Credits: Most scripts have been referenced from Fivetran DW Benchmark and have been adapted to suit our particular usecase.

1. Move dsdgen to a GCS bucket to a specific location as mentioned in the bootstrap script
2. Create a High CPU VM eg. 16vCPU
3. Clone this repository

git clone $REPO_URL

4. Give all script files executable permission

chmod +x *.sh

4. Run bootstrap.sh
   1. This pulls dsdgen binary
   2. Installs Google Fuse; this is to mount GCS bucket as a local folder - More info
5. Run data_gen.sh
   Usage:

   ./data_gen.sh $CPU $SCALE

   1. This is responsible for generating data
   2. $CPU denotes the amount of parallelism must be > 1
   3. $SCALE denotes the scale of data that needs to be generated
   4. This creates and mounts a GCS Bucket and writes data to it
   5. NOTE: Ensure that $CPU is close to number of CPUs in VM for efficient parallel generation
6. Run load_data.sh
   Usage:

   ./load_data.sh $SCALE

   1. This is responsible of loading data in GCP buckets created in step 5 to BigQuery
   2. $SCALE denotes the scale of data that needs to be loaded to BigQuery
   3. Note: Before running this step ensure that data is generated and present in the appropriate GCS Bucket
7. Run benchmark.sh
   Usage:

   ./benchmark.sh $SCALE

   1. This is responsible for running TPC-DS queries and measuring query execution time
   2. Generates a csv file in results folder containing the query start_time and end_time
   3. Saves query statistics in the same directory