This repository contains a basic experimental setup to automatically configure the Cassandra NoSQL database using irace. It uses the YCSB benchmark.
(paper to be submitted)
A full Google Cloud virtual machine to reproduce the experiments can be downloaded here (6.7GB)
- Irace https://iridia.ulb.ac.be/irace/
- Cassandra https://cassandra.apache.org/
- YCSB https://github.com/brianfrankcooper/YCSB
Clone the repository.
$ git clone https://github.com/MoisesSilvaM/Cassandra-Tuning
$ cd Cassandra-Tuning
Create a directory for storing the tuning scenario setup.
$ mkdir ~/tuning
Copy all the files to the scenario directory:
cp 23_params/* ~/tuning
if you want to configure 23 parameters, or
cp 5_params/* ~/tuning
if you want to configure only the 5 most important parameters.
Copy the workload files you're interested in to the YCSB workload folder.
Create a directory called "arena"
$ mkdir ~/arena
Change the experiment budget in the scenario file.
maxExperiments = budget
If you want to include the default configuration, uncomment the configurationsFile option in scenario file.
configurationsFile = "./configurations.txt"
Select the workloads to adjust by commenting/uncommenting the relative workload in the instances file.
Check that all the paths are correct, and point to valid file locations.
Call the script:
$ sh irace.sh
or
$ nohup ./irace.sh > irace.out 2> irace.err &