Houxiang Ji

headshot-2.jpg

213 Coordinated Science Laboratory

1308 W. Main Street

Urbana, IL, 61801

I am a Ph.D. candidate in the Siebel School of Computing and Data Science at University of Illinois Urbana-Champagin, advised by Prof. Nam Sung Kim. I am affiliated with the FAST research group and Coordinated Science Lab. My research interests include computer architecture, memory systems, computer express link (CXL), and system for ML. I am currently working on CXL Type-2 device to boost hyperscaler application performance.

Before joining UIUC, I received my bachelor’s degree in Computer Science and Engineering and Finance from Shanghai Jiao Tong University (SJTU). I have worked as a visiting researcher at Meta and as a research intern at Luminous Computing and Intel Lab.

news

Jul, 2025 Our paper Re-architecting End-host Networking with CXL: Coherence, Memory, and Offloading has been accepted to MICRO 2025! This work pioneers the use of a CXL Type-2 device to build end-host networking, leveraging its cache coherence, memory expansion, and offloading capabilities.
Jun, 2025 Our work on SSD Latency Prediction has been accepted to IEEE Computer Architecture Letter! In this paper, we develop time series machine learning models to precisely predict SSD access latency, enabling smarter storage system design and optimization.
Apr, 2025 Our work on Memory Deduplication Acceleration has been accepted to USENIX ATC 2025! In this work, we redesign the memory deduplication process to be more efficient by enabling parallel function execution, which are further accelerated by the on-chip accelerator, DSA. This is extensive work from our previous CAL work on DSA.
Jan, 2025 Our work on Data Streaming Accelerator has been accepted to IEEE CAL! In this work, we propose a cooperative memory deduplication scheme between CPU and on-chip accelerator, DSA, to reduce the memory tax.
Jul, 2024 Our work on CXL Type-2 device has been accepted to MICRO 2024! This is the first work on a geniue CXL Type-2 device, showcasing its characterization and potential for heterogeneous computing.