RICERCA ITALIANA     

Science and Applications of Advanced Computational Paradigms

Abstract

Advanced computational paradigms are essential to extend the frontiers of our ability to process information in very many sectors of contemporary society. On this theme, the proposed center aims at promoting research, education, and relations with productive organizations, by integrating and sharing expertise from a number of disciplines with a strong computational component.

Scientifically, the center will pursue a deeper understanding of architectural and algorithmic paradigms from a general computer science perspective and the application of such paradigms to problems arising in astronomy, structural engineering, and experimental physics. Astronomy and structural engineering applications are mostly based on finite-difference and finite-element algorithmic paradigms to solve differential equations, which are most efficiently executed by tightly coupled multiprocessors. The physics applications considered in this proposal are instead motivated by the need of managing huge collections of experimental data: relevant algorithmic paradigms are those typical of data bases, data mining, and information retrieval; the most appropriate architecture appears to be the microprocessor clusters. Particular attention will be devoted to themes that are conceptually unifying across architectural paradigms and across application domains, such as the tradeoffs between general-purpose and special-purpose machines, and the tradeoffs between machine-specific and machine-portable software. Special consideration will be also given to to approaches, such as the recently emerging computational-grid paradigm, aiming at unifying the use of computational resources on a global scale.

The applications of the proposed research are innumerable. In the computer industry, the results can have impact on the design of high performance platforms as well as on the development of efficient software. In the field of astronomy, the computational approach will shed new light on fundamental aspects of the structure and the history of the universe. The studies of structural engineering, mostly focused on the physics of porous media, will impact several sectors such as the conservation of the environment (hydrogeologic hazards, impact of extraction of fluids from reservoirs), the production of energy (behavior of dams), the safety of transportation (behavior of materials under fire, e.g., in tunnels), and the development of biomechanical devices (behavior of biological materials). The physics applications pursued within the center will be integrated into larger regional and global efforts devoted to the formidable data management problems posed by the experiments to be performed at the Large Hadron Collider of CERN. The technological solutions emerging in this context have a high potential to lead to significant applications in other domains, as in the case of the world wide web, originally developed at CERN.

While several of the mentioned research lines are already actively pursued by the proponent groups, the center is expected to add a number of significant benefits: (a) the establishment of an advanced computing facility that will greatly enhance the applicability of theoretical results; (b) the synergy among different groups and the definition of common projects, to which most of the positions contracted out by the center would be devoted; (c) the establishment, with the cooperation and additional resources of other interested organizations, of educational initiatives; (d) the definition of a focal point for the further growth of an already significant number of relations between the researchers of the center and various productive organizations. <<<

Principal Investigator

International Relationship

RESEARCH ORGANIZATIONS

In the recent past, the units of the proposed center have actively collaborated with many national and foreign research institutions, as listed below.

Astronomy: University of Torino and Roma Tor Vergata; the Observatories of Bologna, Padova and Milano-Merate; SISSA in Trieste; the CINECA supercomputing centre in Bologna; the Max-Planck-Institute in Garching (Germany); the Institute of Astrophysics in Paris (France); the Fermilab in Chicago (USA); the European Southern Observatory in Garching (Germania); the Royal Observatory in Edinburgh (UK); the University of Nottingham (UK); the University of Kansas (USA).

Computer Science: University of Pisa, Roma La Sapienza, Roma Tor Vergata, and Roma III; Istituto di Matematica Computazionale, CNR, Pisa; INFN, Pisa; CINECA, Bologna; Brown University (USA); Cornell University (USA); Oxford University (UK); University of California at Irvine (UK); University of Cork (IR); University of Glashgow (UK); University of Texas A&M (USA); IBM Research (USA); Akamai (USA).

Physics: Many collaborations are active both through the Italian National Institute of Nuclear Physiscs (INFN) and through CERN.

Structural Engineering: Politecnico di Milano, Politecnico di Torino, Universita' di Genova, Milano, Palermo, Trieste e Trento; ENEA, Roma; CSIC (Spagna); VTT (Svezia); Universita' di Bruxelles, Lingby, Berlino, Atene, Jyvakyla, Eindhoven, Stoccarda; Universidad Politecnica de Catalunya; Universita' del Galles di Cardiff; CIMNE (Spagna).

PRODUCTIVE ORGANIZATIONS

First, some active collaborations with productive organizations that could play a key role for the specific activities of the center are outlined. Then, the potential for developing further relations is briefly discussed.

IBM. A fruitful research cooperation with IBM has been recently established. In 1999, G. Bilardi has spent several months at the IBM T. J. Watson Research Center, developing a collaboration along two main lines. The first line is concerned with the modeling and optimization of the performance of the memory system of the next generation of commercial multiprocessors. The second line is related to the Blue Gene project, launched last December and internally funded by IBM Research at the level of one hundred million dollars, to build by 2004 a one-million processor computer. Aiming at increasing the computational capability by three order of magnitude with respect to the most powerful computer systems of today, Blue Gene represents a mind-boggling project for computer science and engineering, and is likely to be the only project of this caliber for the next several years. Of equally great interest are the protein folding problems that Blue Gene is designed to solve, which are crucial in many medical and farmaceutical applications of the information now becoming available through the Genoma Project. In addition to the above scientific collaborations, IBM Italy has expressed strong interest in advanced education initiatives in the thematic areas of the center.

CINECA. Collaboration with CINECA is active along several lines, including joint research projects and use of their supercomputer systems. The computer scientists from Padova are involved in an ongoing discussion group with CINECA and other Universities, to coordinate the growth of advanced computing facilities at various sites, to perform a joint technology assessment, to better share resources, and to reduce costs. CINECA is also interested in contributing to educational initiatives in high performance computing.

EU INDUSTRIAL PARTNERS. The following companies are partners in various European Union projects of which the stuctural engineering unit is also a partner: Aalborg, Portland; Ind. Math. Institute di Linz; Partek, Helsinki; Metalogenia (Spagna); Liebherr (Germania); Boskalis (Paesi Bassi); Engin Soft (Italia); Quantech (Spagna); Rockfields (Regno Unito); NAFEMS (Regno Unito). L'unita` di ingegneria strutturale

ENEL. A collaboration is active between ENEL and the structural engineering unit, on issues of dam control.

AGIP/ENI. A collaboration is active between ENI and the structural engineering unit, on modelling subsidence with applications to petroleum reservoirs.

AUTOSTRADE DEL BRENNERO. A collaboration is being finalized with the structural engineering unit to study the behaviour of tunnels in case of fire.

BIOMECHANICAL INDUSTRY. Joint projects and collaborations exist with the companies Orotig srl (Italy) e Dental Trey srl (Italy).

The research activities planned for the center are of considerable interest for various productive sectors and, with time, could lead to the development of a wide spectrum of relations. The computer science component of the research will address problems that are central for both the hardware and software vendors of parallel systems, a market that is destined to substantial growth as the evolution of the microprocessor is approaching saturation. The research motivated by the large scale physisc problems is likely to lead to interesting developments for many public and private organizations that need to manage huge amounts of data and make them accessible to a large number of cooperative agents; several European companies have already expressed considerable interest to participate to projects being currently defined, within the V Framework Program of the European Union, to develop the grid paradigm. The investigations of structural engineering have applications in numerous areas, in addition to those involved in the active collaborations mentioned above: hydrogeologic and sysmic hazards, of interest to territorial agencies; nuclear waste storage, of interest to public institutions and to the energy industry; cultural heritage; and development of biomechanical devices. The numerical codes developed by the astronomy unit for the study of the evolution of fluids in three dimensions can be adapted with reasonable effort to a number of problems in fluid turbolence, aerodynamics, dynamics of the oceans, and vulcanology, which have both civil and industrial applications. <<<

Research Objectives

Computing is the process by which some available information is recast into a form more suitable for the intended uses. As such, computing plays a pivotal role in most aspects of the Information Society. In spite of impressive advances in computing technology, spanning almost 10 orders of magnitude in about half a century, there are still many domains where further progress crucially depends on the ability to develop new algorithmic and architectural paradigms to process the relevant information. For example, in 1993, the Committee on Physical, Mathematical, and Engineering Sciences of the US Federal Office of Science and Technology Policy has identified a number of ``Compuational Grand Challanges'', which even today are barely approachable by the most powerful computers, featuring a Teraflops of processing power and a Terabyte of storage capacity. These Grand Challanges include: global climate change, human genome, fluid turbolence, vehicle dynamics, ocean circulation, viscous fluid dynamics, superconductor modelling, quanto chromo dynamics, and artificial vision. These domains are of considerable importance for the progess of the basic sciences, as well as for their numerous applications to areas such as engineering and medicine. In more recent years, with the explosion of the Internet, extremely large amounts of data are becoming accessible electronically; a new set of computational challanges is arising from the opportunity to search and identify patterns in such a massive data set. The ability to win these challanges will enable many new applications, in particular in the business world.

In the outlined context, the general goal of the proposed center is to realize a close cooperation between researchers from various disciplines, to investigate the most advanced architectural and algorithmic paradigms and to develop state-of-the-art applications in a number of engineering and scientific areas. In the longer term, the center will contribute a number of methodologies and technologies which, after reaching a sufficient level of maturity, will be transferred into the society at large, with applications to innumerable aspects of life, as it is currently happening to several information technologies matured during the last decade or so.

Specific objectives:

1. Establish within the University a laboratory with high-performance computing platforms of various types, initially including a parallel multiprocessor and a cluster computer, together with the appropriate software and performance tools. This facility will enable a substantial expansion of the experimental component of the research activities of all groups involved, considerably improving the applicability of the results.

2. Create a forum for scientific collaboration between computer science and other fields of science and engineering. The benefits of the collaboration are expected especially along the following lines:

(a) The sharing of the intellectual and financial resources to achieve the critical mass necessary to justify the cost of both the equipment and the educational programs.

(b) The availability of an environment for faster and more effective transfer of advances in computer science to the other disciplines.

(c) The availability of an environment for the other disciplines to provide computer science with valuable experience and insight on which computational paradigms and theories are relevant to important applications.

3. Investigate a number of fundamental issues concerning the universality of architectural paradigms and the portability of algorithmic paradigms. The architectural paradigms will include computing by tightly-coupled multiprocessors, cluster computing and special-purpose computing. The algorithmic paradigms will include a number of methods to solve differential equations (finite difference, finite elements, and Monte Carlo) as well as a number of techniques for information retrieval, data management, and data mining. The more recently emerging grid paradigm, which combines both algorithmic and architectural aspects will also be considered.

4. Develop computational solutions with significantly better performance than currently available for a number of important engineering and scientific applications that require massive computation and/or massive data management. Areas of particular interest will include multiphysics of porous media (for applications to the study of fires within tunnels, durability of geomaterials, hydro-geologic hazards, sysmic behaviour of earth dams, biomechanical behaviour of biologic materials), cosmology (for applications to the study of the formation and evolution of cosmic structures and to the statistical description of the large-scale structure of the universe), and experimental high energy physiscs (for applications to processing large data collections, such as the thousands of terabytes that will be produced by the Large Hadron Collider at CERN).

5. Contribute to establishing an educational program at the graduate level, on the themes of the center, to form experts who combine a backgroung in any of the disciplines where advanced computing plays a crucial role (including, but not limited to those where the center will conduct research activities) with a solid knowledge of the frontiers of the science and applications of computational paradigms.

Organization The center will be organized into four units, broadly reflecting the major disciplines involved: Astronomy, Computer Science, Physisc, Structural Engineering. The center will be directed by a small board, including a representative for each unit.

The board will coordinate and monitor the progress of the research and educational activities of the center and, in particular, will oversee the development and the operation of the advanced computing laboratory. The board will also develop a long term strategy for the continuation and the evolution of the activities of the center beyond the initial three-year period of funding.

It is a key objective of the center to promote synergy among different units. The activities to be carried out jointly by different units will include a number of workshops, the definition of research contracts for scientists at various levels of seniority with a scope spanning two or more units, and the joint supervising by advisors from different units of both laurea and doctoral theses. <<<