Synthetic Load Generation for Dynamic Cloud Configuration Management

1. Install PostgreSQL 14 or later along with psql
   https://www.postgresql.org/

2. Clone the repository from Github or Zenodo
   https://github.com/TheMonocledHamster/SynConfLoad/
   https://zenodo.org/badge/latestdoi/568378464

3. Download the Azure Traces dataset from https://github.com/Azure/AzurePublicDataset/blob/ef8b2517b27357df0b418b6e6ca4efcdeb5117b0/AzureFunctionsDataset2019.md

4. From the dataset, copy 2 files, function_durations_percentiles.anon.d01.csv and invocations_per_function_md.anon.d01 into the source/traces/AzureFunctions/ directory. Your directory structure should look something like this:
   

5. Ensure that the PostgreSQL Database is properly installed and in operation.

6. Run sudo -u postgres bash source/db/run.sh from the project root directory.

7. The script may take anywhere from 1-3 hours to execute, with a max disk space requirement of around 10 GB.

8. Upon completion, do the following:

9. Install psycopg2 and numpy using
   pip install psycopg2-binary
   pip install numpy

10. Navigate to source/arrival_rates/gen_arrivals.py

11. Change the user to your username

12. Run python gen_arrivals.py

The final data should be populated into sub-folders within the arrival_rates directory, as .csv and .npy files for each SLO bin.