Workers: Proposal

Proposed PPPL Participation in SciDAC Fusion Collaboratory: Other than Remote Control Room (1/2 FTE per year for at least two years):

• Building/extending computational services (40%, 0.2 FTE/year+travel):
  • Transition from GT2 to GT3 on existing services and associated client sites (TRANSP).
  • Conversion of new codes (possibly: GS2, Gyro, TORIC) as services on PPPL Linux Clusters-depending on user requirements.
  • Support of conversion of new codes at remote sites; GS2 will run first at U. Maryland; TORIC will run first at MIT.
  • Develop ability to call TORIC service (at MIT) from TRANSP service (at PPPL); coordinate allocation/reservation of service resources.
  • Develop/improve methods for delivery of time-critical services (between-shots TRANSP: realistic deployment).
• Remote/Collaborative real-time monitoring and graphical trouble-shooting of computational services (ElVis) (20%, 0.1 FTE/year):
  • Develop php version of ElVis Collaboration server to work with clients outside firewalls.
  • Verify/improve ElVis interactive/collaborative performance over a wide-area network.
  • Test as method for sending results of between-shots services to the collaborative control rooms.
• Education & Documentation (20%, 0.1 FTE/year + travel):
  • Tutorial/training on conversion of Linux Clusters to support Computational Services (target audience: Unix systems engineers).
  • Tutorial/training on adaptation of codes as computational services (target audience: Applications scientists).
  • Tutorial/training - conversion of applications to make use of ElVis.
  • Tutorial/training - Globus client/server networks with low impact on firewalls.
• · Firewalls (20%, 0.1 FTE/year):
  • Install test (beta) Globus proxy server at PPPL (trusted machine outside firewall).
  • Integrate services (data, computation) with proxy server to allow access by off-site clients.
  • Trouble-shoot and evaluate performance.

M&S: $2500.00 (1st year only) is needed for a machine with sufficient compute power, memory, and storage to support development, testing and evaluation of FusionGrid data and computational services in test configurations prior to deployment (we need a machine independent of the machines that are now "live" and in production use; the use of transpgrid.pppl.gov during SC03 preparation was a mistake that led to disruption of live TRANSP services, affecting PPPL and remote users).

Travel: extension of FusionGrid services to European sites may require the direct presence of Collaboratory personnel overseas (PPPL expects request from IPP/Garching).

Proposed PPPL Participation in SciDac Fusion Collaboratory: Summary, Remote/Collaborative Control Rooms:

PPPL proposes to dedicate ½ FTE on the work scopes associated with the collaborative control room. Effort at this level is expected to be required for at least two years. Requirements will evolve.

• Constructing, testing networking, getting initial setup of user prototype system in NSTX control room: about 2 man months: 1/6 FTE. PPPL Display Wall to be used for early testing of new system components.
• Evaluating compression algorithms for heterogeneous connectivity (PCS and Utah software components); evaluating PCS window sharing / transfer software, as well as commercial software methods (e.g. Instant Messenger): 1/6FTE.
• User training and support: 1/12 FTE.
• Evaluation of A/G and other AV software: 1/12 FTE.

Test facility (offline, and live for DIII-D Collaboration) - PPPL Display Wall. Test facility (live) - NSTX Control Room.

Evaluate means:

Alpha test: staff to install, test, evaluate, trouble-shoot. If OK, then…
Beta test: train/support users; let users test and give feedback.

User interaction: NSTX team (S. Kaye et al.); DIII-D collaboration (R. Nazikian et al.). Both user groups are supportive of the initiative-we should be able to get letters of support. User satisfaction is the "ultimate" test of the technology.

M&S:

We will use the PPPL Display Wall (not the NSTX control room) for initial testing and evaluation of software and hardware. We need a budget for hardware for tests-machines and equipment that are not intended for production use and can be "hacked on"; testing ground for MAC, PC, Linux, and wireless connections-general portability of software, hardware, networks, drivers, etc.

~3K for MAC, PC, Linux box: $9K
~3K for networking and cameras, headsets, etc.: $3K

TOTAL M&S: $12K - 1st year. Recommend M&S contingency at PPPL, $5K/year for 2nd and 3rd years of the proposal. Our efforts will be slowed if we don't have some M&S flexibility.

Details Below-tasks referenced against requirements document.

7. Collaborative Control Room:

1. One to all within the Control Room
   • One scientist at a computer in the control room sharing their display to the control room tiled display wall.
   • No need to support audio/video here.
   • Supports Linux, Windows, MacOS X.
   • Fast and easy to use
   • Share X-windows & web browsers. Office applications lower priority.

PPPL will continue to implement a collaborative control room. This will involve

• Finish constructing the collaborative display room at PPPL with two 1400x1050 projectors.
• Testing/Optimizing an internal gigabit network of MAC's, Windows, and Linux machine.
• Evaluate different VNC compression algorithms for efficient collaboration.
• Work with PCS to evaluate and test their system to move windows from a PC/Mac to the shared display.
• Give tutorial sessions for using all of the features inside of the collaborative control room.

2. One outside to one inside the Control Room
   • One scientist at a computer outside the control room sharing their display to one scientist inside the control room.
   • Requires audio and video as well
   • Supports Linux, Windows, MacOS X.
   • Fast and easy to use
   • This will be a one-on-one interactive data analysis discussion

• Normal VNC will work in this scenario. PPPL can test compression algorithms.
• PPPL will test VNC reflectors-performance impact of heterogeneous network connectivity amongst clients and amelioration strategies.
• Test/Evaluate Bluetooth headsets.

3. One outside to many inside the Control Room
   • One scientist at a computer outside the control room sharing their display to many scientists inside the control room.
   • Requires audio and video as well
   • Supports Linux, Windows, MacOS X.
   • Fast and easy to use
   • This will be a one-on-one interactive data analysis discussion .

• PPPL will test VNC reflectors-performance impact of heterogeneous network connectivity amongst clients and amelioration strategies.

4. One outside to many inside the Control Room
   • One scientist at a computer outside the control room sharing their display to the control room tiled display wall.
   • Requires audio and video as well
   • Supports Linux, Windows, MacOS X.
   • Fast and easy to use
   • This could be an off-site person who is heavily involved in the experiment and they will be involved in discussions for the entire 8 hours.
   • Alternatively, this could arise from one-on-one discussions mentioned above where the topic of interest is deemed to be of enough general value that the discussion needs to involve the entire control room.

• PPPL will test VNC reflectors-performance impact of heterogeneous network connectivity amongst clients and amelioration strategies.
• Work on testing compression mechanisms-testing software supplied by CS team.
• Also evaluate public/commercial alternatives. For example, enable instant messenger to allow the one outside to communicate with the control room group.

5. One inside to one inside the Control Room
   • This will be accomplished by just walking around, no need to technology solution.

• For the group leader this is true, but for others a use of instant messenger, and having windows which can be shareable one to one, will be implemented.
• Test instant messenger to send image of plots. Train people to do this; evaluate effectiveness of this strategy.

6. Security
   • The same authentication and authorization mechanism that works for Grid computing must be integrated into these tools.

7. Shared display Floor Control
   • A significant amount of floor control by conversation can be accomplished.
   • What if two off-site people want to share to the control room tiled wall. Is this allowed, do we create a queue?
   • These issues need to be addressed and discussed. However, they are secondary to first getting the technology to work. Additionally, floor control solutions should be simple so as to not get into the way (we do not want to do research here).
   • Only edit your own application
   • No restriction on window placement and sizing.
   • Always on top capability
   • Multiple people can interact but limited to limitations above.

• Continue to test/evaluate/trouble-shoot PCS software inside the PPPL control room.

8. Audio needs to work well
   • In a big room (control room) can be a problem due to background noise. Our experience at SC03 is indicative of this problem.
   • Headsets in the control room as broadcasting is not an option. But need to be able to go from headset to talking with the control room staff.
   • Audio floor control is required.

• Test Microsoft A/V tools, AG tools, and VRVS: Compare and contrast tools. Bring tools to a select number of users for further evaluation. Let users decide which A/V tools will best serve him/her.

9. Browser based display of shot cycle, DAM, Logbook, etc. (that was demoed at SC03).
   • SC03 was not a good layout, more generic capability for all tokamaks.
   • Layout should be customizable to persons needs. Easy to use.

team contact info | about the fusion grid | fusiongrid research Last modified 12/12/03. Comments? webmaster