Xian-He Sun
Illinois Institute of Technology, IL·Computer Science
Computers have become very fast at calculating, but moving data in and out of memory and storage is now the slowest step in large-scale science and machine learning. Simulations, telescopes, gene sequencers and AI training jobs all wait on data.
- Fields
Computer Science
Engineering
- Topics
high-performance computing
parallel I/O
storage systems
- Grants
- 5 known grants, through Jul 2029
What they work on
- Engineer IOWarp, a modular and scalable data management platform for AI-infused scientific workflows, using NVMe SSDs, CXL memory devices and CPU-GPU co-design to reduce data access times on HPC systems.
- Build a cross-layer Unified Memory-centric Computing (UniMCC) system for machines with near-memory data processors and disaggregated shared memory, covering architecture, code generation, runtime support and performance modeling.
- Develop LABIOS, a distributed storage system that runs Labels (an operation plus a pointer to its data) asynchronously across tiered NVMe, SSD and hard-disk servers to provide active and software-defined storage semantics.
- Measure how LABIOS changes I/O efficiency for scientific computing, deep learning and microservice workloads, and assess what a label-based I/O paradigm means for existing HPC infrastructure.
- Create ChronoLog, a log store that uses physical time for synchronization-free ordering of activity data, auto-tiering logs across storage-class memory and NVMe flash under multiple-writer, multiple-reader access.
- Make memory access pattern obfuscation practical on current and emerging memory architectures by co-designing cryptographic protocols, protected applications and architectural support against side-channel attackers.
More about the research
The projects here redesign the layers that hold and move data: storage platforms for scientific workflows, systems for recording streams of events, hardware-software designs that compute next to memory, and techniques for keeping memory activity private from attackers.
Why it matters
Faster and more flexible data handling could cut the waiting time between measurement and insight for researchers working with very large datasets in genomics, climate science, cosmology and AI. The logging work is aimed at applications the grant names, including fraud detection in financial transactions, weather prediction, medical and bioengineering data analysis, and more secure web services. Practical obfuscation methods could help keep sensitive computations private on future memory hardware.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Collaborative Research: Frameworks: IOWarp: Bending the I/O Fabric for Advancing AI-Infused Scientific Workflows
NSF · Aug 2024 – Jul 2029
Collaborative Research: CSR: Medium: Towards A Unified Memory-centric Computing System with Cross-layer Support
NSF · Oct 2023 – Sep 2027
OAC Core: LABIOS: Storage Acceleration via Data Labeling and Asynchronous I/O
NSF · Aug 2023 – Jul 2027
Closest labs
Peng Jiang
University of Iowa, IA·Computer Science·Early-career
Develop a runtime system that gives a virtualized view of data partitioned across heterogeneous devices, plus a compiler tool that places data using a statistical performance model of random memory accesses.
Feng Chen
Indiana University, IN·Computer Science
Develop hardware functionality and system-level memory management for in-memory computing over Compute Express Link (CXL), the interconnect standard that lets processors and memory devices share memory, to reduce fragmentation of memory resources.
Mahmut Kandemir
Pennsylvania State University, PA·Computer Science
Characterize diverse HPC workloads on real and simulated computational storage devices to identify which computations benefit from running inside storage, and what performance tradeoffs result.
The school
Illinois Institute of Technology: 5 labs on PI Pages, $38M a year on research, 76 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.