MIDDLEWARE

By Steve Fisher, Editor In Chief -- Last week Supercomputing Online carried an interview with SDSC’s Fran Berman about the recent announcement that the National Center for Research Resources (NCRR) had awarded more than $20 million to a consortium of universities coordinated by the University of California, San Diego (UCSD) to build the first nationwide high-performance computer environment to study diseases of the brain. This week, we are proud to bring you the comments of BIRN project lead Mark Ellisman. Supercomputing: Networking definitely seems like the key in this effort. Can you please tell the readers in detail how the networking aspects of a project of this size will be accomplished? ELLISMAN: All of the partnering sites associated with BIRN are currently participating in Internet2. As such, the broadband Abilene and Calren-2 network backbones will serve as the foundation for the BIRN network. Within the budgets for each partnering site there is funding for a networking engineer (at 50% time). The BIRN coordinating center staff will work very closely with each of these engineers and campus networking authorities to place infrastructure to bridge the last mile from the existing campus Internet2 point-of-presence to each participating laboratory. While the fiber connections for this last mile will come from campus commitments, BIRN engineers from the coordinating center and partnering sites will audit the existing routing and switch infrastructure and specify the appropriate variation on a standard configuration of equipment to guarantee gigabit connectivity to each laboratory and associated BIRN data cache. Each partnering site will be equipped with a network probe to enable the BIRN Networking Operations Center (NOC to be established at SDSC) to monitor and tune the performance between sites. The NOC will be fully equipped and staffed to monitor network traffic and performance, model and solve reported problems by replicating the partnering site hardware and software configurations, and provide remote tuning of all aspects of the BIRN network. The NOC will also provide 24/7 "help-desk" services to maintain a high up-time environment. Training and dissemination of key networking technologies will also be coordinated by the NOC staff. Supercomputing: What sort of tools will be used to mine the mountains of data that will be available? ELLISMAN: The BIRN coordinating center will build a fully integrated architecture on top of the robust and capacious network for sharing and mining data. At the bottom layer, grid services will be employed to provide seamless authentication and access to remote computation and data resources. Globus and the Storage Resource Broker (SRB) are examples of services and applications that will be used to build a secure interoperable system of distributed data, computation, and imaging resources. On top of the data and computational grid, data modeling and integration tools will be incorporated and developed to facilitate the construction of project- and site-specific data models; extend methods to query and retrieve complex data and associations from each partnering site; and develop a body of "domain knowledge" to allow correlation of data in the BIRN archives (see Federation of Brain Data article in EnVision, Jan-Mar 2001 issue). To interactively and collaboratively present the various types of data available across sites, a universal visualization toolkit will be developed, incorporating elements and tools from each of the partnering site along with comprehensive file format converters. Supercomputing: The wireless access mentioned in the release seems like it would be of particular value to doctors and researchers alike. Any comment? ELLISMAN: For the researcher, technologies developed as part of BIRN may eventually replace and improve the common laboratory notebook. Tools for managing and querying relevant data, performing advanced computation, analysis, and visualization, and providing access to instrumentation will be brought out of the computer lab to the workbench and wet lab. Although the focus of BIRN is basic and clinical research, and not patient care, BIRN technologies certainly have the potential to impact patient care as well. As many have envisioned, wireless tools will be able to provide secure access to resources directed at delivering and navigating patient data to facilitate real-time diagnosis and treatment. Upon stepping into a patient's room doctors would have immediate access to monitored telemetry data, current chart and medical histories, tools for collaborative consultation, as well as recommended remedies on demand. These same information assistants will allow doctors to also remotely monitor and participate in imaging exams, further extending the access to physicians and enhancing quality of care. Supercomputing: Patients' rights are a popular topic these days. What measures will be taken to ensure the security of information across this multi-institutional and obviously geographically dispersed system? ELLISMAN: The BIRN coordinating center will work with each partnering site to establish the appropriate security strategies for each type of data (raw and descriptive) while maintaining high transfer rates. The Grid Security Infrastructure (GSI) system within the Globus toolkit will provide services to address security, including authentication, encryption, and enforcement of a signed certificate authority. For the Brain Morphology scientific project, human data will be appropriately sanitized prior to insertion into the BIRN file system as appropriate to satisfy HIPPA requirements. Supercomputing: Will the information from BIRN be shared with pharmaceutical companies and biotech firms right away? In the future? ELLISMAN: The technology trail blazed by this project will benefit industry, basic and applied biomedical research. There will be multiple levels of access enabled. Data will first be shared between collaborating laboratories or centers and over time one can expect more open access to larger and larger portions of the federated databases. Some pharmaceutical companies are already engaged in molecular brain mapping projects on animal model systems. Their efforts are presently comparatively closed. One would hope that, at some level, their databases will be part of what becomes available to the community as part of the federation efforts within the BIRN. There is however little motivation for pharmaceutical companies to share with one another. ---------- Supercomputing Online wishes to thank Mark Ellisman for his time and insights. ----------