##Docker image:

The 'launcher' docker image contains all the necessary client-side software to run queries on the following systems:

• Spark
• Impala
• Oracle
• Vertica

See the Dockerfile for details.

We assume that each system is already up and running with access to the required datasets.

The configuration file 'env.sh' specifies connection parameters for each of the systems.

##Starting a container:

To start a container for launching experiments run:

docker run -t -i --net=host damsl2:5020/damsl/launcher bash

Upon startup, the container will pull the K3-Benchstack repo, load the configuration from env.sh, and start up a webserver in the launcher/web folder for displaying plots.

##Running an experiment on a single system:

Each system provides a shell script for executing a single experiment on a single dataset, then reporting the elapsed time.

The scripts are located at launcher/src/systems//run_.sh

The scripts generally take a query file and a dataset name as arguments, but see usage for exact details. All necessary files will be located within the system's diectory.

##Automated experiments: See the src/ folder.

The python program launch.py is the main driver.

Run:

python launch.py --help

for exact usage.

The arguments are:

• --systems: A list of systems to run the exeriments on.
• --trials: The number of trials for each experiment.
• --tpch10g: A list of queries to run on the TPCH10g dataset.
• --tpch100g: A list of queries to run on the TPCH100g dataset.

For example:

python launch.py --systems Spark Impala --trials 2 --tpch100g 1 3 5 6

Results are stored in a postgres database running outside of the container. More details to come.