A lightweight and high-performance order-book designed to process level 2 and trades data.
Ran a couple of benchmarks showcasing real case scenarios against a naive Vec implementation and orderbook with fixed sizes of 100 and 500 levels.
The benchmarks are run with 300,000 events of level 2 orderbook data . This data is split in 2 chunks:
- First 200,000 events are used to warm up.
- Last 100,000 for the actual benchmark.
The scenarios tested are:
- Process events and stream best bid and ask
- Process events and stream top5 bids and asks
Here are the results:
| bench | iterations | time | ns/iter |
|---|---|---|---|
| ninjabook_bbo | 100,000 | 9.1093 ms | 91.093 ns |
| fixed_100_bbo | 100,000 | 14.253 ms | 142.53 ns |
| fixed_500_bbo | 100,000 | 72.636 ms | 726.36 ns |
| naive_bbo | 100,000 | 263.77 ms | 2,637.7 ns |
| ninjabook_top5 | 100,000 | 16.729 ms | 167.29 ns |
| fixed_100_top5 | 100,000 | 23.532 ms | 235.32 ns |
| fixed_100_top5 | 100,000 | 81.112 ms | 811.12 ns |
| naive_top5 | 100,000 | 275.40 ms | 2,754 ns |
To add a better version, create a new file, implementing the same methods as orderbook.rs (including tests) and add the improved orderbook to the bench optimal_vs_naive.rs. Only order books with a better performance than orderbook.rs will be considered. Lastly, add performance logs to the Pull Request, can just copy paste what cargo bench outputs.
Any issues, refactoring, docs and tests are also welcomed. Feel free to reach out here if you have any questions.