We like our Best Paper Awards around here, but UW CSE Ph.D. student Dylan Hutchison took it to a new level this week by contributing not one, but two Best Papers at IEEE’s High Performance Extreme Computing Conference (HPEC 2016). Hutchison, a member of UW CSE’s Database group, collected the Best Student Paper Award as lead author of “From NoSQL Accumulo to New SQL Graphulo: Design and Utility of Graph Algorithms inside a BigTable Database.” He also co-authored the Best Paper winner, “Julia Implementation of the Dynamic Distributed Dimensional Data Model.”
In the first paper, Hutchison and his colleagues illustrate how Graphulo—a library for executing graph algorithms inside Apache Accumulo—enables the execution of GraphBLAS kernels in a BigTable database. The team, which includes UW iSchool professor and CSE adjunct Bill Howe, and MIT researchers Jeremy Kepner and Vijay Gadepally, compared the performance of two graph algorithms implemented with Graphulo to that of two main-memory matrix math systems. Their work yields new insights into whether it is faster to execute a graph algorithm inside a database versus an external system, showing that memory requirements and relative I/O are critical factors.
The second paper, which was co-authored by a group of MIT researchers that includes student and lead author Alex Chen, professor Alan Edelman, Kepner and Gadepally, details implementation of D4M in Julia, a new language for writing data analysis programs that are easy to implement and run at high performance. The team illustrated that D4M.jl matches or outperforms its Matlab counterpart, thanks to Julia’s well-structured syntax and data structure.
Read the Graphulo paper here and view the conference presentation here. Read the Julia paper here.
Way to go, Dylan!