Changing of the guard: Tom Bettge to become new SCD associate director

Bernie O'Lear and Tom Bettge, current and future SCD associate director

by Lynda Lester

Effective January 2003, Tom Bettge of NCAR's Climate and Global Dynamics Division will become the new associate director of SCD. Lead software engineer for the Parallel Climate Model in Warren Washington's Climate Change Research Section, Tom will take the reins from Bernie O'Lear, who has been associate director since 1994.

"We worked out a transition plan with Warren Washington's group," says SCD director Al Kellie. "Tom began his involvement with SCD starting with the planning for the new IBM supercomputer bluesky. He will be fully at the helm when Bernie O'Lear retires at the beginning of next year."

"Tom comes to us with a 25-year history at NCAR and a strong user background," says Bernie. "He'll bring a solid user perspective to this slot."

In explaining why he was interested in the position, Tom says, "I was looking for an opportunity to move to a different level in applying the skills I have. I have a good deal of experience with SCD mission-oriented projects and have always had a great working relationship with people in SCD. I thought it would be exciting to work more closely with them to meet and provide direction for the continual changes we face in supercomputing. We're now entering the world of remote data access and moving large datasets over networks."

He smiles and adds, "I don't know which is more difficult—to leave a legend or to replace a legend." Tom will no longer be working for the Warren Washington, an icon in the national science community—and he will be taking over from the legendary Bernie O'Lear, a presence at NCAR since before the Mesa Lab was even built.

Bernie's 38 years

Bernie began working at NCAR in 1964. His office was in Cockerall Hall on campus at the University of Colorado in Boulder, when most of NCAR was located in CU buildings on 30th street. He'd produce source code on coding sheets, which were delivered to "keypunch girls" for data entry on punch cards. To build a program, Bernie would assemble punch cards (each of which contained a line of code) into decks, then send them for submission to the CDC 3600, an early computer from Control Data Corporation. While waiting for the output, he and the other coders would play ping pong in the basement. (For a brief historical note on the Cockerall Hall ping pong table, see the February 1998 Staff Notes story, "It happened here.")

Bernie wrote code for the first General Circulation Model (GCM) of the atmosphere, and in 1965 built the first computer movie—of the GCM—at NCAR. Initially an applications programmer, he later became a systems programmer and helped develop NCAR's first Fortran compiler.

Bernie wrote the Request for Proposal (RFP) for the first Mass Storage System at NCAR, which resulted in the procurement of the Ampex TeraBit Memory (TBM) system in 1977. He started the SCD Mass Storage System Group and was a key person in founding the IEEE Computer Society MSS Technical Committee.

Project manager for the University Satellite Network (USAN) in the mid-1980s, Bernie became manager of SCD's High-Performance Systems Section in 1986, expanding activities of the group in 1991 to include networking. He became associate director of SCD in 1994.

Bernie, who is now in his 38th year at NCAR, laughs when he cites "O'Lear's Indispensable Rule": "Get a glass of water. Put your finger in it and take it out. If the hole stays, you're indispensable."

Tom's track record

"I've run on every machine at NCAR except the CDC 3600 and the early IBM that was being used when the Mesa Lab was under construction," Tom says. "From the user perspective, I've seen everything."

Tom came to NCAR in the fall of 1976, joining what was then the Numerical Weather Prediction Group in the Atmospheric Analysis and Prediction Division—today's CGD. He worked on short-term weather forecasts, running code on the CDC 7600, and in the early 1980s wrote code for the MM2 weather model to validate regional forecasts.

His first job for Warren Washington was managing a host on MFENET, a network for the Department of Energy's Magnetic Fusion Energy (MFE) Computer Center—now the National Energy Research Scientific Computing Center (NERSC). MFENET was one of the precursors to the Internet. When an MFENET node was built at NCAR, Tom worked with SCD to maintain it.

"We'd been generating lots of data, running models on a DOE machine at the MFE center," Tom says. "We used to put the data on tapes and ship them back to NCAR. Then we got the host node and could bring data back over the network. It was considered high speed—the Internet didn't yet exist, it was well before T1 lines, and I remember we were getting a megabyte in one hour!"

Tom's job providing network support evolved into his becoming lead programmer for Warren's climate model effort. "We built what was really the first fully coupled climate model at NCAR in the late 1980s," Tom says, "bringing in atmosphere, ocean, and ice models at coarse resolution. We ran first on the CRAY1, then on the CRAY X-MP and CRAY Y-MP." Their model became the predecessor to today's Parallel Climate Model (PCM).

"We started putting together the PCM in 1996, building it in collaboration with SCD's Computational Science Section," Tom says. "The PCM targeted distributed memory, multiprocessor, massively parallel machines such as the CRAY T3D and T3E; it went in a different direction than the Climate System Model at the time, which targeted vector machines." The PCM has had broad success in the DOE Climate Change Prediction Program, he notes.

Tom has managed a number of projects in CGD's Climate Change Research Section and is currently co-investigator on a Scientific Discovery through Advanced Computing (SciDAC) project, involving terascale computing with a fully coupled climate systems model.

In conjunction with Argonne National Laboratory and CGD's Climate Modeling Section, Tom has been involved in designing and building a next-generation coupler. In the last seven years he has worked extensively on distributed cluster machines at NCAR, Oakridge National Laboratory, NERSC, and Los Alamos.

Interestingly, Tom worked with Don Middleton, currently manager of SCD's Visualization and Enabling Technologies Section, in the late 1980s. "Don and I go back to when he first came in the door," Tom says, noting that he and Don produced a climate change visualization in collaboration with the National Center for Supercomputing Applications. The project resulted in a videotape that was popular with the media for many years.

As part of the transition plan to assume his new responsibilities, Tom participated in recent the Live Test Demonstration of the new IBM supercomputer, bluesky, in Poughkeepsie, New York. Bluesky represents phase II of the procurement of NCAR's Advanced Research Computing System. For more information on bluesky, see "New IBM 'bluesky' system to be delivered 2 October 2002" and "ARCS bluesky system passes Live Test Demonstration."