[Abstract]

The vast amounts of digital data generated daily are managed and processed by large-scale systems in data centers. The capital and operational costs associated with these systems are significant, driving a growing demand for solutions that are cost-effective, energy-efficient, and high-performance. I have focused on reshaping entire computing system architectures to meet these demands. In this talk, I will share my journey and insights in developing such systems. I begin with LightStore (ASPLOS’19), a network-attached, embedded-class data store designed as a building block for constructing scalable, energy-efficient data pools. Next, I present PinK (USENIX ATC’20), a key-value (KV) store for LightStore that achieves fast response times and high bandwidth with stable latency through hardware-software co-design. I then discuss KEVIN (OSDI’21), a POSIX-compatible, in-kernel file system that operates over LightStore and PinK. By offloading data indexing to the storage side, KEVIN significantly enhances performance, especially for metadata-intensive applications. Finally, I introduce BigKV (EuroSys’23), a high-performance KV caching system optimized for managing large objects over all-flash arrays.

[Biography]

Sungjin Lee received a BE degree in electrical engineering from Korea University in 2005 and MS and PhD degrees in computer science and engineering from Seoul National University in 2007 and 2013, respectively. Before joining POSTECH in 2025, he was an assistant/associate professor at DGIST (2017-2025) and Inha University (2016–2017), respectively. He was a postdoctoral associate at the Computer Science and Artificial Intelligence Laboratory (CSAIL), Massachusetts Institute of Technology (MIT) (2013–2016). His research interests include operating systems, system software, storage systems, and AI systems.