Bao
A learned database query optimization system proposed by MIT in 2021. It supplies hints to a traditional query optimizer and uses a bandit algorithm to learn optimal hint combinations from execution feedback, while retaining the fallback capability of the underlying traditional optimizer.
Timeline (last 90 days)
Bao (Bandit Optimizer, MIT 2021) provides different hints to the traditional optimizer, allowing it to search within a constrained space, and then uses a bandit algorithm to learn from execution feedback, preserving the fallback capability of the traditional optimizer.
Academic exploration in the direction of learned optimizers includes the DQ method, which uses deep reinforcement learning for join order selection, and the Bao system, which uses graph neural networks to model query structure.
All Facts (2)
Bao (Bandit Optimizer, MIT 2021) provides different hints to the traditional optimizer, allowing it to search within a constrained space, and then uses a bandit algorithm to learn from execution feedback, preserving the fallback capability of the traditional optimizer.
50%UnverifiedAcademic exploration in the direction of learned optimizers includes the DQ method, which uses deep reinforcement learning for join order selection, and the Bao system, which uses graph neural networks to model query structure.
50%