A-Z Index | Phone Book | Careers

InTheLoop 01.26.1998

January 26, 1998


Our universe, which began with the Big Bang, will never come to a standstill or collapse in a Big Crunch, but will expand forever, according to findings announced earlier this month by Saul Perlmutter, leader of the international Supernova Cosmology Project and a member of the Center for Particle Astrophysics based at Berkeley Lab.

Using several ground-based telescopes plus, more recently, the Hubble Space Telescope and a NERSC computer, the Supernova Cosmology Project has determined that the universe was expanding faster seven billion years ago (roughly half the time since the Big Bang) than it is today. Although expansion has slowed, the deceleration is not enough to suggest that gravity can bring outwardly rushing galaxies and other celestial matter to a halt.

"On the basis of both the ground-based data and the new Hubble data, we find evidence for a universe which may ultimately expand indefinitely," Perlmutter said.

The evidence comes from observing Type IA supernovae in very distant galaxies. To look at a distant object in space is to look into the distant past. To measure that distance, astronomers use "standard candles," objects whose intrinsic brightness is the same wherever they are found. Type 1A supernovae at their maximum brightness can be brighter than entire galaxies, bright enough for their light to have traveled billions of light-years and still be visible.

While the evidence for the findings came from astronomical observations, Perlmutter's team also used one of NERSC's T3E supercomputers to double-check some of their work. For example, the team must compare the light from nearby supernovae with that of the distant ones. To make meaningful comparisons, the light measurements from the more distant supernovae (which have been shifted to the red part of the spectrum due to the expansion of the universe) and the closer ones (which are in the blue) were altered slightly to examine the effects of dust along the line-of-sight to the supernovae and slightly different explosion scenarios. Then they were compared to make sure the team's observations matched their theoretical calculations. Because the measurements involved readings taken many times over a 60-day period from 40 supernovae, making the comparisons "is a task you only want to send to a supercomputer," says Lab postdoctoral fellow Peter Nugent.


Gene Logic Inc., which has licensed the Object-Protocol Model data management tools developed by members of the Scientific Data Management Research Group, officially opened their Berkeley office with a reception last Friday. The event was also something of a reunion, as the office is staffed by former CS employees Victor Markowitz, Amy Chen, Anthony Kosky, Ernest Szeto, Thodoros Topaloglou and Bea Edwards.

Gene Logic plans to use the OPM tools to integrate existing databases with new information being generated. The company is primarily involved in the pre-clinical discovery business, identifying new medical targets for drugs, new compounds for use in drugs and illnesses for which there is currently no satisfactory treatment.

Gene Logic President and CEO Michael Brennan credited Victor's group with inventing unparalleled technology that will allow information to be integrated in ways never before possible and help the company improve medical treatments. Chief Scientific Officer Keith Elliston said Gene Logic would continue looking for ways to work with the national laboratories, especially Berkeley Lab.

NERSC Division Director Horst Simon noted that the OPM tools are also benefiting high energy and nuclear physics research, and that when he was thinking of joining the Lab two years ago, the potential of OPM was already well known in the scientific community.

Group Leader Arie Shoshani, whom Victor thanked for nurturing the group and its work for five years, presented the team with a plaque bearing a picture of Wile E. Coyote, the cartoon character adopted as the OPM mascot for his never-say-die attitude.


Kathy Yelick, who holds a joint appointment at UC Berkeley and NERSC, will teach Computer Science 267 on campus during the spring semester. The class, to meet Wednesdays and Fridays from 12:30-2 p.m. in 310 Soda Hall, is intended to provide graduate students in Computer Science other fields with a broad understanding of the application of parallel computers. With hands-on experience with parallel programming as a central theme, it addresses architectures, languages, tools, environments, methods and techniques. It also ties into application domains where parallel computing is essential. A major component of the course is a parallel programming project, conducted in teams of three.


Jed Donnelly, who made the move from LLNL to Berkeley Lab last December, will be part of a three-person panel discussing "The Computing Environment at Livermore in the 1970s" at Sun Microsystems in Palo Alto on Thursday, Jan. 29. The presentation is sponsored by Sun Microsystems and the Computer Museum History Center.

Back 35 years ago, if you took delivery of a new supercomputer by Seymour Cray (like the CDC 6600) you got a box with boards and wires and you had to write an operating system yourself. If you wanted a comprehensive network delivering services to a community of users, you had to do that job yourself too. You had to build the network hardware, write the network software, and integrate it all with your own custom operating system on a wide variety of machines from minicomputers to supercomputers. The result was a complete computing environment built and tailored to serve the needs of a particular group of users.

Such was the environment at Lawrence Livermore in the 1970s. It wasn't possible in the 1960s, and it had already become obsolete in the 1980s, with the arrival of generic network hardware and software. In fact, the 1970s may have been the only time in the history of computing that such an entire custom dedicated environment was possible or practical.

Dick Watson will begin the program with an overview of the period. John Fletcher will take us back to the earliest times, since he came to Livermore first. And Jed Donnelly will discuss operating system developments with the network extensions. The talk will begin at 5:30 p.m. Thursday, Jan. 29, in the Caspian Room at Sun Microsystems, 901 San Antonio Road, Palo Alto. Anyone interested is welcome to attend.


Although the Super Bowl unofficially marks the end of the ever-lengthening holiday season, you can still relive one of the high points of the festivities -- the Computing Sciences Holiday Party. Photos taken at the Dec. 18 event are now available for viewing from the Computing Sciences staff page.