QTune A Query-Aware Database Tuning System with Deep Reinforcement Learning

Existing database tunning

1. reply on a large number of high-quality data DBAs’ experience to train
2. run a SQL query workload multiple times in the database to get an appropriate configuration.
3. cannot provide fine-grained tuning (i.e., tuning for a specific query workload.
4. existing DRL model, which assumes that the environment can only be affected by reconfiguring actions, but cannot utilize the query information,

actor-critic networks is an effective strategy to solve optimal problems with continuous action space by concurrently learning the Q-value function and the action policy.

---

• ← Previous Post
• Next Post →