A new version of swingbench with a version of the TPC-C benchmark

There’s a new version of swingbench that includes the TPC-C benchmark and a tool to extract benchmark results from the database and export them as text or csv files.

My early career was periodically marked with the hyperbole that came with continuously increasing improvements in the TPC-C workloads before it became too expensive to run the benchmarks and, I would argue no longer reflected reality. Even though TPC-C is rarely used for official benchmarks and hardly anyone uses it as a yard stick on the performance of databases, I keep getting asked when I would add support for TPC-C, well that day has come.

I’ve implemented both the data generator and workload for TPC-C using the tools that most people who use swingbench will immediately feel immediately familiar with. The unit of size, and concurrency, is the number of rows in the Warehousetable. This drives the other tables rows counts i.e. 1 rows in the Warehouse table results in 10 rows in the District table. Based on my calculations about 15 rows in the Warehouse table results in 1GB of data, equivalent to a scale of 1.

The image below shows a dataset of scale 1, roughly 1GB in size.

It takes about 1-2 mins per scale unit. i.e. Scale 1 takes about 1 minute to generate. Scale 10 takes about 10 minutes.

I haven’t spent anytime optimizing the code, so there may be a few changes in the coming weeks as people point out changes that might improve the performance. If you see anything don’t hesitate to reach out to me.

I’ve also add a new results2text utility in the bin/winbin directory which will extract results files from the database and write them to stdout and/or csv files. I’ve written another blog about that here.

You can download the latest version of the code in the usual location here

• 2025 4
• 2024 1
• 2023 5
• 2022 8