| John Towns / Meeting Proposal | |
| Title: | Klaus Schulten Mtg | |
| Location: | 2100 NCSA | |
| Guests: | Jay Alameda, Tim Cockerill, Mike Pflugmacher and John Towns | |
| Schedule: | Wednesday, September 19, 2007 2:30 PM to 3:30 PM | |
| Agenda: | X-Envelope-From: jtowns@ncsa.uiuc.edu X-Mailer: QUALCOMM Windows Eudora Version 6.2.5.6 Date: Tue, 18 Sep 2007 11:04:26 -0500 To: Klaus Schulten <kschulte@ks.uiuc.edu>, Joyce Lucas <joyce@ks.uiuc.edu> From: John Towns - NCSA Cog <jtowns@ncsa.uiuc.edu> Subject: Re: Fwd: very hot Abe project: NAMD-millisecond Cc: "Phillips Dr. James C. (Jim)" <jim@ks.uiuc.edu>, Stone John <johns@ks.uiuc.edu>, Peter Freddolino <petefred@ks.uiuc.edu>, "Laxmikant (Sanjay) Kale" <kale@uiuc.edu>, Amber Moore <amber@ncsa.uiuc.edu>, Mike Pflugmacher <mikep@ncsa.uiuc.edu>, Tim Cockerill <cockeril@ncsa.uiuc.edu>, Jay Alameda <jalameda@ncsa.uiuc.edu> X-Null-Tag: 2b3908690abd3fda7a806100b1cecce1 X-NCSA-MailScanner-Information: Please contact help@ncsa.uiuc.edu for more information, zanamavir.ncsa.uiuc.edu X-NCSA-MailScanner: Found to be clean X-Envelope-To: amber X-Deliver-To: amber Joyce-- I can meet between about 2:30-4:30 pm. Please coordinate with Amber. Amber-- We should have Mike Pfl*, TimC, Jay Alameda, if they can make it. -John At 10:47 AM 9/18/2007, Klaus Schulten wrote: Dear Joyce: Could youplease arrange a meeting with John Towns and John, Jin, Peter, and me tomorrow afternoon at NCSA. It has very high priority and other meetings can be cancelled. Thanks, Klaus Begin forwarded message: From: John Towns - NCSA Cog <jtowns@ncsa.uiuc.edu> Date: September 18, 2007 10:21:48 AM CDT To: Klaus Schulten <kschulte@ks.uiuc.edu> Cc: Thom Dunning <tdunning@ncsa.uiuc.edu >, Rob Pennington <robp@ncsa.uiuc.edu>, "Phillips Dr. James C. (Jim)" <jim@ks.uiuc.edu>, Peter Freddolino <petefred@ks.uiuc.edu>, "Laxmikant Kale (Sanjay)" <kale@uiuc.edu>, Amber Moore <amber@ncsa.uiuc.edu> Subject: Re: very hot Abe project: NAMD-millisecond Klaus-- NCSA is certainly interested in principle. There are multiple facets we will need to understand and we will need to plot out a course that is achievable and can be funded as needed. I would like to meet, but would need perhaps a bit more background on what you would be looking for from NCSA so I can pull in the right folks. There is a small window tomorrow afternoon for me to meet. Thu-Fri this week are already packed. Mon-Tue of next week look very good. -John At 09:28 AM 9/18/2007, Klaus Schulten wrote: Dear John: My group began to work on a very hot Abe project that I like to discuss with you at your earliest convenience, best in a group of NCSA staff and my own developers. On our side my three best developers are closely involved: Jim Phillips, John Stone, and Peter Freddolino as well as key people from Sanjay. NAMD has been held back in its capacity by being an all-platform program, i.e., it has the same features from my laptop to all teragrid machines. Our competition (D. E. Shaw) forces us to shed some of the overhead that comes with NAMD being all-platform. We want to build the leanest and meanest NAMD-millisecond that runs so fast that it can achieve eventually millisecond simulations. The project aims at (i) pacing NAMD by avoiding all overhead (like CHARM and MPI) and optimizing code specifically for Abe processors; (ii) carry out the longest ever all-atom molecular dynamics simulations in two categories, all-atom for protein folding, (ii) coarse-graining for cell-size mechanics (bacterial swimming). We want to beat D.E. Shaw in his game of doing millisecond molecular dynamics simulations. We do not expect to get anytime soon, i.e., in 2008, to milliseconds as presently we would need 10,000 days to do an all-atom simulation that long. But we want to simulate as long as possible by the middle of 2008, being confident we can accomplish a 10 millisecond simulation after successes with recoding, i.e., by running simulations 100 days and speeding NAMD up by a factor ten. This factor is quite achievable according to our estimates, and with some new insights, luck, and tricks possibly more. Naturally, for this project we need access to the top NCSA experts and we need to run simulations continuously on rather small (a few hundred) partitions of Abe. Doing long, i.e., millisecond, molecular dynamics simulation is the holy grain in computational biology, opening the path to fold proteins and thereby predict their structure. The project, accordingly, has a huge appeal. It also fits well with our IACAT participation (both the one in the Sanjay and the Wen-mei teams), with our pending petascale proposal, and with a planned (under Wen- mei's leadership) "Expedition in Computing" proposal. Please let me know if you and NCSA is interested in principle and please suggest a contact through which we can schedule a meeting. Many thanks, Klaus |
|