Reinforcement Learning with Tree-LSTM for Join Order Selection

Reinforcement Learning with Tree-LSTM for Join Order Selection(RTOS) is an optimizer which focous on Join order Selection(JOS) problem. RTOS learn from previous queries to build plan for further queries with the help of DQN and TreeLSTM.

Development

This repository contains a very early version of RTOS.It is written in python and built on PostgreSQL. It now contains the training and testing part of our system.

It has been fully tested on join order benchmark(JOB). You can train RTOS and see its' performance on JOB.

This code is still under development and not stable, add issues if you have any questions. We will polish our code and provide further tools.

Important parameters

Here we have listed the most important parameters you need to configure to run RTOS on a new database.

• schemafile
  • a sample
• sytheticDir
  • Directory contains the sytheic workload(queries), more sythetic queries will help RTOS make better plans.
  • It is nesscery when you want apply RTOS for a new database.
• JOBDir
  • Directory contains all JOB queries.
• Configure of PostgreSQL
  • dbName : the name of database
  • userName : the user name of database
  • password : your password of userName
  • ip : ip address of PostgreSQL
  • port : port of PostgreSQL

Requirements

• Pytorch 1.0
• Python 3.7
• torchfold
• psqlparse

Run the JOB

1. Follow the https://github.com/gregrahn/join-order-benchmark to configure the database

2. Add JOB queries in JOBDir.

3. Add sythetic queries in sytheticDir, more sythetic queries will help RTOS generate better results. E.g. You can generate queries by using templates.

4. Cost Training, It will store the model which optimize on cost in PostgreSQL

   python3 CostTraining.py

5. Latency Tuning, It will load the cost training model and generate the latency training model

   python3 LatencyTuning.py 

6. Enjoy your optimizer by running

   python3 run.py

   Sample:

   $python3 run.py
   Enter each query in one line
   ---------------------
   >SELECT MIN(mc.note) AS production_note, MIN(t.title) AS movie_title, MIN(t.production_year) AS movie_year FROM company_type AS ct, info_type AS it, movie_companies AS mc, movie_info_idx AS mi_idx, title AS t WHERE ct.kind = 'production companies' AND it.info = 'top 250 rank' AND mc.note  not like '%(as Metro-Goldwyn-Mayer Pictures)%' and (mc.note like '%(co-production)%' or mc.note like '%(presents)%') AND ct.id = mc.company_type_id AND t.id = mc.movie_id AND t.id = mi_idx.movie_id AND mc.movie_id = mi_idx.movie_id AND it.id = mi_idx.info_type_id;
   --------------
   ('(as DreamWorks Pictures) (presents)', '12 Years a Slave', 1934)
   -------------
   >

Questions

Contact Xiang Yu (x-yu17@mails.tsinghua.edu.cn) if you have any questions.