Partner Organizations: Visual Numerics Inc.: In-kind Support; Collaborative Research
Proposed September 30, 1996: Recently VNI has teamed with Netscape and Javasoft to enhance the Virginia Techn's PV-Wave Java applet created for the Atomistic Simulation Laboratory. This applet is being redesigned as a template for more general purpose, network distributed tools that will allow researchers to create distributed research tools for any discipline. The enhanced version of this applet will be displayed at the upcoming Nescape Developers Conference in New York City on October 17-19, 1996. Dr Kriz's approach and efforts in distributed, collaborative computing environments are consistent with VNI's strategic view of the evolution of the scientific and technical computing markert. Results July 17, 2001: JWave was developed as a commercial product and used in the construction of CRCD modules.
Proposed September 23, 1996: NCSA will assist Virginia Tech researchers in the creation of educational modules that will access supercomputing resources and provide simulation results to be visually interpreted in real-time if using collaborative tools like NCSA's Habanero. NCSA is also interested in exploring the creation of new Java servers and applets that will work with third party visual data analysis tools such as PV-Wave. Results July 17, 2001: Software developed: AtomView, CAVE Collaborative Console (CCC), and CCC_atom.
Activities and findings:
Research and Education Activities: COMPUTER SIMULATION OF MATERIAL BEHAVIOR -- FROM ATOMISTIC TO THE CONTINUUM LEVEL (Engineering Education Program: EEC-9700815) ORIGINAL PROJECT SUMMARY OBJECTIVES PROPOSED IN 1996: Computing has had a tremendous impact on research in materials engineering and engineering mechanics during the last decade, but much of this knowledge remains to be transferred into the classroom. Access to supercomputing has led to the development of new numerical models, simulation-visualization tools, and potentially useful material systems. A sequence of two courses - one for seniors and the other for first year graduate students - is proposed in which state-of-the-art computer hardware and software will be used to teach the concepts and techniques of modern material performance simulation. The focus will be on the range of computer simulations covering and linking the physical length scales necessary for taking a material from the laboratory to the marketplace. These courses will bring into the classroom, research in materials engineering that the Principal Investigators and others have been conducting over the last six years, with an emphasis on computational methods applied to the study of deformation and fracture processes from the atomistic scale to the macroscopic scale. Subjects will be introduced and taught using advanced visual tools and interactive computing in the classroom, and will prepare students for designing complex material systems for the 21st century. Topics to be covered include effect of dislocations, grain boundaries and crystal orientations on material response; dependence of macroscopic behavior upon that of the constituents; localization of the deformation into narrow bands of intense plastic deformation; effects of material anisotropy; crack initiation and propagation in metals and ceramics; performance of advanced composites; modeling of heterogeneous materials; and wave propagation in anisotropic media. An interdisciplinary team of material scientists and engineering mechanicians will develop and teach the courses. The courses will also emphasize design of material systems and thus enhance the capstone design component of the existing engineering curriculum. The courses will have Web-Java based modules on specific topics in atomistic aspects of Materials Science and Micromechanics. The modules will stress the way in which macroscopic materials properties are controlled by phenomena at the atomistic and microstructural levels. Advanced computational and scientific visualization techniques will be used to incorporate research into the modules. The National Center for Supercomputer Applications (NCSA) will work with Virginia Tech as part of NCSA's new Partnership in Advanced Computational Infrastructure (PACI) proposal (pending). Together researchers will create modules that will access NCSA computing resources and provide visual interpretation of simulation results both in real-time on the Web and by Java enabled batch jobs submitted to remote supercomputers both at Virginia Tech and NCSA. Access can then also be extended to PACI industrial partners and other Universities. Virginia Tech has also been funded by NSF (CDA-9601874) to build a CAVE(tm) virtual environment in partnership with NCSA. The virtual immersive CAVE environment will be used to demonstrate complex structure property relationships. Educational pedagogy will follow the current research methodology on notions of learning higher level problem solving skills, and the requisite knowledge necessary to operationalize them, in contexts which are as realistic as possible. For example, useable knowledge is situated in the environment in which it is needed. Knowledge and skills learned in the classroom are situated in the classroom and, as a result, best remembered in the classroom. Conversely, knowledge and skills learned in the problem solving context are most usable in that context as well. A critical evaluation component will then be used to measure the transfer of what is learned to novel, realistic contexts. Visualization, by providing a more context rich set of stimuli, should produce the most 'portable' knowledge. An evaluation team will consist of: 1. Industry: Drs. Buddy Poe (NASA-Langley ); Dr. Vijay Stokes (General Electric Co.), 2. Virginia Tech: Dr. R. E. Denton, Jr., Head of Communication Studies and one faculty member from the Mechanics and Materials Sceince Departments; 3. Other Universities: 1. D.J. Srolovitz (Univ. of Michigan), D.M. Barnett (Stanford), A. Gilat (Ohio State Univ.), R. Talreja (Georgia Tech), G.J. Weng (Rugers), P.K. Law (Univ. Tennessee), and Elias Aifantis (Michigan Technological Univ.). Together these individauls will formulate and evaluate the course content in anticipation of needs of the materials engineers in industry. Some of these individuals have volunteered to participate in the creation of courseware. The course contents, philosophy, modules, and the feedback received from various segments of the scientific and industrial community, will be disseminated to the engineering community at large through presentation of the work at suitable conferences, workshops, short courses, seminars, the Web, and educational journals. _______________________________________________________________ CAVE: is a trademark of the Electronic Visualization Laboratory of the University of Illinois 1.0 RESEARCH AND EDUCATIONAL ACTIVITIES: 1.1 RESEARCH ACTIVITIES (Development of a Web-based Java Interface): CRCD educational modules were developed and distributed on the 'Wave-Java' Web-server, http://www.jwave.vt.edu, which was part of our early efforts to create a distributed, Web-based, visual computing environment in collaboration with Visual Numerics Inc. (VNI) and Sun Microsystems Inc. Development of this Web-server together with the creation of the Scientific Modeling and Visualization Classroom (SMVC) was supported by a foundation grant from SUN, and VNI which was to be located in Virginia Tech's new building called the Advanced Communications and Information Technology Center (ACTIC), see Figure 1. The ACITC is the corner stone of Virginia Tech's new information technology initiative. The NSF ARI proposal (CDA 9601874) to build a CAVE was also funded to be part of the UVAG of the ACITC. The University Visualization and Animation Group (UVAG) was organized within the ACTIC and included the SMVC, CAVE, Virtual Environment Head Mounted Display Lab, Immersive Desk, and Immersive Workbench. Together all of these resources were targeted as resources for this CRCD project. Construction of the CAVE, SMVC, and Wave-Java Web-server in the UVAG of the new ACITC building represented a significant effort that required additional support by the university, which was not included as cost sharing in these proposals. The final Fastlane report on the NSF ARI CAVE project has been posted at WEB-SITES: [1], which describes how this CRCD project used the CAVE. Figures 2 - 8 shows the layout and facilities of the UVAG used by this CRCD project. The research portion of this project used existing Web-based commercial software and created new Web-based Java Graphical User Interfaces (GUIs). With this software instructors created CRCD modules that maintained and archived numerous parametric studies based on computer simulation legacy code. The Web-based Java interface was used as the vehicle to create and transfer the collective knowledge of the instructors to students in two courses. Students used this Web-based Java interface to: 1) enter information required by the simulation, 2) compile that information into a data file, 3) submit the file as a batch job to the server or on a remote computer (optional), and finally to 4) send raw data of simulation back to the server (optional) where images of data were generated for viewing back to the student's Web-browser. Unique to this project was the level of industrial participation in the creation of the Java Web-based interface. Early Java/PV-Wave applet prototypes developed at Virginia Tech have been replaced with the JWave interface developed by VNI. The Network Programming Interface Builder (NPIB) replaced the original Web-based Java interface developed prior to this CRCD project, SOFTWARE: [1]. Development of the Web-server continued with funding from the CRCD project. The server was upgraded to a SUN Sparc 10-Ultra with 1Gbyte of memory using the most recent SUN operating system and commercial software. With this upgrade it was possible to execute larger simulations and generate more representative results for analysis. However most of the existing simulations where already configured to execute on larger remote site computers and could not be moved onto the Web-server. The reason for this is that typically professors in engineering and the sciences configure their own computers with 'unique' (typically older versions) software, compilers and operating systems which are tailored for executing their particular legacy code. Software running on remote site computers was difficult if not impossible to install on the current Web-server because of compatibility issues with newer and different versions of operating systems, compilers, and commercial software. Another important factor was that most simulations required more memory and multiple CPUs not available on the Web-server. Hence, further development of NPIB required that NPIB be extended to coordinate execution of simulations on these larger remote site computers. This problem is typical and will continue to persist because professors outside of computer science are not rewarded to rewrite their legacy code to be compatible with the most current commercial software and operating systems. Although 'old' legacy code has created this unique situation, NPIB has adapted to it, which was not anticipated at the beginning of this project. With modifications to NPIB (version 1.6) larger simulations can be realized on these 'uniquely-configured' remote site computers, which will provide students access to larger and more realistic simulation results. Details of the creation of JWave and NPIB modules, with access to remote site computers, is given in JOURNALS: [1]. For comparison the same legacy code was used to develop CRCD micro-scale module01 (JWave-GUI) and module02 (NPIB-GUI). These two micro-scale modules have been specifically organized and documented as working examples in 'Interactive Computer Program Modules' section of the 'Micro' section of the CRCD Web-server home page, WEB-SITES: [2]. Studying these two working examples together with the 'How-To' tutorials in the NPIB Users Guide, SOFTWARE: [2], demonstrate how to setup and create modules using the JWave and NPIB GUI-interfaces. For this project comprehensive documentation was created to provide developers the necessary information to build other modules that extend beyond the objectives of this CRCD project. Figure 9(a) shows the interactive JWave module01 applet GUI. This GUI form allows students to fill in parameters. Information shown in Figure 9(a), is needed by the legacy code to calculate elastic property polar plots. When the 'Update Plot' button is selected the applet transfers parameters to the PV-Wave procedure which, plots results shown in Figure 9(b). Module01 (micro-scale) was created for students to experiment with different mechanistic modules that predict anisotropic elastic properties at the macro-scale, which are calculated from micro-scale constituent (fiber/matrix) elastic properties. Here only graphical results are returned through the JWave applet. The same micro-macro models were used in module02 (micro-scale) for comparison, where the NPIB GUI-interface returns the same graphical results but along with all of the simulation model source code and data sets for further analysis. Simulation parameters are viewable by the client (student/researcher) in the NPIB form with a web browser, see Figure 10. Specific information about the remote site computer is included at the bottom of the NPIB form below the 'Submit' button. When the 'Submit' is selected, simulation results are returned to the client (student/researcher) by opening a new Web browser window, and by sending the client email when the simulation is completed. When the file transfer to the remote computer is complete, a new Web-browser window is opened and displays the directory where simulation results are archived by creation of a unique directory structure on the Web-server. With only a Web browser the client can view simulation results in real-time, see Figure 11, as they are created and appear in this directory. This comparison demonstrates two different but equally useful GUIs and gives the developer a choice. In some cases only graphical results are needed, JWave module01 (micro-scale). In another case clients (students/researchers) may need more information for analysis and interpretation of simulation results, that is organized in NPIB module02 (micro-scale). 1.2 EDUCATION ACTIVITIES (Building modules for classes on mechanical behavior): The first NSF-CRCD senior-level course (ESM-MSE4984) was taught in the Fall semester, 1998 and a second class (ESM-MSE5984), a first-year graduate level class, was taught in the Fall Semester, 2000. Both classes are three-credit hour classes, meeting for one hour three times a week: Monday, Wednesday, and Friday. Mondays and Wednesdays were reserved for lectures. On Fridays students met with instructors in the Scientific Modeling and Visualization Classroom (SMVC) of the ACITC. The students used the SMVC more for the undergraduate class than for the graduate class. Because of the improvements in the CRCD Web-server the students used the CRCD modules outside of the SMVC lab, which explains the decreased use of the SMVC for the graduate class. Results of the undergraduate class are posted in the 'ESMMSE4984' section of the CRCD Web-server, WEB-SITES: [2]. Student projects were also posted at this same location. For both classes, a list of lecture topics and interactive computer simulation modules is organized in Table 1 for each of the scales: nano-scale, mirco-scale, and macro-scale. Most of the interactive computer modules were associated with the lectures and homework assignments. Two nano-scale modules and one micro-scale module were designed for use in the CAVE. For these three modules students could open the Web URL address on the CAVE computer and directly load the inventor (*.iv) files, which are listed in the archived Web-server directory, into the CAVE using standard virtual navigation applications such as Pfnav or SARAnav. The students could then navigate through these complex three-dimensional (3-D) structures in the CAVE. Figure 12 shows a student, red spherical head with two black eyes, observing a mode-I crack propagating along a Ni-Al grain boundary. From this simulation students can observe dislocation emission for the crack tip which results in a crack with a larger crack radius, hence higher fracture strength. For nano-scale models several publications are referenced here that show how the CAVE was used as an effective research tool JOURNALS: [2-7]. For micro-scale models, Figure 13 shows the elastic anisotropy of a special orthorhombic crystal class symmetry that results in a single connected stress wave surface glyph: longitudinal and both transverse wave surfaces connect into a single surface, JOURNALS: [8]. In both cases the fully immersive experience of the CAVE can be used to study and analyze these structures and their corresponding material properties. Several virtual environment software tools were developed for analysis and collaboration of simulation results in the CRCD nano-scale section, SOFTWARE: [2-4]. These software tools are briefly described here, 1) Atomview (desktop simulator, I-Desk, or CAVE): developed in collaboration with NCSA by John Shalf used in module01 and module02, 2) CAVE Collaborative Console (CCC): EVL-Limbo based collaborative virtual environment linking desktop-simulators, I-Desks, and CAVEs into a collaborative workspae, 3) CCC_atom: CCC and Atomview combined into a collaborative shared virtual environment workspace to analyze and interpret simulation model results. All of these software tools only run on SGI hardware using CAVE libraries sold by VRCO Inc.
Findings: 2.0 MAJOR FINDINGS AND RESULTS: 2.1 Things that worked well Except for the occasional server downtimes, the NPIB and JWave interfaces worked well for both the undergraduate and graduate CRCD classes, CONFERENCE: [1]. Results from the undergraduate class, revealed that the most productive time spent using the CRCD modules was when students and instructors met in the SMVC on Fridays, JOURNALS: [9,10]. Friday classes resembled lab sessions where students could ask questions and try out their ideas with comments from the professors who also helped interpret the simulation results. Instructors also received valuable feedback on how the JWave and NPIB forms were working and what needed to be improved. Friday sessions also built student confidence for successful completion of their homework assignments. Results from the graduate class revealed that significant improvements on the CRCD Web-server modules contributed to decreased use of the SMVC by the students. 2.2 Things that need more work For the first class, although the NPIB (version1.4) form worked well, the 'builder' part of the NPIB was improved with more features; however it was not stable enough for the instructors to build their own forms. Consequently the technical support team members built all the NPIB version 1.4 forms using a scripting syntax that was difficult for the professors to learn. NPIB was upgraded to version 1.5 where the scripting language was replaced with a Java form builder. The Java form builder enables instructors, who are not Java literate, to build interactive NPIB forms. For the undergraduate class NPIB 1.4 forms only worked on the SMVC UNIX workstations with Netscape 4.5. Near the end of the semester we did get the NPIB 1.4 forms to work on Windows-NT Web browsers. This was largely due to the way Windows-NT handles screen refresh. For the first class the CAVE was located off campus and for the second class the CAVE was under reconstruction in the ACITC. So the CAVE was never fully utilized. 2.3 Lessons Learned Java interface development is a difficult, if not impossible, for most engineering professors without backgrounds in computer science. Even using Javascript with HTML is beyond the abilities of most professors. These same professors are also not experienced in routine systems administration needed for configuring and maintaining Web-servers. Hence there must be a commitment from the department or college to support a courseware Web-server and to train professors how to access and use systems such as the NPIB. Because of limited resources and reluctance to accept new technology, building and supporting courseware servers has been the most difficult aspect of this project. Because of these difficulties, the CRCD Web-server was maintained by the Problem Solving Environment (PSE) group in the Department of Computer Science, WEB-SITES: [4]. In both classes we also discovered, based on previous experience, that programmers and system administrators need to work more closely. Typically in the past professors used standard language compilers that ran on computers maintained by the university or department, with access to technical support staff to answer any questions. With the advent of the network, professors are required to use a variety of different resources distributed over a network of computers and operating systems. Technical support staff are no longer isolated to support one or two computers in the 'new world' of computing where popular Web-based software applications are distributed and constantly changing. Successful projects now require that professors devote more time learning computing skills and how to work more closely in teams with their system administrators. We also experienced first hand how Java needs to be maintained as a standard: early interfaces developed in Netscapes' IFC had to be rewritten, CONFERENCE: [2]. Our experience in team-teaching this class was a rewarding but difficult. In the first class more time was spent solving technical problems than was spent developing course content. This trend reversed in the second class because many of the modules were already developed and many of the technical problems solved. For the second class more modules were written for the 'Micro' section. To continue the learning experience, witnessed on Fridays in the first class that used the SMVC, students needed access to the Java-Web server from outside the SMVC. Although convenient to manage, students should not be required to go to any particular workstation classroom environment. Some universities, because of security issues and convenience of management, prefer to isolate these resources from remote access. Such policies are counterproductive when every student is also required to own his/her own personal computer and where professors, who are located off-site, are expected to create courseware materials but cannot do so remotely. Improvements on the CRCD Web-server allowed students in the second class to work mostly from the CRCD Web-server. Remote access will aid in future development of these courses for distance learning. Researchers and students did not use Atomview or CCC_atom on the desktop-simulator, because most of the simulation results were too large. Even for results that were not too large, students tended not to use Atomview or CCC_atom, because this software would only run on SGI operating systems and would not run on the students personal computer. The cost of SGI workstations together with the cost of VRCOs desktop CAVE libraries excludes the 'low-end market' of the desktop computing environment for both students and professors. The growth of open-source copyrighted software (i.e. GNU-GPL/LGPL) has been motivated by similar situations were students and professors can benefit more from open-source copyright. Atomview was an exception and used extensively in the CAVE only for research as noted in JOURNALS: [2-7]. But since the desktop is very important, the link from the desktop to the CAVE is also important, hence affordability at the desktop by association is important. 2.4 Future Developments Because the CAVE and SMVC were not fully operational, when we taught the second course, we hope to teach this course again now that these resources are operational. Since the second class was taught, there have been several improvements. The Java-Web server was upgraded to JWave 3.0, better security measures were implemented without restricting access to the anonymous ftp site, and NPIB was upgraded to version 1.6. Thus professors can build their own JWave and NPIB forms in their public_html directory. Other ideas for future development are: 1) validate and verify data before they are submitted to the server to check for data types: int, floats, etc., and 2) auto-generate NPIB forms given a specific input file (work with existing legacy data files). Although our current Sparc10 Ultra server has been upgraded to 1GB of memory with 92GB of disk space, in the near future we will move the CRCD Web-server onto an SGI Origin 2000 server with 8 R10K CPUs, 8GB of memory, and 73GB disk space. With the current version of NPIB 1.6 students can now access remote site computers, get their results back quicker, and submit larger simulations, which will yield more realistic simulation results. With the NPIB 1.6 upgrade working we plan on accessing several new high performance computing systems at Virginia Tech: 1) Sun Enterprise 6500 computer with seventeen 400MHz (8MB cache) processors, 18GB memory, and 144GB of RAID disk, 2) SGI Origin 3400 rack with 8 R12K CPUs, 8GB of memory and 146GB of disk space, and 3) Beowulf cluster with 1GHz 200 CPUs and 1GB memory per CPU. CRCD modules will be extensible to other classes taught in the Engineering Science and Mechanics (ESM) Department. We hope that this interest will grow to other ESM classes and that eventually the ESM Department will support their own JWave courseware Web-server. Virginia Tech is developing an new virtual environment software API called DIVERSE, SOFTWARE [5], that is licensed GNU General Public License, similar to GNU/Linux. This type of licensing scheme together with the development of the OpenGL interface will promote research collaboration at the desktop. Future funding will be pursed to rewrite Atomview, CCC, and CCC_atom using the DIVERSE API. This will solve the problem of both cost and running on other operating systems other than SGI as previously noted.
Training and Development: Already reported in the sections: 1.0 Research and Educational Activities 2.0 Major Findings and Results
Journal Publications:
Other Specific Products:
SOFTWARE: (Brief Software Description) 1. Network Programming Interface Builder (NPIB): A rapid application development tool that researchers and educators can use to create, maintain, and archive numerous parametric studies based on their legacy computer simulations 2. Atomview: An SGI Performer based VE application used by material science research to visualize and interpret supercomputer simulations of nano-structure physics based models. 3. CAVE Collaborative Console (CCC): A general collaborative VE that links desktop workstations, I-Desks, IWBs, and CAVEs into a shared working ('design') environment. CCC is an SGI Performer based application, that was built on top of CAVERNsoft and Limbo in collaboration with the NSF PACI project. 4. CAVE Collaborative Console Atomview: A Performer based VE application that specifically linked CCC and Atomview to enhance collaboration between material scientists. 5. Device Independent Virtual Environment: Reconfigurable, Scalable, and Extensible (DIVERSE) API, is used to build collaborative desktop to the CAVE applications.
Each software item listed above has a Web site where the software can be downloaded and installed with tutorials and examples. These Web sites are respectively: 1. NPIB: http://www.jwave.vt.edu/npib/npib.html 2. Atomview: http://www.sv.vt.edu/future/cave/software/atomview/ 3. CAVE Collaborative Console (CCC): http://www.sv.vt.edu/future/cave/software/ccc/ 4. CAVE Collaborative Console Atomview (CCC_atom): http://www.sv.vt.edu/future/cave/software/cccatom/ 5. DIVERSE: http://www.diverse.vt.edu
WEB-SITES: 1. NSF-ARI Final Report (Fastlane Format) on Virginia Tech CAVE Project: http://www.sv.vt.edu/future/ari/final_rpt/final_rpt.html 2. Batra, R.C., Farkas, D., and Kriz, R.D., Combined Research and Curriculum Development: Mechanical Behavior of Materials: http://www.jwave.vt.edu/crcd/ 3. Parikh, S.D., VRML 1.0 format necessary for viewing in both the CAVE and VRML Web-based viewer: http://www.sv.vt.edu/classes/vrml/exercise3.html 4. Shaffer, C., Problem Solving Environment: http://www.cs.vt.edu/~pse
Information is shared by access to Web URLs listed.