SQL 百分位数计算运行非常缓慢 - 需要帮助加快速度

Question

我有一个大表（目前大约有 8.5 亿行），并且需要每周计算百分位值，因为插入了新数据并且静态数据变脏了。但是，该过程非常慢（我的硬件/当前查询需要 5-6 小时）。

如何更改我的查询以加快查询速度？

现在，我的查询基本上是这样的：

SELECT DISTINCT [ident1]
    ,[ident2]
    ,[ident3]
    ,[ident4]
    ,percentile_cont(0.05)
        WITHIN GROUP (
            ORDER BY [value] ASC
        ) OVER (
            PARTITION BY [ident1]
                ,[ident2]
                ,[ident3]
                ,[ident4]
        ) AS [percentile_5]
    ,percentile_cont(0.10)
        WITHIN GROUP (
            ORDER BY [value] ASC
        ) OVER (
            PARTITION BY [ident1]
                ,[ident2]
                ,[ident3]
                ,[ident4]
        ) AS [percentile_10]
    ,percentile_cont(0.25)
        WITHIN GROUP (
            ORDER BY [value] ASC
        ) OVER (
            PARTITION BY [ident1]
                ,[ident2]
                ,[ident3]
                ,[ident4]
        ) AS [percentile_25]
    ,percentile_cont(0.50)
        WITHIN GROUP (
            ORDER BY [value] ASC
        ) OVER (
            PARTITION BY [ident1]
                ,[ident2]
                ,[ident3]
                ,[ident4]
        ) AS [percentile_50]
    ,percentile_cont(0.75)
        WITHIN GROUP (
            ORDER BY [value] ASC
        ) OVER (
            PARTITION BY [ident1]
                ,[ident2]
                ,[ident3]
                ,[ident4]
        ) AS [percentile_75]
    ,percentile_cont(0.90)
        WITHIN GROUP (
            ORDER BY [value] ASC
        ) OVER (
            PARTITION BY [ident1]
                ,[ident2]
                ,[ident3]
                ,[ident4]
        ) AS [percentile_90]
    ,percentile_cont(0.95)
        WITHIN GROUP (
            ORDER BY [value] ASC
        ) OVER (
            PARTITION BY [ident1]
                ,[ident2]
                ,[ident3]
                ,[ident4]
        ) AS [percentile_95]
FROM dataTable

我认为问题的一部分是，如果没有DISTINCT 选择，我在数据库中的每个值都会得到一行。 SQL 是否足够聪明，可以只为每个组计算一次百分位数？还是对每个值重复计算？

任何帮助将不胜感激。

Answer 1

将 distinct 放入子查询中：

select . . .
from (select distinct . . . ) s;

distinct 应该在 计算完所有 select 列之后发生。