GridPort Launches Computational Science Applications on the Web

omputational grids promise to harness the combined power of networked computing resources for solving the most complex scientific problems. In this environment, the entire grid can be treated as if it were one large, almost infinitely powerful computer. However, many technical obstacles remain to be overcome before programmers and scientists can view all of the networks, machines, and storage devices as a unified whole. NPACI's GridPort project, developed at SDSC, is creating a toolkit to overcome these obstacles and make it easy to create computational science portals for chemists, physicians, and other scientists, unlocking the full potential of the grid.

A GRID PORTAL TOOLKIT

EVOLUTION FROM THE NPACI HOTPAGE

DRUG, CHEMISTRY, AND HEART MODELS

A GRID PORTAL TOOLKIT

The Grid Portal Toolkit (GridPort) software and architecture is being designed by a large team of SDSC staff led by Mary Thomas, manager of the Computational Science Portals group. GridPort will allow application developers to build Web-based portals to the computational grid while hiding the complexity of managing grid resources so that users can concentrate on the scientific components of the problem, with little knowledge of the underlying portal infrastructure.

The toolkit is comprised of a collection of modules such as the grid portal services module which runs on commercial Web servers, and provides authenticated connectivity to the grid that is used by client applications. The application module provides a Web-enabled interface that allows customized science portal development by end users. A key feature of the architecture is that the client application and grid portal services can run on separate Web servers (Figure 1).

"We are designing GridPort to enable scientists to build new computational portals rapidly," Thomas said. "GridPort provides the underlying security and infrastructure required to use the grid and to access distributed HPC systems." GridPort was announced at the 3rd Grid Forum meeting, held at UC San Diego March 22–24, 2000.

GridPort scripts and tools provide consistent interfaces between the underlying infrastructure and security, and are based on grid technologies such as the Grid Security Infrastructure (GSI), Globus, and standard Web technologies such as the Common Gateway Interface (CGI) and Perl. GridPort is an open architecture that is designed to be flexible and capable of incorporating other grid services and technologies, such as the SDSC Storage Resource Broker (SRB) data-handling system, Application Level Schedulers (AppLeS), and the Network Weather Service as these become available.

To accomplish these tasks in a secure environment, projects such as GridPort and the SDSC SRB are building on the GSI, in an effort led at SDSC by Wayne Schroeder of the Data-Intensive Computing group. Developed by Globus staff for use in Globus and other grid applications, GSI consists of a core library and some utilities that make use of an X.509 public key infrastructure to provide secure authentication in grid applications. The GSI library is layered on top of the SSLeay or OpenSSL library, which performs the X.509 certificate handling and secure sockets layer (SSL) protocol. X.509 certificates, like the ones used in Web-based e-commerce and other applications, are used in grid applications to securely identify users and servers across the Internet and across administrative domains.

In addition to Thomas and Schroeder, the GridPort team spans many SDSC technology groups and includes Jay Boisseau, SDSC associate director for Scientific Computing; Steve Mock, Ray Regno, and Maytal Dahan of the Computational Science Portals group; Joshua Polterock and David Archbell of the Server Systems group; Victor Hazlewood of the HPC Systems group; and Keith Thompson and Bill Link of the Distributed Computing group.

EVOLUTION FROM THE NPACI HOTPAGE

The GridPort tools evolved from work done on the NPACI HotPage, which allows NPACI users to access documentation and related resources on the Web. The latest version of the HotPage extends the informational capabilities with interactive features that let portal users access their HPC accounts directly through a Web interface. Mock is overseeing implementation of the interactive HotPage, which uses GridPort-based services to allow users to submit jobs, manage files and data, and communicate with NPACI's Globus-ready systems.

"The creation of the original HotPage was designed to make it easier for users to get information about NPACI systems," Boisseau said. "In creating the new version that allows users to actually manipulate their jobs, files, and data, we discovered we had developed a very general, robust, and reusable infrastructure for developing many kinds of computing portals: user portals, scientific application portals, and even education and outreach portals."

"The HotPage has driven success after success in terms of deployment of NPACI metasystems tools," Thomas said. "All of these successes and learning experiences have been folded into GridPort. Our objective is to provide portal users and developers access to computational grids in a way that is familiar, effective, and secure."

Science portals based on HotPage and GridPort technologies have been implemented as the NASA Information Power Grid HotPage and for the Department of Defense Naval Oceanographic Office Major Shared Resource Center. The GridPort research effort has also resulted in collaboration among portal development groups at NPACI, NCSA, the Globus project, and NASA IPG to develop a common PACI-wide grid portal, which is planned for the end of the year. The goal of the collaboration is to develop a common grid infrastructure so that user portals can access resources across all the collaboration sites. New infrastructure requirements are leading researchers to develop new solutions for user interfaces, publishing of data using XML technologies, and new security models.

DRUG, CHEMISTRY, AND HEART MODELS

While the HotPage is a low-level portal, essentially allowing authenticated users to log on to HPC systems through a Web browser, GridPort also makes the key components available to developers interested in providing Web access to their scientific applications. The resulting applications approach the true promise of the grid: A scientist interacts with a standard Web interface, while an application runs on potentially many high-performance computers at remote locations in response to the scientist's request.

GridPort is currently being used to create Web interfaces to chemistry, drug dosage, and other applications in collaboration with computational scientists at SDSC.

The GridPort team is working with SDSC applied mathematician Bob Leary to create a Web interface for a drug dosage modeling package from the Laboratory for Applied Pharmacokinetics (LAPK), led by Roger Jelliffe, a physician and professor of medicine at the University of Southern California. Jelliffe and the LAPK have been developing software for 30 years that allows physicians to tailor drug dosage regimens to an individual. While simplified statistical models run on desktop computers, more accurate, nonlinear models require supercomputing power.

For example, in developing a dosage regimen of several drugs, some of which may interact with each other, there are nonlinear relationships between blood stream, peripheral, and stomach concentrations; therapeutic effects; and toxic effects. The nonlinear models are important for studying drug behavior in infectious diseases, AIDS, cancer, and for optimal combination chemotherapy. GridPort tools are making it possible to create a portal through which physicians can run the nonlinear models, interacting with NPACI's high-performance computing resources only through a Web interface (Figure 2).

SDSC computational chemists Kim Baldridge, Jerry Greenberg, and Laura Brovold are working with the GridPort team to develop a Web interface to the GAMESS ab initio quantum chemistry application. Much effort has gone into parallelizing GAMESS, which stands for General Atomic and Molecular Electronic Structure System, on many platforms to make advances in areas such as materials science, electronics, environmental chemistry, and medicinal chemistry.

"With a GAMESS portal, chemists will not have to worry about platform-specific details to perform their computations," Greenberg said. "And to study larger and more complex systems, GridPort-based services will eventually allow us to apply more computing power without making the experience more complex for chemists."

The National Biomedical Computation Resource, sponsored by the NIH National Center for Research Resources, is creating a GridPort-based interface to the Continuity heart modeling code from the Cardiac Mechanics Research Group at UCSD. Continuity is research software for modeling biological systems including cells, tissues, and organs. Based on the finite element method, Continuity currently runs on SGI workstations and the Cray T3E.

Continuity gives special emphasis to solving problems in 3-D image modeling and other problems in bioengineering and physiology, especially those related to cardiac mechanics and electrocardiology research. With a Continuity portal based on GridPort services, bioengineers can expand their studies from workstations to supercomputers without having to learn a new computing environment.

"With GridPort technologies, Web portals can access a world of resources by using any modern Web browser regardless of location," Thomas said. "Scientists can run complex models from laptops in their hotel rooms, kiosks in airports, and eventually wireless handheld devices. GridPort efforts are helping to bring scientists and end users closer to the goal of a ubiquitous and transparent computational grid." --DH