Kevin Oberman Engineered Engaging Career at ESnet

June 13, 2011

Fortunately, Kevin Oberman’s career path has been marked by surprising events trumping his expectations and plans, leading to a 37-year career in networking and engineering first at Lawrence Livermore and then at Berkeley Lab as a senior network engineer with ESnet.

For example, he joined ESnet in 1994 when the network was still managed by NERSC and both were located at LLNL. The next year, DOE asked for proposals to manage the network and computing center, and Berkeley Lab’s plan was chosen over Livermore’s.

Once they heard the news, Oberman and his ESnet colleagues headed to the Matador, a local watering hole, to commiserate over the shocking news. They were sure the move couldn’t turn out well.

“In retrospect, the move was one of the best things that happened to ESnet and to me,” Oberman says.

To start with, ESnet and NERSC were funded by DOE’s Office of Science, making them relatively small players in Livermore’s predominantly classified research environment. But at Berkeley Lab, where ESnet was established as a separate department, the focus is on open scientific work. “LBNL is a much more comfortable, natural home for ESnet,” Oberman says. And the network is much more engaged in research itself these days.

For Oberman, the move meant that he would remain a UC employee and retain his associated benefits, including the retirement plan which he plans to start enjoying come July 1. In fact, it was the UC benefits package that led him to accept his first job out of college with LLNL. At the time, he had three offers. Sperry-Univac wanted him to travel around the country supporting computer systems at military facilities and was offering a so-so benefits package, while HP promised outstanding benefits along with a “tiny paycheck.” “I decided the UC fringe benefits and stability at LLNL were a much better offer.”

That was 1974, and Oberman was finishing his second associate degree—he has one in computer engineering and one in electronics—at Trinidad State Junior College in Trinidad, Colo. At the time, LLNL actively recruited from the school, which explains why ESnet is home to three Trinidad locals: Oberman, Jim Gagliardi and Joe Burrescia. At the peak, Livermore had 23 Trinidad graduates on the payroll – not bad for a southern Colorado town of 10,000 which based its economy on coal until area mines began closing down in the 1950s.

Oberman’s father was a radio technician for the Federal Aviation Administration with a ham radio in the basement. He had a second career as an appliance repairman. His mother was a bookkeeper for local firms, including one of the largest fireworks distributors. One of Oberman’s first jobs was selling fireworks from a shop adjoining the company office. His favorite? The Silver Salute, which made about the same bang as a cherry bomb or M-80, but was much flashier as it was loaded with magnesium.

In high school and college, he was also a member of the debate team, where he developed skills that he later applied as a radio disc jockey and then as a speaker at conferences and workshops.

At LLNL, Oberman began his career in the Engineering Department on the hardware side, designing interfaces, installing and maintaining computers and setting up disc and tape drives. Although he had taken some programming in college and wrote some assembly codes on PDP-8s and the Bendix (later CDC) G-15, he began getting more into the software side of things and learned assembly language for the PDP-11, Intel MCS8 and others so he could develop device drivers. His first foray into networking came in 1979 when he was asked to network five computers. He designed and installed the network — which ran at 1 Mbps — using dual triaxial cables, one to transmit and one to receive.  “I got them all talking to each other and was just pleased as punch,” he recalls. About the same time, Oberman heard about this cool new technology called Ethernet and started getting more into networking on the side.

Back to school plans derailed

While he had only planned to stay at LLNL four or five years and then go back to school to earn his bachelor’s degree, “It never happened, I was having too much fun.” His next step was more ambitious. He finagled funding to buy Ethernet cable— yellow-clad cables about half an inch in diameter and known as “fat yellow hose” — and spent his weekends wiring the entire Engineering building. Before long, he was the network manager for the Engineering Department.

By the early 1990s, however, his job had become “too cookie-cutter” and wasn’t holding his interest. He was looking around when a former boss of his informed him of an opening at ESnet. He applied, got the job – and had to start anew. One of his first lessons was the difference between managing a local area network in one building to working on a national network with connections around the globe. To start with, the protocols were different. Then latency emerged as a consideration – sending a signal to Europe took a lot longer than sending it down the hall, and this in turn affected network performance and tuning.

But all networks have one thing in common, he adds – they all have their own strange features. For instance, when NERSC converted its network connection with the Princeton Plasma Physics Lab, fusion scientists in New Jersey complained that the inherent latency gave LLNL researchers an edge in submitting their jobs. The solution? Livermore networking staff built a “satellite delay simulator” to equalize things.

But quirks can also be good. In 1988, ESnet and NASA had a leased 9.6 Kbps line connecting Fermilab with a NASA facility in Ohio. The line was just being kept until a replacement line was up. But when a fire in a Hinsdale, Ill., switch room knocked out nearly every national network (as well as O’Hare airport), ESnet and NASA stayed up thanks to the almost obsolete link.

Oberman himself is well connected in the networking community. He was a volunteer for many years with SCinet, the network team that supports the annual SC (Supercomputing) conference. That often meant stringing dozens of miles of fiber optics throughout various convention centers to deliver the needed bandwidth.

He has also been a key player in developing plans for the transition from the current Internet Protocol (IPv4) to IPv6, which will address a number of challenges, such as the rapidly dwindling number of IP addresses. ESnet has been leading the IPv6 conversion project for DOE, and Oberman and his successor, Michael Sinatra, have documented the best practices and developed a checklist for others to use. Sinatra will also take over his work with DNS, or Domain Name Service, for ESnet and DOE.

Over the years, Oberman has served on a number of networking committees and conference organizations, such as his five-year membership on the Internet2/ESnet Joint Techs conference Program Committee. He will continue to serve his four-year term on the North American Network Operators’ Group (NANOG) until his term ends in October.

Oberman has given a number of presentations at networking meetings and will give one last talk at the NANOG meeting during his last work week at the Lab. Right after that, he and his wife will kick off his retirement with a trip to Hawaii. Also on the retirement agenda are continuing to attend San Jose Sharks hockey games and finally getting around to taking in some SF Giants games at their ballpark.

He had planned to take off sooner, but when ESnet was funded to develop the 100 Gbps prototype network as part of the Advanced Networking Initiative, he delayed his retirement. But he’s not ready to completely pull the plug on networking — he’s already planning to attend the SC11 conference this November in Seattle.

--Jon Bashor