac.org/users/fox/portalresearch/ and http://www.computingportals.org/. As part of the core effort we will continue to track technology from very specific portal ideas like those from http://www.desktop.com and general Internet distributed systems like Ninja (from UC Berkeley) and E-Speak. Some of the key concepts that underlie the emerging portal architectures include: o 3-tier architecture with two very defined interfaces -- client/server and server/back-end resource. The latter is largely the responsibility of the Grid Forum while the former is a key contribution of the Computing Portal Group with the current whimsically termed "betterportalML" XML interface o Personal servers containing logic to support individual clients with hand-held and PC/UNIX displays. We move away from the choice made in earlier systems with powerful browser based logic. Currently browsers do not provide robust virtual machines and further in a future where hand-held devices are important, we want to provide portal services for a variety of rendering devices -- this seems most elegantly implemented as a client side personal server that can drive any device. o A powerful event model integrating servers and clients and integrating both asynchronous and synchronous delivery. Events are just "time stamped information nuggets" and this concept supports both collaboration and user customization. We are proposing two focused efforts supporting these ideas; one described here is aimed at a prototype "betterportalML" as part of Communication and Collaboration; the second is aimed at a prototype event service as part of Training. Both technologies will be useable in either type of portal. We are also partnering with other PET partners in multi- institution focused efforts that are applications that will use and thereby test these technologies. XML is growing in use and popularity as a convenient way of specifying all (semi-) structured data or metadata. Note that one can use XML as an object specification even when the object is implemented in CORBA, COM or Java -- using XML to specify a structure does not imply that XML is used as the object model. For instance, some attributes in "betterportalML" are designed to specify the layout of a portal in a syntax like: ... The layout could be implemented either by transforming an XML file using an XSL style sheet or it could be the control file for a Java applet using a Flow layout mechanism. Our current work for ERDC on the LMS application already uses XML extensively to specify the code object structure and these ideas will be incorporated into the design of betterportalML. We will use the new coupled-watershed/hydrosphere and other applications from the Computing Portals group to establish requirements for betterportalML. Another important capability is the support of "general simple applications". The new portal architecture has one very important goal – enable users to build new portals very quickly. The result of this focused effort will be both a draft specification of betterportalML but also a prototype "portal construction wizard" that would allow users to choose the objects (and object parameters including layout and collaborative structure) and personal server services and so generate a betterportalML specification of their computing portal. Our idea is that "computer scientists" develop distributed system infrastructure, and services. Application scientists define objects (programs, data sets, resources) and can use default or customize portals that use the objects and services. Currently (as in LMS), computer scientists are needed every time one needs a new portal -- in the new approach, application scientists or their staff can build their own portals and define their own objects. New computer science work is only needed when bugs need to be fixed or improved or new services required. Hewlett Packard marketing says that e- Speak is designed to cut construction time of e-commerce sites to 2 hours (a factor of 100 reduction they claim). This focused effort has a similar goal for computing portals. The betterportalML will incorporate the lessons from our year-4 collaborative computing focused effort in terms of XML specification for asynchronous and synchronous collaboration. Realistically we will not have a "full set of capabilities" available year 5 but we must have enough to allow these new techniques to benefit ERDC. Thus we will also provide two important tools -- a portal authoring tool (the wizard described above) and the capability to convert betterportalML to an HTML browser page. This will enable ERDC users to construct their own portals for a class of simple applications. Deliverables o 4 months: Prototype definition of betterportalML o 6 months: Prototype betterportalML XML to HTML layout tool and illustrative demonstration of its use in a simple application. o 9 months: Prototype betterportalML generation wizard (portal authoring tool). o 12 months: Integrated Prototype allowing generation of customizable portal for several applications. o Presentations for PET Midyear and Annual Reviews o Inputs to PET Annual Report o Written progress reports in June and December 2000 o Final Technical Report in March 2001 Required Resources: $51,381 Item Base Fringe Overhead Total Geoffrey Fox 7,500 1,380 4,129 13,009 1.5 Graduate Students 24,000 144 11,228 35,372 Tuition 3,000 - - 3,000 TOTAL 51,381