Kevin Skadron
University of Virginia, VA·Computer Science
As data volumes grow, analyzing them takes ever more time and energy. Processors are now much faster than memory, so most of that time is spent moving data rather than computing on it.
- Fields
Computer Science
Engineering
- Topics
computer architecture
processing in memory
near-data processing
- Grants
- 1 known grant, through Sep 2027
What they work on
- Develop ways to run database analytics queries directly inside DRAM main memory, and explore how intelligent DRAM can work together with other processing units.
- Design processing-in-memory (PIM) for SRAM-based caches using associative processing, which computes across many stored values at once, and test it on data analytics workloads.
- Create a domain-specific language that breaks analytics queries into dataflow graphs and optimizes their execution across SRAM PIM, DRAM PIM and the CPU.
- Build open simulation infrastructure and benchmarks for evaluating database-hardware co-designs, for use by architecture and database researchers.
More about the research
The steady speedups that once came from shrinking transistors have also slowed. One answer is to make memory 'intelligent', so it can compute results close to where the data are stored.
Why it matters
Companies and researchers depend on fast data analysis to make decisions, but costs rise as data volumes grow. Memory that computes on stored data could make database analytics faster and more energy-efficient. The shared simulators and benchmarks could help other researchers build and compare such systems.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Jose Martinez
Cornell University, NY·Computer Science
Develop mechanisms for running database analytics queries in place inside DRAM, a computer's main memory, and explore how such intelligent DRAM can cooperate with other processing units.
Josep Torrellas
University of Illinois Urbana-Champaign, IL·Computer Science
Develop approximate algorithms for graph clustering in static and dynamic settings, graph construction that preserves key information, and machine learning, for computational biology, social network and software verification problems.
Minxuan Zhou
Illinois Institute of Technology, IL·Computer Science
Design a domain-specific programming language that makes it easier to write general fully homomorphic encryption (FHE) applications across different algorithmic schemes.
The school
University of Virginia: 43 labs on PI Pages, $814M a year on research, 413 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.