Event 18 Dec Bruce Allen, UWM and AEI History and Status
Event 18 Dec pre-history When LIGO construction began in 1994 it was known that searches for unknown pulsars would require huge (unaffordable) computing resources. In the mid -1990s showed one solution. was the most powerful computer on Earth. Many people independently had the idea of doing something similar for LIGO. For example in 1996 or 1997 I had a lunchtime conversation with Stuart Anderson (now head of LIGO Laboratory Computing). We agreed that it was technically feasible to do but we wouldn’t be able to publicize it or get people to use it! So we dropped the idea.
Event 18 Dec History of Einstein Jahr In early 2003, the American Physical Society (APS) was working with the International Union of Pure and Applied Physics (IUPAP) planning events for the “World Year of Physics 2005”(Einstein Jahr). Alan Chodos, Executive Director of the APS, asked his staff to come up with some ideas for events. James Riordon, Director of Media Relations for the APS (and a long-time user) suggested that APS could do something similar for gravitational waves. Chodos liked this idea. and told Riordon to look into it further, and see if he could find someone interested in working on it. In mid-2003, Riordon contacted several people working on gravitational waves, and working in the LIGO project. He proposed that APS would provide publicity and advertising if they could do the technical work.
Event 18 Dec History of my involvement I personally learned about the APS proposal for the first time at a telephone conference meeting of the Executive Committee of the LIGO Scientific Collaboration in late September Some internal study and discussion had already taken place, and the response was negative. There was too little time to prepare, a distributed computing search could not be “meshed” with existing analysis plans, the pulsar search code was designed for Unix not Windows, it would distract people from other ongoing efforts, there was not enough scientific motivation, etc. I thought that this was too good a scientific opportunity to pass by. Other influential members of the committee also thought it would be a wonderful opportunity for publicity and outreach. I volunteered to lead a effort, and the LIGO Scientific Collaboration accepted the APS proposal.
Event 18 Dec History of M.Alessandra and the pulsar search I was already participating in the LIGO pulsar search work, because the effort was led by my wife, Maria Alessandra Papa. Her team at AEI and UWM had already developed and used a search code based on optimal methods proposed by Jaranowski, Krolak, and Schutz.We thought that this search code could be adapted to James Riordon had talked with the US National Science Foundation about funding and they recommended that we get in touch with a Berkeley computer scientist named David Anderson, who had built and was now working on a related project called “BOINC”.
Event 18 Dec History of BOINC and David Anderson Marialessandra and I were intrigued but were suspicious because we had previously had difficulty in working with computer scientists. We had a long phone conversation with David Anderson in early October David seemed very knowledgable and “down to earth” and explained how his experiences with had led him to develop BOINC. We realized that David Anderson’s experiences with had taught him a lot about volunteer distributed computing, and it would be better to take advantage of this experience than to “do it ourselves”. So we decided to adopt BOINC for
Event 18 Dec History of first proposal to the NSF We needed people to work on For this, we needed funding. In January 2004, I gathered together people from AEI, the LIGO Lab, UWM, Cardiff, Glasgow, and Berkeley, to submit a funding proposal to the US National Science Foundation. We needed a better name. I proposed but in developing the proposal, Rejean Dupuis asked “Wouldn’t the name be much better”? The NSF proposal was submitted in February 2004 and rejected in July The reviewers were also not convinced that members of the general public would be interested in this project. We had a choice: try again to get funding, or just go ahead. We decided to just go ahead. “Suffice it to say, the project has been almost TOO well conceived and organized (again, I found the level of detail quite singular).”
Event 18 Dec History of development and launch We gathered volunteers: Reinhard Prix (AEI) David Hammer (UWM) Eric Myers (Vassar) Bernd Machenschalk (AEI) Badri Krishnan (AEI) Xavier Siemens (UWM) Greg Mendell (LIGO Lab) Steffen Grunewald (AEI) Teviet Creighton (LIGO Lab) and some others. Between July and December 2004 we worked very hard. We wrote the screensaver, developed the science application and validator, and started testing. There were lots of bugs and problems. Until then, BOINC had only been used seriously by ClimatePrediction.Net and by a test project. There was no screensaver interface or BOINC Manager for Linux or Macintosh. We helped develop these, and found and fixed a lot of bugs in the BOINC application libraries and BOINC server side software.
Event 18 Dec History of development and launch Starting in November 2004 we began to test using members of the LIGO Scientific Collaboration as testers. In January 2005 we began accepting test users from outside the LIGO Scientific Collaboration. We officially launched on February 19th, 2005, at a press event at the annual meeting of the American Association for the Advancement of Science. In August 2006 received funding from the US National Science Foundation. With MPG and NSF support, the project should improve significantly over the next couple of years.
Event 18 Dec How big is
Event 18 Dec User/Credit History
Event 18 Dec Current performance is currently getting 80 Tflops
Event 18 Dec Tflops is a lot of computers! For comparison purposes… World’s largest web hosting center is in Karlsruhe, Germany (owned by 1&1) Hosts more than 1/3 of the web sites in Europe Technical staff: 100 System administrators: 20 Construction cost: $28 M servers (total) Computer rooms: m 2 15 MW power consumption About 60 Tflops total
Event 18 Dec S3 results: no pulsars found All sky maps With fake signals No fake signals
Event 18 Dec Work currently in progress Postprocessing of the S4 results (Holger Pletsch) First S5 analysis (will be finished around the end of January 2007). The first hierarchical search of S5 data: code and data preparation is underway right now! We hope to have this running by the time that the first S5 analysis is finished. “Targeting” the search to the most likely regions of sky and frequency.
Event 18 Dec Users: THANK YOU!