Third Quarter FY 2003 Report - The National Fusion Collaboratory

Edited by D.P. Schissel

General Atomics (schissel@fusion.gat.com)

Overview

The National Fusion Collaboratory Project was reviewed twice this quarter. The DOE MICS office held a peer review at MIT while the DOE OFES funded Plasma Science Advanced Computing Institute's (PSACI) Program Advisory Committee held a review at Princeton University. Both reviews were positive on the impact the project is having on the Fusion Energy Sciences.

General accomplishments include:

General

Significant time and effort was spent this quarter planning, preparing, and attending the DOE MICS review of the NFC Project. Held over two days at MIT, the review included presentations, demonstrations, and a discussion session with the review panel. Review comments were very positive and also provided valuable suggestions on how to further improve the Project. In addition to this review, the DOE OFES funded Plasma Science Advanced Computing Institute¼s (PSACI) Program Advisory Committee Meeting held a review of the NFC project. This review examined the project exclusively from the standpoint of its impact on NFC science. Quoting from the PSACI PAC report, "The National Fusion Collaboratory has continued to deliver impressive technical results."

Substantial time was spent this quarter working within the U.S. on our country's strategy for participating in the ITER project. Culminating in a paper and presentation at the ITER U.S. forum, agreement was reached that a high priority for the U.S. was to lead the effort in data acquisition, data management, and remote participation for the entire ITER project. The NFC's team of people and technologies deployed are a central component of our country¼s resume in this area of work and was instrumental in placing this area of leadership high on the U.S. ITER task list.

The NFC project was involved in the DOE Science Networking: Roadmap to 2008 report. The Technology and Services Work Group was co‚lead by Bill Johnston and David Schissel.

The NFC project was involved in several meetings this quarter:

The Project web site continued to be updated as required.

Security/Remote Computing

This quarter the Akenti authorization package was integrated into GT2 GRAM for alpha testing on FusionGrid. A paper on this work was submitted to CHEP03 as well as another paper presented at the International Workshop on Grid Computing with the invitation for a more detailed follow‚up paper to be presented at Concurrency and Computation: Practice and Experience. In conjunction with this work, a dynamic account factory service and dynamic account service was designed and prototyped allowing a user to flexibly request, configure, and manage dynamic accounts for Grid execution on a need‚basis.

Work continued on a prototype of a network services infrastructure with a reservation‚based QoS interface module enabling user to run fusion applications as network services guaranteeing a certain QoS which will enable Fusion scientists to run codes between tokamak pulses.

Detailed design work has started on incorporating the capabilities of GridFTP or striped socket communication into the MDSplus mdsip client‚server protocols. The decision was made to use the next generation Globus I/O, called XIO, for this purpose. To facilitate this work, a specification for a set of client‚server programs was created that would demonstrate how to use the new API to provide the functionality required by FusionGrid. The substantially larger data throughput that will be realized by the completion of this task is critical to the successful usage of MDSplus by the Fusion simulation community.

The transition from the original doesciencegrid CA to the new DOEGrids CA has continued this quarter. Work included testing new certificates with a variety of Globus versions and configurations. Information on how to transition from old to new certificates has been documented. FusionGrid will be updated to the new certificates next quarter.

Visualization

This quarter saw continued development of collaboration technology including the creation of a Fusion Virtual Venue for AG meetings and the deployment of a small AG node at DIII‚D. Work also continued on the VenueVNC application that is a mechanism for integrating a VNC session into a Virtual Venue. This capability is critical for integrating the AG node into tokamak operations since scientists need to simultaneously share in real time audio/video and shared data analysis applications. In preparation for testing this new technology during tokamak operations, the small AG node was made operational at DIII‚D. Although the AG software and hardware were easy to install and configure, significant time was spent configuring DIII-D¼s network connection into ESnet. For future planning, when new AG nodes are added into the MFE community, the time and experience required for proper network configuration cannot be ignored. A second AG PIG node, dedicated for the C‚Mod control room, was ordered this quarter.

Tiled display walls have been designed for both the NSTX and DIII‚D control rooms. The NSTX system is a permanent 2‚tile front project system while the DIII‚D system is a prototype for evaluation. In conjunction with these designs, work continued on automatic alignment software so that control room systems are easier to install and maintain.

The ELVis visualization tool that is being used for TRANSP data display was enhanced this quarter. 3‚D surface plots with collaboration between display clients was implemented as well as integration with the legacy RPLOT graphical application. Data monitoring during a TRANSP run for simple f(x) signals was also implemented.

Work on SCIRun continues to move its infrastructure towards GUI separation that will allow among other features macro or super modules. The Teem software package was added to SCIRun to allow for better animation of time dependent fusion data.


Appendix A: Non‚Edited Reports from Individual Institutions

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

This quarter¼s effort focused on continued development of collaboration technology, focusing on putting a Fusion Virtual Venue in place. ANL has created a „Fusion Collaboratory Lobby¾ on its ag-2.mcs.anl.gov venue server. One room has been added to support the work demonstrated at the review and described below. Currently only ANL is administrator for the venue, capable of adding new rooms, but we would like to see others adding needed meeting spaces. The use of a pure 2.0 venue server does not for users to participate from 1.0 venue clients, this will help motivate the move to 2.0 and allow for more feature rich spaces to be constructed.

In addition to the work described above ANL has worked to harden and simplify the VenueVNC application. VenueVNC is a set of scripts and mechanisms for integrating a VNC session with a Virtual Venue. There are two components, a server that wraps the execution and termination of the VNC Server and provides configuration and authentication information to the Virtual Venue, and a client that provides execution and termination management of VNC Clients, as well as automated authentication, under both Linux and Windows. Using VenueVNC, a user can enter a Virtual Venue and join a VNC session without any prior knowledge of passwords, host names or display numbers.

VenueVNC relies on the standard VNC communication mechanisms for the actual client session, taking advantage of the high speed and reliability inherent in the well-tested protocol. VenueVNC takes advantage of the Shared Application API within the Virtual Venue framework provided by the Access Grid Toolkit to provide connection and authentication information in a secure context, secured by the Globus security mechanisms inherent in the AG Toolkit. These Fusion venue on AG 2.0 venue server currently has Fusion Lobby and one room. Currently supports VNC application demonstrated at the project review demo. Virtual Venue integration mechanisms are what allow the client component to connect to a requested VenueVNC session without knowledge of the session password or the hostname/display number usually required to initiate a VNC client session. The password in a standard VNC session is widely regarded as the weakest link in VNC¼s security model. With this in mind we¼ve included measures to minimize this security risk by using the maximum length password and by randomly selecting the password from the complete list of valid characters, many of which are non-printable and not easily producible on a standard keyboard. Also, we make use of the inherent security and encryption that Globus provides to initiate the VNC session providing a secure tunnel for the password authentication. Also during this quarter the ANL continued to hold teleconference¼s with Princeton Computer Science working on long range planning and cross use of software.

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

In the last quarter we made contributions in the following areas:

  1. Work on authorization and enforcement
    1. We worked with our collaborators at LBNL to help integrate Akenti authorization into the alpha version of the GT2 GRAM authorization package delivered last quarter. We also improved documentation for this feature available at at http://www-unix.globus.org/gram_authz/. The package is expected to be integrated into a stable release of Globus after a period of alpha testing by our collaborators.
    2. We presented a paper describing and analyzing the implementation of the GRAM authorization package at the International Workshop on Grid Computing held in conjunction with the Middleware 2003 conference. An extended version of this paper was invited for publication at Concurreny and Computation: Practice and Experience.
    3. We contributed to writing a paper with our LBNL collaborators on the integration of Akenti authorization into the GT2 GRAM authorization package.
    4. We designed a dynamic account factory service and dynamic account service allowing a user to flexibly request, configure and manage dynamic accounts for Grid execution on a need-basis and prototyped it in GT3. The existence of such mechanism allows the users to obtain accounts on a by-need basis and alleviates the administrative burden on system administrators. It also provides for more effective enforcement of the authorization techniques developed in the work described above.
  2. Work on QoS-based application servers
    1. We continued our work on a prototype of a network services infrastructure with a reservation-based QoS interface module enabling users to run fusion applications as network services guaranteeing a certain QoS which will enable Fusion scientists to run code between pulses. We are currently developing a code using statistical methods to estimate end-to-end execution times for realistic fusion simulations
    2. We also contributed the requirements and experiences of this project to the OGSI-Agreement currently under consideration at the Global Grid Forum. This effort will lead to the development of Grid standards, eventually enabling Fusion scientists to obtain agreements for QoS-based execution of their codes.
  3. Collaboration, Papers, and outreach Demonstrations
    1. We contributed to writing of a paper summarizing the Computer Science aspects of our experiences, in particular the release of a first Fusion production service, submitted to the Grid2003 Workshop to be held in conjunction with SC03.
    2. We contributed to writing of a paper summarizing the Fusion user experiences relevant to using the TRANSP service, submitted to the IAEA workshop.
    3. An abstract for a talk at the IAEA conference, describing Grid technology to Fusion scientists was submitted and accepted.
    4. We helped with preparation for, and participated in the review of the project held in May.
  4. General interactio
    1. We worked with our MIT collaborators on designing ways in which GridFTP and the Globus XIO package could be used in Fusion
    2. We provided technical support, assisted with debugging and troubleshooting of Globus installation and associated technologies. This involved providing explanations and technical discussions of Globus functionality as well as providing examples, help with troubleshooting and answering questions.
    3. We also interacted with our collaborators through teleconferences, meetings, discussions, etc.

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

General

Security/Remote

Visualization

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

The transition from the original doesciencegrid CA to the new DOEGrids CA has been in progress. I helped test the new certificates with a variety of Globus versions and configurations. Information on how to transition old certificates to new ones was added to the Grid Identity web page as well as the new CA certificates and signing policies. The NFC servers were all updated with the new certificates. New members of the fusion collaboratory are now getting and using certificates from the DOEGrids CA.

Continued to issue certificates and answer questions from new users on how to get and use certificates to access NFC services.

Continued to work on the Globus/Akenti integration including upgrading Akenti to newer versions of external packages to attempt to resolve a problem in using the Akenti code in the Globus environment. We have been trying to link the Akenti libraries directly into the Globus Job manager to eliminate the need to run a separate Akenti server, but are having difficulties with missing C++ standard functions. We may need to go back to the previous model of just linking against a client interface and running the Akenti server separately from Globus. Finished and submitted the CHEP03 paper about this integration.

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

We have begun discussions with the Globus developers about incorporating the capabilities of GridFTP or striped socket communications into the MDSplus mdsip client ‚ server protocols. During the discussion it was decided that the preferred approach would be to use the next generation Globus I/O, called XIO, for this purpose. This new I/O architecture enables protocols to be layered on top of one another by adding special handlers or drivers. To facilitate the migration of the Globus enabled version of MDSplus mdsip, we created a specification for a set of client server sample programs which would demonstrate how to use the new API to provide the functionality required by the Fusion Grid. The Globus team will implement these sample programs. This will ensure that the functionality needed by the Fusion Grid is available in the standard API and that no special coding or calls into internal Globus routines are needed as was the case with the original implementation of the Globus enabled MDSplus communications.

The demonstration suite will consist of client and server applications that use three different levels of communications. The first level would be a simple client which uses tcpip to connect to a server activated via tcpip. The client would send a block of data to the server (approximately 1 megabyte in size) and the server would send a similar sized block of data back to the client. The second level would be essentially the same except the client would send the user's credentials to the server and the server would send the host credentials back to the client. The server would then inherit the users credentials. The third level would add the socket striping to level 2. The client and server would communicate over multiple sockets for higher throughput rates.

Work was begun on updating PostTransp, the last of the three TRANSP utilities used by researchers to manage runs. (The other two are PreTransp, which is used to prepare data for runs and MultiGraph, which is a TRANSP output viewer.) These utilities are written in IDL and provide easy to use graphical interfaces for TRANSP users. PostTransp will be modified to a) run under linux, b) use the secure connections provided by gsi and mdsip to provide access to runs at any of the three facilities, c) interact with the new run management database.

The MIT PIG node (personal access grid) was operated with GA and Argonne in a 'bridged mode' using a multicast bridge located at Argonne. With this configuration the negative impact on our LAN was minimal. We plan to set up a standing bridge for our venue, and eventually install hardware locally to act as a multicast bridge. With the help of Argonne personnel we installed the AG version 2 software on our PIG node. Using this we demonstrated a remote meeting with shared visualization during the NFC review meeting in May. The AG2 tool set includes secure user identification, local venues, local bridges, and shared workspaces. We plan to try it out with NFC members other than Argonne as soon as possible. We have begun the procurement of a PIG node for the C-Mod control room. This will include a computer and video projector for the front of the room. This will allow outside collaborators to interact with the session leader and physics operator during tokamak operations. The same hardware will use the display "push" technology developed by Princeton University to allow sharing of displays during the tokamaks operation.

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

PPPL continued to develop and test collaborative computational services (TRANSP) and visualization services (ElVis). PPPL participated in the May, 2003 NFC review, with presentations covering the PPPL work given by D. McCune.

Collaboratory Computational Services:

Ronald V Bravenec ........ rbravenec@mail.utexas.edu

Perry Phillips ........... p.phillips@mail.utexas.edu

 

Peter Martin ............. pem@ipp.mpg.de

 

Thomas W. Fredian ........ twf@psfc.mit.edu

Catherine L. Fiore ....... fiore@psfc.mit.edu

Alan Lynn ................ lynn@psfc.mit.edu

William Davis Lee ........ w_davis@psfc.mit.edu

 

Charles M. Greenfield .... Chuck.Greenfield@gat.com

Justin Burruss ........... burruss@fusion.gat.com

Masanori Murakami ........ murakami@fusion.gat.com

Toshihiro Oikawa ......... oikawa@fusion.gat.com

 

Qian Peng ................ peng@fusion.gat.com

 

Alex Pankin .............. pankin@fusion.physics.lehigh.edu

 

Robert G Andre ........... randre@pppl.gov

Robert Budny ............. rbudny@pppl.gov

Christiane Ludescher ..... cludescher@pppl.gov

Douglas C McCune ......... dmccune@pppl.gov

Theodore B. Terpstra .... tterpstra@pppl.gov

Lew Randerson ............ lranderson@pppl.gov

 

Balamurali Ananthan ...... bala@mcs.anl.gov

Sam Lang ................. slang@mcs.anl.gov

Prathibha Deshikachar .... deshpra@mcs.anl.gov

 

Mary R. Thompson ......... mrthomps@lbl.gov

PPPL Contributions to Development of Collaboratory Visualization Capabilities:

ElVis:

NFC Review:

Collaborative Display:

A.7 G.Wallace for the Princeton University Computer Science Department

A.8 Sanderson for the University of Utah Center for Scientific Computing and Imaging

Much of the focus has been on SCIRun infastructure as we are moving towards the GUI separation which will allow among other features macro or super modules which are the highest priority. At the same time we have added more functionality into the Fusion Package by integrating the Teem Package which will allow for better animation of time dependent data. We continue to work on stability issues. We have also integrated the HDF5 into SCIRun which will allow more data reading capability.

Research has continued on the new representation for vector data. A technical report has been written and publically availible. Also a viewpoint paper has been submitted and accepted to IEEE Computer Graphics and Applications (Sep 2003) on Visualisation Uncertainty which features some of this research.