Databend 向量化引擎性能
tip
- 仅适用于内存中的 SIMD 向量处理性能
- 数据集：100,000,000,000（千亿级）
- 硬件： AMD Ryzen 9 5950X（16 核，32 线程）
- Rust: rustc 1.61.0-nightly (8769f4ef2 2022-03-02) :::
| 查询 | DatabendQuery (v0.6.87-nightly) |
|---|---|
| SELECT avg(number) FROM numbers_mt(100000000000) | 1.682 s. (59.47 billion rows/s., 475.76 GB/s.) |
| SELECT sum(number) FROM numbers_mt(100000000000) | 1.621 s. (61.67 billion rows/s., 493.37 GB/s.) |
| SELECT min(number) FROM numbers_mt(100000000000) | 3.962 s. (25.24 billion rows/s., 201.93 GB/s.) |
| SELECT max(number) FROM numbers_mt(100000000000) | 2.792 s. (35.82 billion rows/s., 286.54 GB/s.) |
| SELECT count(number) FROM numbers_mt(100000000000) | 1.172 s. (85.31 billion rows/s., 682.46 GB/s.) |
| SELECT sum(number+number+number) FROM numbers_mt(100000000000) | 6.032 s. (16.58 billion rows/s., 132.63 GB/s.) |
| SELECT sum(number) / count(number) FROM numbers_mt(100000000000) | 1.652 s. (60.52 billion rows/s., 484.16 GB/s.) |
| SELECT sum(number) / count(number), max(number), min(number) FROM numbers_mt(100000000000) | 6.212 s. (16.10 billion rows/s., 128.78 GB/s.) |
| SELECT number FROM numbers_mt(10000000000) ORDER BY number DESC LIMIT 10 | 1.414 s. (8.76 billion rows/s., 70.09 GB/s.) |
| SELECT max(number), sum(number) FROM numbers_mt(10000000000) GROUP BY number % 3, number % 4, number % 5 LIMIT 10 | 5.791 s. (1.73 billion rows/s., 13.81 GB/s.) |
在你的笔记本电脑上即可畅享千亿级数据的查询， 现在部署，体验 Databend 的强劲动力 。