Jaechan An

PhD Student · Computer Science · University of Maryland, College Park
DSLAM (Data Systems Lab) · Advisor: Daniel Abadi

I am a PhD student in Computer Science at the University of Maryland, advised by Daniel Abadi, specializing in distributed database systems. My current research focuses on building low-latency, geo-distributed multi-writer database systems that provide strong consistency.

Previously, I worked on MVCC databases for HTAP workloads and LSM-tree performance optimization at Hanyang University's DBOS Lab, advised by Hyungsoo Jung (now a professor at Seoul National University).

Email Google Scholar GitHub GitHub (Personal) LinkedIn

Research

Low-latency Multi-writer Distributed Database Systems with Strong Consistency Current

Developing a geo-distributed multi-writer disaggregated database architecture that provides serializability. The protocol is designed to be concurrency control-agnostic, making it compatible with S2PL, OCC, and MVCC-based systems.

MVCC Databases for Hybrid Transactional/Analytical Processing

Designed and integrated new mechanisms leveraging hardware-assisted (i.e., SmartSSD) and software-based approaches to support HTAP workloads on disk-oriented MVCC databases. Redesigned the GC mechanism to mitigate space amplification, reducing space usage by 93.6% in Sysbench experiments.

Read/Write Amplification in LSM-trees under HTAP

Introduced a new filter layer orthogonal to existing filters (e.g., Bloom filter), reducing read latency by 35.9% in RocksDB. The paper also presents aligned compaction to mitigate write amplification, reducing total disk writes by up to 14.9% under CH-BenCHmark.