The following report is available in PDF format for high quality printing.

Third Quarter FY 2002 Report – The National Fusion Collaboratory

Edited by D.P. Schissel1

1General Atomics (schissel@fusion.gat.com)

Overview

The focus this quarter was on the two demonstrations given at the Sherwood and TTF fusion science meetings, debriefings after these events on lessons learned, and the planning on future work for presentation at SC02 and APS/DPP 2002.

General accomplishments include:

·             Demonstration to several hundred fusion scientists on the initial production Fusion Grid and on new advanced visualization capabilities. A Collaboratory workshop was held at PPPL after the demonstrations to discuss lessons learned and plan future work.

·             Demonstration of wall–to–wall shared visualization capabilities to the NSTX and DIII–D scientific teams leading to a better understanding of requirements.

·             NFC project information was presented at the US–Korea Advanced Physics and Control Collaboration workshop and at the ISOFS FESAC subcommittee meeting.

·             The NFC Project received a very positive review by the PSACI PAC.

·             The NFC participated in the Web Portal Project kick–off meeting.

·             Components of the NFC project software development goals were integrated into the draft 5–year plans (FY04–FY08) of the Alcator C–Mod and DIII–D facilities.

·             An abstract outlining the first year experiences of the Collaboratory Project was submitted to the NeSC Workshop on Applications and Testbeds on the Grid.

General

The Program Advisory Committee (PAC) of the Plasma Science Advanced Computing Institute (PSACI) met to assess progress on the Fusion Energy Science projects within the SciDAC portfolio and to provide recommendations for possible enhanced productivity. This review included the National Fusion Collaboratory Project. The work of the NFC was highly praised. To quote from their report about the NFC: “The PAC was impressed with the illustrative demonstrations of advanced collaboratory software tools to various fusion theoretical and experimental groups and the effective solicitation and use of the feedback received. There’s also been a very effective engagement with OSCAR SciDAC groups to bring advanced network and visualization software tools into the fusion program to facilitate nationally distributed computing using the MDSplus framework. We endorse early deployment of these tools.”

The work and plans of the NFC were presented at the US–Korea (KSTAR) Bilateral Advanced Physics and Control Collaboration Meeting. There is significant synergy in the development of collaboration technology between the fusion programs of both countries and future joint–work in this area is being explored. The long–term goals of the NFC were integrated into the draft 5–year plan (FY04-FY08) documents of both the Alcator C–Mod and DIII–D facilities. This represents a strong endorsement by upper management of both facilities regarding the planned software development of the NFC. The NFC project was represented at the initial kick–off meeting of the Web Portals project. It is expected that the fusion energy sciences, and therefore the collaboratory project, will be a significant customer of web portal technology. The collaboratory’s work, plans, and experiences of forming a unified team of fusion scientists and computer scientists were presented to the Integrated Simulation and Optimization of Fusion Systems FESAC subcommittee whose charge is create a roadmap for the formation of a joint OFES/OASCR program.

A Collaboratory workshop was held at PPPL to discuss lessons learned from the two demonstrations and to make future plans. One lesson learned was that both remote computing and shared visualization demonstrations ran into difficulties with site–security policies during software implementation. The Collaboratory is presently summarizing our needs on security issues and is working with the PPDG and ESNET to reach an understanding on how the needs of site–security administrators and the concept of Grid computing can coexist. Feedback has also been given to the Globus team that should be integrated into GT3. The Collaboratory Project is planning on a release of the TRANSP Grid computation software and the SCIRun visualization package at the APS/DPP meeting in November 2002. Accompanying that release will be major demonstrations on both remote computing and visualization to the over 1000 fusion scientists who attend the meeting.

The Project web site continued to be updated with numerous new individuals being granted access to the workers area.

Security/Remote Computing

The Collaboratory Project demonstrated new remote computing capabilities at two major Fusion Science meetings this quarter. Several hundred scientists attended the Sherwood Theory Meeting in Annapolis and the Transport Task Force Meeting in Rochester. These first–of–their–kind demonstrations provided valuable feedback from the scientists on the direction of work for the Collaboratory Project. Additionally, they provided the initial introduction to this scientific community of advanced computer science techniques that can be deployed today to assist in their research. Accompanying the demonstrations, handouts and posters were prepared and distributed to the attendees.

Figure 1 Scientists at the Sherwood theory meeting discussing and giving feedback to the Collaboratory Team on the remote computing provided by the Fusion Grid as well as the advanced scientific visualization capabilities.

The remote computing demonstration used the TRANSP time dependent power balance and simulation program and was composed of six components: (1) PreTRANSP controller for setting up the TRANSP inputs and starting the run (at meeting), (2) An MDSplus data server where TRANSP data is read from and written to (at GA), (3) TRANSP, a tokamak data analysis and simulation package (at PPPL), (4) Remote monitoring of the status of the TRANSP analysis job (at meeting), (5) Scientific visualization of the data (at meeting), and (6) Visualization of the interactions between components (at meeting). After logging into the Fusion Grid using Globus credentials, the PreTRANSP controller was started on the showroom floor workstation. The controller allowed the preparation of TRANSP input data and to request the execution of the TRANSP run. After authorization, the TRANSP code was started on the PPPL Linux cluster using Globus remote startup mechanisms. After the analysis was complete, the results were written to the MDSplus server at GA. While TRANSP was executing, monitoring software followed the status of the run at anytime. When the run was completed, a graphics program running on a local workstation read data from the GA MDSplus server for data visualization. This demonstration was made possible by the completion of several technical tasks. Among those was the prototype integration of the Globus Job Manager and the Akenti authorization server allowed authorization policies to be set to allow only certain users to start a TRANSP job.

Visualization

The Collaboratory Project demonstrated new visualization capabilities, in conjunction with the remote computing demonstration, to several hundred scientists at the Sherwood Theory meeting in Annapolis. The demonstration involved the combination of a Fusion Package with SCIRun to create 3D visualizations of NIMROD data stored in MDSplus. This first–of–the–kind demonstration provided the initial introduction to this scientific community of advanced visualization techniques that can be deployed today to assist in their research. Accompanying the demonstration, a handout and poster were prepared and distributed to the attendees. Feedback from the scientists was very positive and two clear needs emerged. One need is the ability for SCIRun to visualize 2D as well as 3D imagery and the other is the need for different levels of user interface that trade off complexity of use with the ability to customize a visualization.

Figure 2 Simulation of fusion plasma magnetic field lines and pressure by the NIMROD code and visualized by SCIRun. The adoption and expansion of SCIRun for fusion data has given researchers unique visualization capabilities.

 

Work this quarter included numerous experiments in shared tiled display technology. This work culminated in a demonstration to the NSTX and DIII–D experimental teams. The wall at PCS was shared with the wall at PPPL (NSTX) and the wall at ANL was shared with the wall at GA (DIII–D). Using VNC, DMX, and Chromium technologies, these shared visualization demonstrations included tools presently used in fusion research as well as the new SCIRun tool being developed by the Collaboratory. The feedback from the scientific teams proved especially valuable as the Collaboratory begins to outline how this technology would be used in a tokamak control room. Work this quarter also included progress towards a software package for automatic tiled display alignment that will be crucial for minimizing maintenance costs.

An Access Grid meeting was also conducted for the first time to a number of fusion scientists. AG nodes at ANL and PPPL joined with nodes at Boston University and the San Diego Super Computer Center to form the fusion meeting; MIT/C–Mod staff went to BU, GA/DIII–D staff went to SDSC, and the PPPL tiled wall was attached to their AG node. The meeting provided a lively debate on the use of AG technology in fusion research. During the meeting the new Personal Interface to the AG was discussed and there was considerable interest. The software for this lower–cost reduced-functionality node has been released and will be tested by the Collaboratory in actual fusion science research.

Figure 3 Fusion scientists participating in introduction to the AG meeting

Appendix A: Non–Edited Reports from Individual Institutions

A.1      M. Papka for the ANL MCS, Futures Laboratory

·             Performed numerous experiments in shared tiled display environments, including special demonstration to General Atomic fusion scientists. Work including debugging the security issues associated with running VNC in wide-area use, optimizations to VNC transport to operate at efficiently in wide-area use, and understanding of user interaction issues when sharing across a wide-area connection. Other tests performed included using Mesa version of SCIRun and using DMX to share regions of a desktop between two sites.

·             Introduction of the Access Grid (AG) to the fusion scientists using the AG. Participants at PPPL used the tiled display node at PPPL, while General Atomic scientists used the node at SDSC and MIT used the node at Boston University. This included a presentation on the use, features, and history of the AG from ANL. An open discussion session followed with the fusion scientists discussing how they thought the AG might be useful to them.

 

Figure A.1–1 Fusion scientists participating in introduction to the AG meeting

·             A binary distribution of the Personal Interface to the Access Grid was released, more information can be found at www.mcs.anl.gov/fl/news.html on how to get the software, release notes, and tested hardware. Release of software required release engineering, construction of new software installer. The new installer allows for users to choose which type of install they want to do, either standard AG or new Personal Interface.

·             Finished the tests on getting SCIRun to work on top of the Chromium tools (http://graphics.stanford.edu/~humper/chromium_documentation/). Only missing feature is some the selection mechanisms which the current version of Chromium does not support.

Figure A.1–2 SCIRun running on Argonne’s
micro–Mural displaying a standard SCIRun medical demonstration

·             Based on resent improvements to the Chromium tools, we have implemented a new version of the tiled video tools. The new version simplifies the over code structure of the SPU. This new version will serve as the building block for future remote visuallization work. Building on the success of getting SCIRun to work with Chroimum, figure 3 shows the use of the University of Utah’s fusion package running distributed across 6 H.261 video streams.

Figure A.1–3 SCIRun running the University of Utah fusion package on top of
Argonne’s video Chromium SPU allowing for remote visualization over the AG.

A.2      K. Keahey for the ANL MCS, Distributed Systems Laboratory

In the last quarter ANL accomplished the following tasks:

 

 

 

 

 

 

 

A.3      D. Schissel for the General Atomics Fusion Group

General

·             Organized successful demonstrations of NFC software at two fusion physics meeting in Rochester and Annapolis (Schissel).

·             Presented the work and plans of the NFC to the US–KSTAR Bilateral Advanced Physics and Control Collaboration Meeting (Schissel). There is significant synergy in collaboration technology between the fusion programs of both countries and future joint–work in this area is being explored.

·             Presented the work and plans of the NFC to the Integrated Simulation and Optimization of Fusion Systems FESAC subcommittee whose charge is create a roadmap for the formation of a joint OFES/OASCR program (Schissel).

·             Assisted in organizing and running a 1.5 day NFC working–group workshop that was held at PPPL that acted as a lessons–learned discussion as well as a future planning meeting (Schissel).

·             Participated in the initial kick–off meeting of the Web Portals project held at Indian University where the NFC is expected to be a significant customer (Schissel).

·             Presented for review the work and future plans of the National Fusion Collaboratory Project to the PSACI PAC Review held at PPPL; PAC response was very positive and supportive of our plans (Schissel).

·             Contributor to the first draft of the DIII–D National Fusion Facility 5–Year Plan Proposal (FY04–08) to USDOE/OFES where the software products of the Collaboratory play a significant role in the plan for successful facility operation (Schissel, Peng, Burruss, Flanagan).

·             The project web site (http://www.fusiongrid.org) was maintained including updates for the April demonstrations at the TTF and Sherwood meetings.

Security/Remote

·             An NFC workshop was held at PPPL to discuss security and remote computing lessons learned from the TTF and Sherwood demonstrations and to put together a set of plans for the November demonstrations at SC02 and APS/DPP (Schissel).

·             The majority of work early this quarter was focused on the two demonstrations at Sherwood and TTF fusion science meetings. Several days in advance of both meetings were required for hardware and network setup. The Grid–enabled TRANSP application was successfully demonstrated to several hundred scientists at both meetings. This demonstration involved secure communication to MDSplus and SQL Server data repositories at GA, secure and authorized communication to a Linux cluster at PPPL to run the power balance code TRANSP, and control of the demonstration and visualization of the results from the meeting floor (Peng, Burruss, Schissel).

Figure A.3–1 Scientists at the Sherwood theory meeting discussing and giving feedback to the Collaboratory Team on the remote computing provided by the Fusion Grid as well as the advanced scientific visualization capabilities.

 

·             Designed the poster and handout material used at Sherwood 2002.

·             The PreTRANSP GUI (PreTRANSP) used preparing and securely invoking TRANSP at PPPL is being modified so that it may be ready for usage by DIII–D scientists for actual data analysis. This system orchestrates the creation of input data that is stored in MDSplus and the Run Management database, and starts the TRANSP run on the PPPL linux cluster. All communication is secure utilizing the Globus toolkit (Burruss).

Visualization

·             A shared visualization demonstration between ANL’s and GA’s Tiled Walls was given to the DIII–D scientific staff. This educated the scientists and allowed them to give the NFC feedback on how this technology could be used for shared visualization in either the tokamak control room or for more one–on–one interactions (Peng, Schissel).

·             The Utah application SCIRun, with the Fusion Package for MDSplus data access, has been installed at GA for visualization of NIMROD data (Peng).

·             Substantial improvements to the writing of NIMROD data from NERSC to the GA MDSplus server were accomplished this quarter. This work included porting the NIMROD post–processing code to the NERSC computer Seaborg, the opening of NERSC firewall ports for MDSplus I/O, and modifications to MDSplus to allow for larger NIMROD data storage (Flanagan).

A.4      M. Thompson for the Lawrence Berkeley National Laboratory

During this quarter the protoype integration of the Globus job-manager and the Akenti authorization server was completed and used during the two physics meeting demonstrations in April. A simple set of Akenti policies were created and X.509 identity certificates were issued to the physicists taking part in the demo. In this prototype the authorization policy controlled the right to start a TRANSP job on the server and disallowed the right to run any other programs on the server machine. It did not attempt to control other job-manager functions such as querying a job status and terminating a job. Design was started on the modifications required by the job-manager to check for authorization at these other points. The new design is being implemented at ANL, and I am providing support on getting a test Akenti environment working there.

Some work was done on clarifying the Globus firewall issues and attempting to facilitate the use of the Globus tool kit in the presence of firewalls. This issue is being worked on by the DOE Science Grid and PPDG projects as well. The immediate goal is to have complete documentation of what ports are needed by both the server and client side, a recommended port range for the ephemeral ports and possible work arounds for NAT firewalls. It is hoped that this docuement and a co-ordinated effort by these three projects will help convince the firewall maintainers to open up the ports required by Globus. Input is also being given to the Globus OGSA team on the desire to allow pull only communication from the client applications which are more likely to be running behind NAT firewalls at meetings or at sites who are not yet committed to the OGSA distributed computing model.

A.5      T. Fredian for the MIT Plasma Fusion Science Center

During this period

·             Assisted with prepartion for demos.

·             Did some network throughput analysis in an attempt to understand performance of high speed network links between geographically distant systems.

·             Developed prototype "GRIDPST" service which overcomes some of the network latency issues when transmitting data between geographically distant systems on the Internet. Have demonstrated up to 10 fold increase in throughput over single connection transfers. GRIDPST (Grid Parallel Socket Tunnel) splits communications into multiple streams and reassembles those streams on the remote system. Applications that send data over the Internet using large data packets can generally use GRIDPST services without any modification to the application.

·             Spent a fair amount of time working with the new alpha release of globus_io for Windows. Have not been successful using this release with MDSplus and have sent feedback to the globus developers.

A.6      D. McCune for the Princeton Plasma Physics Laboratory

PPPL hosted a workshop, attended by representives of all the institutions participating in the development of the Collaboratory, on May 16-17. There were numerous useful exchanges of information between fusion applications scientists and computer scientists, touching on issues of security, remote computation, and visualization.

Grid–Enabled Remote Computation (L. Randerson, C. Ludesher, D. McCune)

Development and testing of the TRANSP service continued. Access to the Globus-based service through firewalls is a major concern, and some progress has been made in this area:

In addition, the PPPL group worked with Bill Dorland (U. Maryland) to begin preparations for the deployment of the plasma microturbulence code GS2 as a Collaboratory service. This involved assisting with setting up a Globus system on a U. Maryland Linux cluster, as well as providing an MDSplus-capable tool for extracting TRANSP results needed as input to GS2.

Portions of the NTCC website (http://w3.pppl.gov) are being used to distribute PPPL-contributed client software for the TRANSP service.

 

Collaborative Visualization (E. Feibush, S. Klasky)

PPPL and Princeton University (CS) worked together to produce a demonstration of collaborative visualization capabilities, which was presented to members of the NSTX research team on May 3. Two approaches for sharing images between two tiled display walls, and between display walls and workstations, were shown:

PPPL's display wall was also attached successfully to the Access Grid and this was used effectively for audio-visual conferencing between the PPPL and Princeton University display walls (separated by 5 miles). The combination of simultaneous shared scientific visualization and audio–visual conferencing was quite impressive to a number of the attendees.

Work continued on the SciVis software, with further extensions to the API and underlying system architecture.

A.7      A. Finkelstein for the Princeton University Computer Science Department

PCS-PPPL collaboration

 

o           Research paper on tiled display alignment accepted by IEEE Visualization 2002. “Scalable Alignment of Large-Format Multi-Projector Displays Using Camera Homography Trees.”

o           Began development of a distribution software package for automatic alignment.

o           Modified Wiregl and VNCWall to make use of alignment information.

o           Began modifications to Chromium to make use of alignment information.

o           Modified VNCWall to tile a single display. Hoping this would allow for an inexpensive display wall - single PC with quad graphics card. Unfortunately, quad card performance not satisfactory yet, will explore other techniques.

A. Sanderson for the University of Utah Center for Scientific Computing and Imaging

The general direction of the third quarter was to get SCIRun and the Fusion Package out into the hands of fusion scientists for evaluation. This was done at the Sherwood meeting in April and with a meeting with scientists at GA and SAIC in May. Both meetings were very productive since we were able to interact with fusion scientists and get their direct reaction to using SCIRun.

Three main issuses came out of the these meetings.

·             The need for SCIRun to have 2D tools not just 3D. The need for a data access tools through the MDS+ server and local access via HDF5. The need for different levels of user interface. (e.g. IDL/ReviewPlus)

·             To address the 2D needs we have evaluated several methods of viewing 2D data within SCIRun using prototypes developed for other projects using SCIRun. We have not decided on a particular path but will make that decision in Q4.

·             Data access tools will be worked on during Q4 as joint project with Argonne/PPPL/UTAH/GA

The different level of user interface is currently being addressed by another group at SCI (BioPSE) using SCIRun and we piggyback our needs with theirs.

Other specific accomplishments include:

·             Developement of a data structure (Structure Grid) for fusion data with SCIRun. This allows for a more natural handling of the data in SCIRun.

·             Development several support modules for the Fusion Package in SCIRun that allow quick and easy sub sampling of the data. The sub sampling allows either slice (2D) or volume (3D) viewing.

·             Development of a 3D phi slicer for the 3D volume viewing.

·             Created a serries of demo networks for fusion scientists to see how specific viz. tools in SCIRun work.

·             Public distribution of SCIRun and the Fusion package.

 

This report is available in PDF format for high quality printing.

team contact info | about the fusion grid | fusiongrid research

Last modified 07/12/02. Comments? webmaster