Emma on CoGaDB

CoGaDB is a column-oriented GPU-accelerated database management system.

The aim of this project is to execute emma dataflows on CoGaDB.

The following guide for setting up CoGaDB is taken from the CoGaDB repository, which also contains the latest project code. Use of the latest code is recommended (falcon branch).

Recommended system: Ubuntu 16.04 or higher (14 should work as well).

The following steps should guide you through the installation and basic configuration for running the emma-cogadb tests.

Clone the Repository

Get access to the repository and clone as follows.

# password-based authentication
hg clone https://USERNAME@bitbucket.org/cogadb/falcon
# ssh-based authentication
hg clone ssh://hg@bitbucket.org/cogadb/falcon

Update the Falcon Branch

cd falcon
hg pull && hg update falcon

Install Build Dependencies

Install system specific dependencies using the provided script.

bash utility_scripts/install_cogadb_dependencies.sh

Compile the Project

mkdir debug_build
cd debug_build
cmake -DCMAKE_BUILD_TYPE=Debug ..
make -j$(grep -c ^processor /proc/cpuinfo)

Add Test Data (Optional)

Add some tpch and ssb data as follows.

bash utility_scripts/setup_reference_databases.sh

Launch

Browse to the bin directory and launch the service as follows.

cd bin
./cogadbd

If you plan using CoGaDB in standalone mode, consider using startup.coga as a configuration file. It should be located in the bin directory of your build and can be useful for executing generated (or hardcoded) dataflows directly at startup. A sample configuration file can be found here

In order to run the emma-cogadb tests, define a COGADB_HOME variable in your ~/.profile (you have to restart your Linux session before the changes can take effect).

export COGADB_HOME=/path/to/your/build/directory