Behind the scenes at Labs
Showing results for 
Search instead for 
Did you mean: 

How Rob Schreiber’s Supercomputer Paper Stood the Test of Time




Schreiber, 5th from left, with his co-authors

By Curt Hopkins, Managing Editor, Hewlett Packard Labs

In November, Rob Schreiber, Distinguished Technologist at Hewlett Packard Labs, and his co-authors, were awarded Supercomputing's Test of Time Award at SC15, the latest SC Supercomputing conference in Austin, Texas.

SC is the “International Conference for High Performance Computing Networking, Storage, and Analysis.” Their Test of Time Award recognizes “an outstanding paper that has appeared at the SC conference and has deeply influenced the HPC discipline.” 

This was only the third time in 25 years that the award has been given.

Schreiber was recognized for the paper ”The NAS Parallel Benchmarks - Summary and Preliminary Results,” which he co-authored with David H. Bailey, Eric Barszcz, John Barton, D. Browning, Robert L. Carter, Leonardo Dagum, Rod Fatoohi, Paul Frederickson, Tom Lasinski, Horst Simon, V. Venkatakrishnan and Sisira Weeratunga, and which was presented at the SC91 conference.

About Schreiber’s paper, the supercomputing conference board said, “It is a mark of historical impact and recognition that the paper has changed HPC trends.” The authors made a presentation at this year’s conference.

At the time it was written, Schreiber said, most supercomputers were “vector” machines. But Schreiber and others realized that “parallel machines would soon eclipse vector machines.” Parallel computing was going to be necessary, he said, to achieve further breakthroughs in scientific computing. And if parallel machines were going to dominate, the world would need new “realistic ways to measure computing performance” and allow users to “select the best machine for a given job.”

schreiber.jpgSchreiber’s paper provided a way to do just that.

The best way to benchmark those machines with all their architectural diversity and different software, said Schreiber, was to specify a problem and ask the computer manufacturers to provide an algorithm to solve it. With vector machines, the testers provided everything but the machine itself. This new approach let the new machines’ vendors create a “tuned implementation” for their own system. 

“We put in rules,” said Schreiber. “Programming methodologies would be available to customers, but could be adapted to take advantage of the best characteristics of their machines.”

At SC15, Schreiber and about half the original authors gave a presentation to the attendees. Or at least they were present while their most demonstrative member took the audience for a time-travelling trip down memory lane.

“We gave a terrific presentation,” said Schreiber, “and by we I mean Dave Bailey.”

Although like most scientists he remains focused on whatever challenge is in front of him, Schreiber said he was flattered by the award.

“I was tickled,” he said. “The paper has been cited more than 2500 times and had a much bigger impact than we thought we were going to have at the time. It was nice that the community viewed it as important. It was very pleasant reminder of the terrific people involved and how much we enjoyed working together.”

Schreiber’s boss was more direct.  

 “Rob’s work with the NAS parallel benchmarks team shaped over two decades of system designs,” said John Sontag, VP of Systems Research at Hewlett Packard Labs. “We’re proud of Rob and glad he’s gotten recognition for the lasting impact of his work. We’re glad he’s part of the Labs team and that he continues to set the industry up for the next generation of computing, including Labs’ current research on The Machine.”

About the Author


Managing Editor, Hewlett Packard Labs

Online Expert Days - 2020
Visit this forum and get the schedules for online Expert Days where you can talk to HPE product experts, R&D and support team members and get answers...
Read more
HPE at 2020 Technology Events
Learn about the technology events where Hewlett Packard Enterprise will have a presence in 2020
Read more
View all