Second Quarter FY 2002 Report – The National Fusion Collaboratory
Edited by D.P. Schissel1
1General Atomics (schissel@fusion.gat.com)
The focus this quarter was on creating the first example of a production Fusion Grid, on demonstrating different shared visualization techniques, and on demonstrating innovative 3D visualization techniques. This work is aimed towards giving demonstrations at two Fusion scientific meetings in April that will enable the Project to receive detailed feedback from the fusion scientific community. This feedback will be used to adjust the short–term working plan so that scientists can use the Grid and visualization tools for work that will be presented at the APS/DPP meeting in November 2002.
General accomplishments include:
· Demonstrated production Fusion Grid using the TRANSP code on a PPPL Linux cluster including an alpha version of Globus IO on Windows for secure connections via MDSplus to an SQL database and an Akenti–enabled Globus GRAM functionality for authorization services
· Demonstrated shared IDL–based visualization between two tiled display walls and a reduced functionality AG node for the desktop
· Created a fusion–SciRUN prototype for 3D visualization of NIMROD data stored in MDSplus and established a minimum hardware configuration for desktop visualization
· Held a collaboratory workshop at MIT, and participated in the SciDAC PI’s meeting, a SciDAC Network panel discussion and the AG retreat
· Two papers were submitted to HPDC 2002 & one abstract was submitted to WACE 2002
· An agreement was reached with General Atomics to use their new 8–projector Tiled Display Wall (16’ x 8’) for shared visualization research being conducted by the Collaboratory
The Collaboratory Project was represented at the SciDAC PI’s meeting where linkages to other projects were established. Specifically, the Scientific Annotation Middleware, Particle Physics Data Grid (PPDG), Pervasive Collaborative Computing Environment, Reliable and Secure Group Communication, and Storage Resource Manager Projects all have pieces that could potential tie into the Collaboratory Project. The strongest near–term linkage is with the PPDG and Mary Thompson (Collaborartory) and Doug Olsen (PPDG) have been identified to be the communication points between the two projects. As part of the PI’s meeting, the Collaboratory Project was represented on a Network Panel. There was considerable interest in the Collaboratory’s network usage scenario and a contact with network specialist Nagi Rao was established for future exploratory discussions and work.
A Collaboratory workshop on security and remote computing was held at MIT in January. A presentation on the Collaboratory was given at the Access Grid retreat in San Diego. Two papers were submitted to the High Performance Distributed Computing Conference on a cluster based image viewer and real time performance visualization on distributed systems. One abstract was submitted to the Workshop on Advanced Collaborative Environments giving an overview of progress made in the Collaboratory Project. Several Collabortory members attended the Globus 2.0 training tutorial.
The Project web site continued to be updated with numerous new individuals being granted access to the workers area. The mailing lists for the Project were updated to reflect mailings to the fusiongrid.org domain.
During this quarter a number of site–security and firewall issues were encountered that needed to be reconciled before work could continue. As software moves beyond the single laboratory network and security policy to a larger U.S. implementation, site–security issues will become more prevalent for the Collaboratory Project. Future work will need to address solving these issue in a general way so that solutions scale to building a large Fusion Grid. It is possible that other SciDAC programs are encountering such issues and a linkage not established at the SciDAC PI’s meeting might be forged to solve the site–security issues in a cross–cutting way.
Security/Remote Computing
After the successful demonstration at SC01, the Security and Remote Computing participants held a workshop at MIT to plan for two demonstrations at the April Fusion scientific meetings. The SC01 demonstration involved resources from the CSET partners and the work this quarter involved bringing fusion resources into the Grid environment. The scientific code to be used in the April demonstrations is the TRANSP power balance and simulation program that is significantly larger than the EFIT code used at SC01 and therefore stretches the Grid computational scenario to a much larger problem. This effort represents the first step towards building a production Fusion Grid and it is anticipated that Fusion Scientists will use this TRANSP service for physics work presented at the APS/DPP meeting later in the year.
The planned demonstration at the April meetings goes substantially beyond what was illustrated at SC01 by creating new capabilities and extending others. Globus IO has been made functional on Windows enabling secure Globus–based interactions with a Windows–based MDSplus server that acts as a gate way to the SQL run management database. This relational database is already used within the Fusion community to track simulation and analysis code runs. A prototype of the Akenti–enabled GRAM job manager has been developed to perform authorization on job creation. The monitoring capabilities of Globus were enhanced to meet the functionality required by the fusion community. New MDSplus and SQL data servers at DIII–D and the TRANSP Linux cluster at PPPL were brought into the Fusion Grid environment. Moving these resources into the Grid environment required substantial work and interaction with the site–security staff to modify firewalls at both PPPL and DIII–D.
The planned demonstration is 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). Along with preparing for the demonstrations, handouts and posters were prepared for presentation to the scientific community.
Visualization
The visualization working group independently explored a variety of options for sharing remote visualization with the aim of selecting one method that will be deployed throughout the fusion community. Sharing the IDL based visualization application ReviewPlus between two tiled walls was accomplished by using VNC (Fig. 1). This work is prototyping the Xplit environment discussed in the Collaboratory proposal. In addition to VNC, DMX and Xinerama are being investigated to bolster the VNC capability. The interaction classes of vtk were extended to support XMLRPC and thereby allowed the prototyping of collaborative advanced visualization applications compared to the 2D visualization capability of ReviewPlus. The Java based SciVis application was ported to Java 1.4 and used to display many graphs on a Tiled Display Wall.

Figure 1 Picture of ReviewPlus running in a shared mode on 2 tiled displays.
In early February General Atomics completed the construction of an 8–projector Tiled Display Wall that is operated by a 9–node Linux cluster. An agreement has been reached so that this system can be used by the Collaboratory for visualization research. Some initial shared visualization tests between two Tiled Display Walls were conducted between ANL and GA. Results are very preliminary but network speed and firewall access are issues that become critical when sharing displays over a WAN compared to sharing within a single labs LAN.
Another thrust this quarter was the creation of a prototype advanced 3D visualization application using NIMROD data stored in MDSplus (Fig 2.). This work required the creation of a Fusion Package for SciRUN that contains modules specific to the magnetic fusion community. Two NIMROD scientists visited Utah to provide feedback on this prototype. Their feedback was very positive and it is expected that this system will be demonstrated in April at the second fusion science meeting. Along with the creation of this visualization prototype, an evaluation has begun on the minimum hardware and software requirements for a desktop “visualization workstation” that could be used for the new software.
Work on Access Grid technology this quarter included some initial work to create an AG node at PPPL using their existing Tiled Display Wall. Firewall and network multicast issues have been identified that must be resolved before the node is fully functional. Responding to the

Figure 2 Visualization of DIII–D Pulse 10089 simulated by the NIMROD code
needs of the fusion community, a “reduced functionality” AG node has been specified and a prototype desktop solution has been created (Fig. 3). This system only generates two outgoing video streams in order to handle reduced network capabilities but still supports the same screen real estate and hands free audio with active echo cancellation as a full node.

Figure 3 Prototype Desktop Access Grid Node
· MDSplus client installed and working at ANL – this was done in order to have access to sample DIII-D data and for use with ReviewPlus.
· ReviewPlus & IDL– ReviewPlus is installed and working via MDSplus to display datasets. We have demonstrated large format display using VNC on tiled displays, demonstrated running on two different displays (distributed) in a shared access mode (See Figure 1).

Figure 2 Picture of ReviewPlus running in a shared mode on 2 tiled displays.
· Continued prototyping of the Xplit environment using VNC, and have started to investigate DMX and Xinerama as additional technologies for the prototype. Tests included running standard AG applications such as vic and web browsers along with fusion specific applications such as IDL and ReviewPlus.
· Development of first reduced functionality AG node (See Figure 2). Prototype of a desktop solution for the AG. The desktop solution offers the same screen real estate as a full node, same hands free audio with active echo cancellation, but generates only two outgoing video streams. Modifications have been made to the virtual venue server to support the selection of a subset of video streams in order to handle reduced network capabilities and to better balance the performance due to the resource limitations of a single computer implementation. Reduced functionality AG prototypes have been tested with both single and dual processor systems. The prototype required the additional modifications:
o Configured to support a maximum two outgoing video streams, using OpenMASH Vic as an alternative to the standard UCL vic in the AG toolsuite, since the UCL Vic does not work in a SMP multiprocessor environment
o Support added to the AG tool suite for both UCL direct draw Vic and OpenMASH Vic.

Figure 2 Prototype Desktop Node
· Extended vtk interaction classes to support XMLRPC, allowed for the prototyping of collaborative visualization applications. Prototyped the delivery of interfaces over the network on demand using Python. This will allow for users of a remote visualization tool to acquire the interface on the fly.
· Attended the SciDAC Pis meeting in Virginia and participated in a Networks panel discussion (Schissel).
· A one–day visit to ANL was used to discuss both remote computing and visualization work for the Collaboratory (Schissel, Peng).
· A presentation was given at the Access Grid retreat in San Diego on the requirements of the U.S. Fusion Community for wide deployment of Access Grid technology. Specifically, a case was presented for a lower cost AG node due to limited funding by the majority of the 80 institutions that do fusion research in the U.S. (Schissel).
· An abstract summarizing work done by the Collaboratory Project was submitted and accepted for presentation at WACE 2002 (Schissel).
· The project web site (http://www.fusiongrid.org) was maintained including updates for the April demonstrations at the TTF and Sherwood meetings. Numerous people were added to the workers portion of the web site.
· A security and remote computing workshop was held at MIT to discuss the lessons learned from the SC01 demonstration and to put together a set of plans for the April demonstrations at TTF and Sherwood. This meeting also offered several project members their first tour of a magnetic fusion facility (Schissel).
· The majority of work this quarter was focused on creating a demonstration of secured grid computing for two April fusion scientific meetings. This represents the first step to deploying the production fusion computational Grid. The demonstration involves secured 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.
· A new Windows PC was purchased and configured for use as a Microsoft SQL Server relational database system. The DIII–D Run Management Database was installed for testing of Grid computing (Burruss)
· A Globus–enabled MDSplus data server was installed on the new PC using an Alpha release of Globus. Grid related SQL requests are securely handled via MDSplus without the need to coordinate Globus with Microsoft SQL server software (Burruss).
· An OSF Unix computer was donated to the project by the DIII–D National Fusion Facility for use as a Globus–enabled MDSplus data server test platform for secured Grid computations (Burruss).
· MDSplus storage of TRANSP results was upgraded to allow for grid–based computation. This included the storage of TRANSP input quantities as well as information for the MultiGraph visualization package (Burruss).
· A GUI (PreTRANSP) for preparing and securely invoking TRANSP at PPPL was created. 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. PreTRANSP was also instrumented to log data transfer using netlogger for demonstrations at future scientific meetings (Peng).
· Designed the poster and handout material used at TTF 2002.
· Visits to Utah were made to offer feedback during the creation of a prototype visualization system for MacroStability data (Schissel, Kruger).
· Substantial improvements to the storage of NIMROD data in MDSplus were accomplished this quarter. NIMROD data is being used to test the prototype visualization application. This work allows the Collaboratory to evaluate the efficiency of large–scale data retrieval into an interactive visualization code (Flanagan).
· An agreement was reached with General Atomics to use their new 8–projector Tiled Display Wall (16’ x 8’) for shared visualization research being conducted by the Collaboratory. This represents a substantial increase in capability for DIII–D with no cost to the Collaboratory Project.
A.4 M. Thompson for the Lawrence Berkeley National Laboratory
The main LBL accomplishment of this quarter was the beginning of the integration of Akenti authorization with the Globus job manager. The focus was to limit access to run TRANSP code on the PPPL servers to only those users who had been specifically granted such rights. A rudimentary integration with the job manager start-job method was completed and used during the remote TRANSP demo at TTF.
Akenti code was initially installed on a test machine at ANL along with the latest version of the Globus job manager code. A call-out to Akenti for authorization was added to the job-manager code just before it would normally execute the specified job. If the reply from the authorization call was positive, the job execution would continue, otherwise an error would be returned to the user. The integration required placing all the public credentials of the collaboratory members in a local credentials directory, since the two CAs we are using, Globus and DOEScienceGrid, do not currently publish user certificates; setting up policy files for a test resource; and coordinating the function call and error returns with the job-manager.
Once the Akenti/Globus integration worked at ANL it was ported over to a server machine at PPPL where it could be used to authorize attempts to run the TRANSP code on that machine. At this point the policy files are very rudimentary, but even so it was necessary to add a resource name mapping functionality to Akenti to map the full path name of the executable file which is seen by the job manager to the type of resource names that Akenti expects.
I continued to work with the DOEScienceGrid Certificate Policy group to define a CP/CPS that will support the Fusion collaboratory's needs for certificates. Most of the collaboratory members now have credentials and are using them to run TRANSP jobs and for some MDSplus interactions.
Participated in the January workshop at MIT and the weekly phone conferences.
During this period progress was made in prototyping the addition of secure authentication and authorization to the remote access capabilities of the MDSplus data handling system used widely in fusion energy research. MDSplus provides scientists with the ability to access fusion research data located essentially anywhere accessible via the Internet. A scientist can run an analysis program that pulls in data from numerous fusion experiments anywhere in the world and write the analysis results into a remote data store. As the use of this capability has grown it has become increasing apparent that a sophisticated method of providing secure authentication and authorization was needed to ensure the integrity of the research data was maintained and access to it was limited to authorized users. In pursuant to this goal, the client/server communications used in the MDSplus remote access services were converted to use Globus I/O services. Globus provides secure authentication and authorization based on pass phrase protected X509 certificates. Communications between the client and server can be further secured by data encryption.
During this quarter the MDSplus server and client code modified to use secure communications with “globus_io” routines was ported to additional computer platforms used by the fusion community. The original prototype done in the previous quarter was implemented for the Linux operating system. Both the C-Mod experiment at MIT and the NSTX experiment at PPPL use data servers based on the Compaq OpenVMS operating system. The DIII-D experiment in San Diego use servers based on the Compaq Tru64 operating system. All three sites also maintain relational databases containing electron logbook information and physics results using Microsoft SQL Server on the Windows platform. Work was performed this quarter to add security to remote access to this data.
The Globus enabled MDSplus client and server software was successfully ported to OpenVMS, Tru64 and Windows. MIT also ported the Globus libraries used by MDSplus to the OpenVMS platform.
Access to the SQL Server relational databases was originally implemented using the native remote access protocol of SQL Server called Dblib. To provide Globus authentication to remote access to SQL Server data, an MDSplus gateway to SQL Server was implemented. In this way, the client software opens a secure connection to an MDSplus data server on the Windows system with the SQL Server database. Queries to the database are forwarded over this secure connection. Additional work will be required both in Globus development and MDSplus data server code to be able to provide secure MDSplus data servers as native Windows services. This prototype version requires a user on the Windows machine to manually run the MDSplus server application.
A.6 D. McCune for the Princeton Plasma Physics Laboratory
The past quarter has been spent on developing apps. for downloading data from the MDSplus server and reading it to SCIRun. Once into SCIRun we have developed several networks for interactive 3D visualization and exploration.
· Created an application to download data from the MDSplus server at GA and other locations. This allows data to be down loaded and store locally for analysis. Local storage is needed when accessing a data set repeatedly since network traffic may make downloading impractical at times. This data can then be used by SCIRun and other apps. The app was based upon some example code from the NIMROD group and is freely availible..
· Established a Fusion package for SCIRun. The package contains modules that are specific to the fusion project. This package will continue to grow as more modules are developed. Currently the package includes for following modules:
o MDSplusFieldReader - this allows downloading of data directly from the MDSplus server on a slice by slice basis into SCIRun.
o FusionFieldReader - Reads previously downloaded data from the MDSplus server that is stored locally.
o MeshBuilder - Builds a mesh specific to data that is in a cylindrical coordinate system. Since there is a large amount of data this module allows the user to pair down the data for fast interactive viewing.
o Two other misc modules for analysis, Radial Differential Module and a VectorMagnitude module.
· Created several networks for visualizing (Fig. 3) the Pressure, B Fields, and V Fields. These nets use various techniques ranging from isosurface and volume rendering for scalar data (pressure) to streamlines for vector data (B Fields). These will be presented at the Sherwood conference.
· Created a network that allows the 3D fuion data in SCIRun to be read directly by the RealTimeRayTracer (RTRT). This allows for time varying field analysis of the data using the RTRT
· Hosted visited with Eric Held and Scott Krugger, NIMROD scientists at Utah State U and SAIC respectively. These interactions have been on going and have given feed back on our visualization work.
· Now have IDL on site so that we can look at the current state of ReviewPlus and see some of the current tools being used for data visualization.
· Started a process of evaluating different hardware and software configurations to specify a $2000 viz box. The goal is to find several configuations that give the most for around $2000. We are looking at CPU speed, Graphics Cards, OS, and memory.
· Speced out a low cost AG Node with Argone. It will be a three LCD panel display system, 2 cameras, and a personal echo cancellation device all running under a dual CPU windows 2000 machine. A Wacom drawing tablet for shared whiteboard and we'll replace some of the video cards with GeForce series cards for experiments in remote visualization.a

Figure 3 Visualization of DIII–D Pulse 10089 simulated by the NIMROD code
team contact info |
about the fusion grid | fusiongrid research |