Five years ago, as a Ph.D. student at the Massachusetts Institute of Technology, Sang-Woo Jun co-authored a paper suggesting a novel way to boost system efficiency without sacrificing performance. The paper, “GraFboost: Using Accelerated Flash Storage for External Graph Analytics,” discussed a way to design algorithms and physical systems using solid-state secondary storage (such as solid-state drives) instead of the main memory (dynamic RAM) for I/O intensive graph analytics.
“We had shown that our system can improve the cost and power efficiency of the system by more than four times, often without any significant loss of performance,” says Jun, now an assistant professor of computer science in UC Irvine’s Donald Bren School of Information and Computer Sciences (ICS). “As memory capacity and performance scaling is tapering out, similar to processor performance, I believe our approach is one of the more promising ways to continue improving scalability in the future.”
Highlighting the significance of this work, the paper is now part of a 25-year retrospective for the 50th edition of the International Symposium on Computer Architecture (ISCA 2023), the premier forum for new ideas and experimental results in computer architecture. This curated collection showcases the progress of computer architecture research over the past quarter-century, capturing the essence of this evolving field.
“I am quite honored to be selected as part of the 98 papers for this year’s ISCA-50 retrospective,” says Jun. “The list of papers from the last time they did this in 1998 included many seminal works that shaped our field, so I am very excited and grateful to be included in this version.”
— Shani Murray