PhoebeDB: A Disk-Based RDBMS Kernel for High-Performance and Cost-Effective OLTP
Boge Liu, Chunling Wang, Xiaoshuang Chen, Yu Hao, Zhengyi Yang*, Yi Jin, Yixing Yang, Wenke Yang, Wanchuan Zhang, Wenjie Zhang
International Conference on Extending Database Technology (EDBT)
RAIDS Lab Authors
Details
Research Area
Tags
Resources
Abstract
Relational databases have long been fundamental to data management. This paper presents PhoebeDB, an enterprise- and commercial-oriented RDBMS kernel that integrates recent research with practical innovations to deliver high-performance, cost-efficient OLTP solutions. It features: 1) an in-memory data-centric storage design optimized for parallel access and data temperature-based buffer management, 2) a co-routine pool-based runtime with a smart scheduler that maximizes CPU utilization for high-concurrency workloads, and 3) optimized transaction management with in-memory UNDO logs, hybrid concurrency control, parallel Write-Ahead Logging with Remote Flush Avoidance, and enhanced snapshot isolation. Experiments show that PhoebeDB achieves nearly 13.7 million tpmC and 30 million tpm on the TPC-C benchmark using a single machine, delivering a 27x improvement over PostgreSQL.
Author Affiliations
BibTeX
@misc{liu2025phoebedb,
title = {PhoebeDB: A Disk-Based RDBMS Kernel for High-Performance and Cost-Effective OLTP},
author = {Liu, Boge and Wang, Chunling and Chen, Xiaoshuang and Hao, Yu and Yang, Zhengyi and Jin, Yi and Yang, Yixing and Yang, Wenke and Zhang, Wanchuan and Zhang, Wenjie},
doi = {10.48786/EDBT.2025.82},
url = {https://openproceedings.org/2025/conf/edbt/paper-292.pdf},
keywords = {Database Technology},
language = {en},
publisher = {OpenProceedings.org},
year = {2025}
}