Question

我有一个耗时查询，耗时5分钟，我想将其结果存储在一个表中，该表将由后端查询。

此外，该表应每小时更新一次，并用新数据完全替换该表的内容。

我发现了this Server Agent solution，根据我的理解，我应该这样做：

创建结果表。
创建一个存储过程，以删除结果表数据，运行5分钟查询，然后插入新数据。
创建间隔运行该过程的服务器代理作业。

这是最佳方法吗？

这是有问题的查询。它遍历了很多行，我认为这是主要的速度影响器。

WITH 
salesCTE AS (
    select
        concat(year(m.addate), '-', format(m.addate, 'MM')) AS ym,
        year(m.addate) as y,
        format(m.addate, 'MM') as m,
        sum(M.anvalue) as salesRev
    FROM tHE_Move m
    WHERE ( 
        RIGHT(LEFT(M.acKey,5),3) = '300'
        OR RIGHT(LEFT(M.acKey,5),3) = '305'
        OR RIGHT(LEFT(M.acKey,5),3) = '319'
        OR RIGHT(LEFT(M.acKey,5),3) = '380'
        OR RIGHT(LEFT(M.acKey,5),3) = '355'
        OR RIGHT(LEFT(M.acKey,5),3) = '360'
        OR RIGHT(LEFT(M.acKey,5),3) = '3X1'
        OR RIGHT(LEFT(M.acKey,5),3) = '395'
    ) and m.adDate between '01.01.2014' and '01.01.2030'
    GROUP BY 
        concat(year(m.addate), '-', format(m.addate, 'MM')),
        year(m.addate),
        format(m.addate, 'MM')
),
retailCTE AS (
    select
        concat(year(m.addate), '-', format(m.addate, 'MM')) AS ym,
        year(m.addate) as y,
        format(m.addate, 'MM') as m,
        sum(M.anvalue) as retailRev
    FROM tHE_Move m
    where (
        RIGHT(LEFT(M.acKey,5),3) = '321'
        OR RIGHT(LEFT(M.acKey,5),3) = '322'
        OR RIGHT(LEFT(M.acKey,5),3) = '323'
        OR RIGHT(LEFT(M.acKey,5),3) = '324'
        OR RIGHT(LEFT(M.acKey,5),3) = '325'
        OR RIGHT(LEFT(M.acKey,5),3) = '326'
        OR RIGHT(LEFT(M.acKey,5),3) = '327'
        OR RIGHT(LEFT(M.acKey,5),3) = '328'
        OR RIGHT(LEFT(M.acKey,5),3) = '329'
        OR RIGHT(LEFT(M.acKey,5),3) = '331'
        OR RIGHT(LEFT(M.acKey,5),3) = '332'
        OR RIGHT(LEFT(M.acKey,5),3) = '333'
        OR RIGHT(LEFT(M.acKey,5),3) = '334'
        OR RIGHT(LEFT(M.acKey,5),3) = '335'
        OR RIGHT(LEFT(M.acKey,5),3) = '336'
        OR RIGHT(LEFT(M.acKey,5),3) = '337'
        OR RIGHT(LEFT(M.acKey,5),3) = '338'
        OR RIGHT(LEFT(M.acKey,5),3) = '339'
        OR RIGHT(LEFT(M.acKey,5),3) = '341'
        OR RIGHT(LEFT(M.acKey,5),3) = '342'
        OR RIGHT(LEFT(M.acKey,5),3) = '343'
        OR RIGHT(LEFT(M.acKey,5),3) = '344'
        OR RIGHT(LEFT(M.acKey,5),3) = '345'
        OR RIGHT(LEFT(M.acKey,5),3) = '346'
        OR RIGHT(LEFT(M.acKey,5),3) = '347'
        OR RIGHT(LEFT(M.acKey,5),3) = '348'
        OR RIGHT(LEFT(M.acKey,5),3) = '349'
        OR RIGHT(LEFT(M.acKey,5),3) = '352'
        OR RIGHT(LEFT(M.acKey,5),3) = '353'
        ) and m.adDate between '01.01.2014' and '01.01.2030'
    GROUP BY 
        concat(year(m.addate), '-', format(m.addate, 'MM')),
        year(m.addate),
        format(m.addate, 'MM')
)
SELECT 
    s1.ym,
    s1.salesRev,
    (s1.salesRev / s2.salesRev - 1) * 100 salesDelta,
    r1.retailRev,
    (r1.retailRev / r2.retailRev - 1) * 100 retailDelta,
    s1.salesRev + r1.retailRev totalRev,
    ((s1.salesRev + r1.retailRev) / (s2.salesRev + r2.retailRev) - 1) * 100 totalDelta
FROM salesCTE s1
    left join salesCTE s2
        on s2.y = s1.y - 1 and s1.m = s2.m
    left join retailCTE r1
        on s1.ym = r1.ym
    left join retailCTE r2
        on r2.y = r1.y - 1 and r1.m = r2.m
order by s1.ym desc

Answer 1

如评论FORMAT中所述，这是一个巨大的性能冲击力；而我的庞大，我确实的意思是庞大。将此DB<>Fiddle用作FORMAT和CONVERT来更改日期的值。使用CONVERT的查询在46ms时执行（在数据库提琴上），但是FORMAT查询花费了1218 ms！慢了26倍。

将format(m.addate, 'MM')更改为RIGHT('00' + CONVERT(varchar(2),DATEPART(MONTH,DATEADD(DAY,N-1,0))),2)将对您的查询产生重大的性能优势。尽管后者看起来更复杂，但是（如小提琴所示）它将以显着优势胜过FORMAT。 老实说，我建议不要使用FORMAT，Microsoft对该功能确实有误。

不过，我也建议将RIGHT(LEFT(M.acKey,5),3)作为持久列添加到您的表中：

ALTER TABLE tHE_Move ADD {Meaningful Name} AS RIGHT(LEFT(M.acKey,5),3) PERSISTED;

然后，您还可以将该值添加到索引（新索引或现有索引）中，这也将大大提高查询的性能；也许只推几秒钟。

SQL Server：如何每小时运行一次查询，并将结果存储在表中？更新时替换它们

1 个答案: