Closed momo54 closed 1 year ago
@Yotlan Could you implement this but comment out the modifications for now?
Another issue is what to do when measurements don't agree?
instance 1: [timeout, 482, 496, timeout] instance 2: [537, 493, 496, 519]
- A outlier removal strategy is needed:
- [ ] Inter-Quartile Range proximity rule
- [ ] Z-score
We need to see the effect of increasing the number of sources on the source selection time per federated Query engine !
In Kobe, they also include the planning time (time to optimize the query after source-selection ie. Join ordering...)