Intelligent routing of queries in a polystore system (Master Project, Finished)
Query routing is the process of selecting on which data store(s) a query or parts of it should be executed. Polypheny-DB currently has two query router implementations. While both implementations work fine for their intended use-case, there is not yet a universal router implementation. Furthermore, there is still a huge potential for further optimizations of the query routing.
The goal of this project is to develop a new approach to query routing in Polypheny-DB. The project consists of the following parts:
- Classification: An important part of query routing is the classification of the work- load. It would be interesting to apply machine learning based techniques.
- Estimation: In order to route a query, there needs to be a component which es- timates the execution time of a (partial) query plan on an underlying data store. This estimator should use machine learning techniques.
- Execution: The actual core of the implementation which utilizes the previously implemented estimator to find the optimal combination of data stores.
The implementation can optionally be extended to also adjust the system according to the workload. This could include adding and removing data stores and adjusting the data placement.
Start / End Dates
2021/01/01 - 2021/08/15