clickhouse 在数据分析技术领域早已声名远扬,如果还不知道可以 点这里 了解下。
最近由于项目需求使用到了 clickhouse 做分析数据库,于是用测试环境做了一个单表 6 亿数据量的性能测试,记录一下测试结果,有做超大数据量分析技术选型需求的朋友可以参考下。
服务器信息
- CPU:Intel Xeon Gold 6240 @ 8x 2.594GHz
- 内存:32G
- 系统:CentOS 7.6
- Linux内核版本:3.10.0
- 磁盘类型:机械硬盘
- 文件系统:ext4
Clickhouse信息
- 部署方式:单机部署
- 版本:20.8.11.17
测试情况
测试数据和测试方法来自 clickshouse 官方的 Star Schema Benchmark
按照官方指导造出了测试数据之后,先看一下数据量和空间占用情况。
数据量和空间占用
表名 | 列数 | 数据行数 | 原始大小 | 压缩大小 | 压缩率 |
supplier | 6 | 200,000 | 11.07 MiB | 7.53 MiB | 68 |
customer | 7 | 3,000,000 | 168.83 MiB | 114.72 MiB | 68 |
part | 8 | 1,400,000 | 34.29 MiB | 24.08 MiB | 70 |
lineorder | 16 | 600,037,902 | 24.03 GiB | 16.67 GiB | 69 |
lineorder_flat | 37 | 688,552,212 | 111.38 GiB | 61.05 GiB | 55 |
可以看到 clickhouse 的压缩率很高,压缩率都在 50 以上,基本可以达到 70 左右。数据体积的减小可以非常有效的减少磁盘空间占用、提高 I/O 性能,这对整体查询性能的提升非常有效。
supplier、customer、part、lineorder 为一个简单的「供应商-客户-订单-地区」的星型模型,lineorder_flat 为根据这个星型模型数据关系合并的大宽表,所有分析都直接在这张大宽表中执行,减少不必要的表关联,符合我们实际工作中的分析建表逻辑。
以下性能测试的所有分析 SQL 都在这张大宽表中运行,未进行表关联查询。
查询性能测试详情
Query 1.1
SELECT sum(LO_EXTENDEDPRICE * LO_DISCOUNT) AS revenue FROM lineorder_flat WHERE (toYear(LO_ORDERDATE) = 1993) AND ((LO_DISCOUNT >= 1) AND (LO_DISCOUNT <= 3)) AND (LO_QUANTITY < 25) ┌────────revenue─┐ │ 44652567249651 │ └────────────────┘ 1 rows in set. Elapsed: 0.242 sec. Processed 91.01 million rows, 728.06 MB (375.91 million rows/s., 3.01 GB/s.)
扫描行数:91,010,000大约9100万
耗时(秒):0.242
查询列数:2
结果行数:1
Query 1.2
SELECT sum(LO_EXTENDEDPRICE * LO_DISCOUNT) AS revenue FROM lineorder_flat WHERE (toYYYYMM(LO_ORDERDATE) = 199401) AND ((LO_DISCOUNT >= 4) AND (LO_DISCOUNT <= 6)) AND ((LO_QUANTITY >= 26) AND (LO_QUANTITY <= 35)) ┌───────revenue─┐ │ 9624332170119 │ └───────────────┘ 1 rows in set. Elapsed: 0.040 sec. Processed 7.75 million rows, 61.96 MB (191.44 million rows/s., 1.53 GB/s.)
扫描行数:7,750,000775万
耗时(秒):0.040
查询列数:2
返回行数:1
Query 2.1
SELECT sum(LO_REVENUE), toYear(LO_ORDERDATE) AS year, P_BRAND FROM lineorder_flat WHERE (P_CATEGORY = 'MFGR#12') AND (S_REGION = 'AMERICA') GROUP BY year, P_BRAND ORDER BY year ASC, P_BRAND ASC ┌─sum(LO_REVENUE)─┬─year─┬─P_BRAND───┐ │ 64420005618 │ 1992 │ MFGR#121 │ │ 63389346096 │ 1992 │ MFGR#1210 │ │ ........... │ .... │ ..........│ │ 39679892915 │ 1998 │ MFGR#128 │ │ 35300513083 │ 1998 │ MFGR#129 │ └─────────────────┴──────┴───────────┘ 280 rows in set. Elapsed: 8.558 sec. Processed 600.04 million rows, 6.20 GB (70.11 million rows/s., 725.04 MB/s.)
扫描行数:600,040,000大约6亿
耗时(秒):8.558
查询列数:3
结果行数:280
Query 2.2
SELECT sum(LO_REVENUE), toYear(LO_ORDERDATE) AS year, P_BRAND FROM lineorder_flat WHERE ((P_BRAND >= 'MFGR#2221') AND (P_BRAND <= 'MFGR#2228')) AND (S_REGION = 'ASIA') GROUP BY year, P_BRAND ORDER BY year ASC, P_BRAND ASC ┌─sum(LO_REVENUE)─┬─year─┬─P_BRAND───┐ │ 66450349438 │ 1992 │ MFGR#2221 │ │ 65423264312 │ 1992 │ MFGR#2222 │ │ ........... │ .... │ ......... │ │ 39907545239 │ 1998 │ MFGR#2227 │ │ 40654201840 │ 1998 │ MFGR#2228 │ └─────────────────┴──────┴───────────┘ 56 rows in set. Elapsed: 1.242 sec. Processed 600.04 million rows, 5.60 GB (482.97 million rows/s., 4.51 GB/s.)
扫描行数:600,040,000大约6亿
耗时(秒):1.242
查询列数:3
结果行数:56
Query 3.1
SELECT C_NATION, S_NATION, toYear(LO_ORDERDATE) AS year, sum(LO_REVENUE) AS revenue FROM lineorder_flat WHERE (C_REGION = 'ASIA') AND (S_REGION = 'ASIA') AND (year >= 1992) AND (year <= 1997) GROUP BY C_NATION, S_NATION, year ORDER BY year ASC, revenue DESC ┌─C_NATION──┬─S_NATION──┬─year─┬──────revenue─┐ │ INDIA │ INDIA │ 1992 │ 537778456208 │ │ INDONESIA │ INDIA │ 1992 │ 536684093041 │ │ ..... │ ....... │ .... │ ............ │ │ CHINA │ CHINA │ 1997 │ 525562838002 │ │ JAPAN │ VIETNAM │ 1997 │ 525495763677 │ └───────────┴───────────┴──────┴──────────────┘ 150 rows in set. Elapsed: 3.533 sec. Processed 546.67 million rows, 5.48 GB (154.72 million rows/s., 1.55 GB/s.)
扫描行数:546,670,000大约5亿4千多万
耗时(秒):3.533
查询列数:4
结果行数:150
Query 3.2
SELECT C_CITY, S_CITY, toYear(LO_ORDERDATE) AS year, sum(LO_REVENUE) AS revenue FROM lineorder_flat WHERE (C_NATION = 'UNITED STATES') AND (S_NATION = 'UNITED STATES') AND (year >= 1992) AND (year <= 1997) GROUP BY C_CITY, S_CITY, year ORDER BY year ASC, revenue DESC ┌─C_CITY─────┬─S_CITY─────┬─year─┬────revenue─┐ │ UNITED ST6 │ UNITED ST6 │ 1992 │ 5694246807 │ │ UNITED ST0 │ UNITED ST0 │ 1992 │ 5676049026 │ │ .......... │ .......... │ .... │ .......... │ │ UNITED ST9 │ UNITED ST9 │ 1997 │ 4836163349 │ │ UNITED ST9 │ UNITED ST5 │ 1997 │ 4769919410 │ └────────────┴────────────┴──────┴────────────┘ 600 rows in set. Elapsed: 1.000 sec. Processed 546.67 million rows, 5.56 GB (546.59 million rows/s., 5.56 GB/s.)
扫描行数:546,670,000大约5亿4千多万
耗时(秒):1.00
查询列数:4
结果行数:600
Query 4.1
SELECT toYear(LO_ORDERDATE) AS year, C_NATION, sum(LO_REVENUE - LO_SUPPLYCOST) AS profit FROM lineorder_flat WHERE (C_REGION = 'AMERICA') AND (S_REGION = 'AMERICA') AND ((P_MFGR = 'MFGR#1') OR (P_MFGR = 'MFGR#2')) GROUP BY year, C_NATION ORDER BY year ASC, C_NATION ASC ┌─year─┬─C_NATION──────┬────────profit─┐ │ 1992 │ ARGENTINA │ 1041983042066 │ │ 1992 │ BRAZIL │ 1031193572794 │ │ .... │ ...... │ ............ │ │ 1998 │ PERU │ 603980044827 │ │ 1998 │ UNITED STATES │ 605069471323 │ └──────┴───────────────┴───────────────┘ 35 rows in set. Elapsed: 5.066 sec. Processed 600.04 million rows, 8.41 GB (118.43 million rows/s., 1.66 GB/s.)
扫描行数:600,040,000大约6亿
耗时(秒):5.066
查询列数:4
结果行数:35
Query 4.2
SELECT toYear(LO_ORDERDATE) AS year, S_NATION, P_CATEGORY, sum(LO_REVENUE - LO_SUPPLYCOST) AS profit FROM lineorder_flat WHERE (C_REGION = 'AMERICA') AND (S_REGION = 'AMERICA') AND ((year = 1997) OR (year = 1998)) AND ((P_MFGR = 'MFGR#1') OR (P_MFGR = 'MFGR#2')) GROUP BY year, S_NATION, P_CATEGORY ORDER BY year ASC, S_NATION ASC, P_CATEGORY ASC ┌─year─┬─S_NATION──────┬─P_CATEGORY─┬───────profit─┐ │ 1997 │ ARGENTINA │ MFGR#11 │ 102369950215 │ │ 1997 │ ARGENTINA │ MFGR#12 │ 103052774082 │ │ .... │ ......... │ ....... │ ............ │ │ 1998 │ UNITED STATES │ MFGR#24 │ 60779388345 │ │ 1998 │ UNITED STATES │ MFGR#25 │ 60042710566 │ └──────┴───────────────┴────────────┴──────────────┘ 100 rows in set. Elapsed: 0.826 sec. Processed 144.42 million rows, 2.17 GB (174.78 million rows/s., 2.63 GB/s.)
扫描行数:144,420,000大约1亿4千多万
耗时(秒):0.826
查询列数:4
结果行数:100
性能测试结果汇总
查询语句 | SQL简要说明 | 扫描行数 | 返回行数 | 查询列数 | 耗时(秒) |
Q1.1 | 乘积、汇总、4个条件、首次运行 | 91,010,000 | 1 | 2 | 0.242 |
Q1.2 | Q1.1增加1个条件运行 | 7,750,000 | 1 | 2 | 0.040 |
Q2.1 | 汇总、函数、2列分组、2列排序、首次运行 | 600,040,000 | 280 | 3 | 8.558 |
Q2.2 | Q2.1增加1个条件运行 | 600,040,000 | 56 | 3 | 1.242 |
Q3.1 | 汇总、函数、3列分组、2列排序、首次运行 | 546,670,000 | 150 | 4 | 3.533 |
Q3.2 | Q3.1更换条件运行 | 546,670,000 | 600 | 4 | 1 |
Q4.1 | 相减、汇总、函数、2列分组、2列排序、首次运行 | 600,040,000 | 35 | 4 | 5.006 |
Q4.2 | Q4.1增加2个条件运行 | 144,420,000 | 100 | 4 | 0.826 |
在当前软硬件环境下,扫描 6 亿多行数据,常见的分析语句首次运行最慢在 8 秒左右能返回结果,相同的分析逻辑更换条件再次查询的时候效率有明显的提升,可以缩短到 1 秒左右,如果只是简单的列查询没有加减乘除、聚合等逻辑,扫描全表 6 亿多行数据首次查询基本可以在 2 秒内执行完成。