Hey, if you’re into data engineering and want to see how different databases handle large language model (LLM) style queries, I’ve got something for you. Someone just shared a database benchmark and ‘chat latency simulator’ app on Reddit that’s pretty interesting. It’s designed to test LLM style queries on Postgres and Clickhouse, with datasets ranging from 10,000 to 10 million rows.
The app lets you run the benchmark yourself, adjusting container resources and data size to see how it affects performance. Plus, you can simulate chat latency to get a sense of the user experience. The creator is looking for feedback, so if you’re familiar with benchmarking projects, they’d love to hear your thoughts.
I think this is a great resource for anyone working with large language models or trying to optimize their database performance. The project is open-source, so you can dig into the code and learn from it.
Check it out and let me know what you think!