Cassandra Benchmarking Project

Installing dependencies

• On each of the 5 cluster nodes, install Cassandra 3.11 by following the instruction to install Cassandra from tarball here. Add $CASSANDRA_HOME/bin to your PATH so that you can run cassandra and cqlsh from anywhere.
• You should also have Python 2.7 and pip installed. If you have no sudo access to the server, refer to this guide on how to install pip without sudo access.
• Install the python dependencies using pip install --user -r requirements.txt

Configuring Cassandra

• Specify your cluster nodes' ip addresses and login details in config.txt. Make sure they are in order.
• Run ./configure.sh

In the directory conf, you should find the cassandra configuration files for each of 5 cluster nodes. Replace the cassandra.yaml in each of your cluster nodes with these files.

Setup the schema

• Start the cassandra servers on all 5 nodes (by running cassandra)

• Go to any one of your cluster nodes, again make sure that cqlsh can be run from anywhere, and run ./setup.sh

Running benchmarks

• On each of the cluster nodes, download the project codes from here and unzip it inside the home folder

• For benchmarking, it's required that you have a server that has passwordless ssh access to the 5 cluster nodes.

• On this server that has passwordless ssh access to all the cluster node, run ./benchmark.sh <NUMBER_OF_CLIENTS> <CONSISTENCY_LEVEL> to start benchmarking. For CONSISTENCY_LEVEL, 1 denotes consistency level ONE, while 4 denotes consistency level QUORUM. Log files will be created in your working directory when the benchmark script is running, so it is recommended that you run the benchmarking inside a separate directory, and allocate enough memory to it (around 2GB).