I am Yibo Wang, a junior majoring in Computer Science at College of Computer Science, Sichuan University (Chengdu, China). Currently, I am in AI-System Lab, under the supervision of Prof. Mingjie Tang.
My research interests lie in the intersection of database management systems (DBMSs) and machine learning (ML), especially using ML/AI techniques to automate database administration/tuning to remove human impediments.
My aim is to make significant contributions to the field of self-driving DBMS through advanced research and collaborative efforts.
I am a highly self-motivated undergraduate eagerly seeking admission to a Ph.D. program.
Email: wangyibo2@stu.scu.edu.cn
🔥 News
- 2024.03: 🎉🎉 Demo of GPTuner is accepted by SIGMOD, 2024 !
- 2024.03: 🎉🎉 GPTuner is accepted by VLDB, 2024 !
- 2024.01: 🎉🎉 A video demonstration of GPTuner is available on YouTube!
- 2023.12: 🎉🎉 GPTuner is under revision of Proceedings of Very Large Data Bases Conference (VLDB) !
- 2023.07: 🎉🎉 I become one of AI-System lab at College of Computer Science, Sichuan University!
📖 Educations
- 2021.09 - 2025.06, Bachelor of Computer Science, College of Computer Science, Sichuan University, Chengdu, China
📝 Publications
GPTuner: A Manual-Reading Database Tuning System via GPT-Guided Bayesian
Jiale Lao, Yibo Wang, Yufei Li, Jianping Wang, Yunjia Zhang, Zhiyuan Chen, Wanghu Chen, Mingjie Tang, Jianguo Wang
Proceedings of Very Large Data Bases Conference (VLDB), 2024.
A Demonstration of GPTuner: A GPT-Based Manual-Reading Database Tuning System
Jiale Lao, Yibo Wang, Yufei Li, Jianping Wang, Yunjia Zhang, Zhiyuan Chen, Wanghu Chen, Yuanchun Zhou, Mingjie Tang, Jianguo Wang
Proceedings of ACM Conference on Management of Data (SIGMOD), 2024.
📝 Research
- Engaged users to probe into the ingenious LLM-powered pipeline which refines and unifies heterogeneous knowledge to guide system optimization.
- Unleashed the potential of everyday users, enabling them to delve into the nuances of knob features and maximize the efficiency of their tailored DBMS seamlessly.
- Empowered DBAs to supercharge GPTuner with their priceless tuning expertise expressed in natural language and witness how it can be customized to the Coarse-to-Fine Optimization Framework.
- Outcomes: a demo paper accepted by SIGMOD 2024, and an open-source project with more than 3000 views, 200 clones and 50 stars on GitHub.
- Designed and implemented GPTuner, a novel manual-reading database tuning system that automatically exploits domain knowledge to enhance the knob tuning process.
- Developed a LLM-based data pipeline, a prompt ensemble algorithm, a workload-aware and training-free knob selection strategy, and a Coarse-to-Fine Bayesian Optimization Framework.
- Evaluated GPTuner under different benchmarks, metrics and DBMS. It identifies better configurations 16x faster and achieves 30% performance improvement over the best-performing alternative.
- Outcomes: a research paper accepted by VLDB 2024.
- Collaborated with AntGroup to develop an automated optimization system for Flink, reducing resource consumption to cope with tight budget while maintaining SLA adherence.
- Proposed a rule-based method to get pod features based on the degree of parallelism of vertexes.
- Implemented an ML-based evaluator to estimate resource utilization of a pod given its features.