Sql server sql server 2008对总和的慢速查询
我对管理sqlserver数据库还不熟悉。我收到了这个请求 我有一张有一百万张唱片的大桌子。要运行此报告。使用SQL Server 2008需要一个多小时 Rowid是PK,在dx和varcharfield上有一个索引Sql server sql server 2008对总和的慢速查询,sql-server,sql-server-2008,cumulative-sum,Sql Server,Sql Server 2008,Cumulative Sum,我对管理sqlserver数据库还不熟悉。我收到了这个请求 我有一张有一百万张唱片的大桌子。要运行此报告。使用SQL Server 2008需要一个多小时 Rowid是PK,在dx和varcharfield上有一个索引 SELECT T1.VarcharField, T1.DX, T1.ROWID, (SELECT Sum(t2.QC) AS Total FROM tbl t2 WHERE T2.VarcharFie
SELECT T1.VarcharField,
T1.DX,
T1.ROWID,
(SELECT Sum(t2.QC) AS Total
FROM tbl t2
WHERE T2.VarcharField = T1.VarcharField
AND t2.dx <= t1.dx) AS Total
FROM tbl AS T1
我认为这将运行得更快:
SELECT
T1.VarcharField,
T1.DX,
T1.ROWID,
SUM(t2.QC) AS Total
FROM
tbl t2
JOIN tb1 T1 ON T2.VarcharField = T1.VarcharField
GROUP BY
T1.VarcharField,
T1.DX,
T1.ROWID
WHERE
t2.dx <= t1.dx
在SQL Server 2012中,这一点要容易得多
SELECT T1.VarcharField,
T1.DX,
T1.ROWID,
SUM(QC) OVER (PARTITION BY VarcharField ORDER BY DX) AS Total
FROM tbl AS T1
由于您在2008年,上述语法不起作用,您可以尝试添加此索引
CREATE INDEX ix ON tbl(VarcharField,DX) INCLUDE (ROWID, QC)
然后使用以下查询
WITH RecursiveCTE
AS (SELECT VarcharField,
DX,
ROWID,
QC,
QC AS Total
FROM (SELECT *,
ROW_NUMBER() OVER (PARTITION BY VarcharField ORDER BY DX) AS RN
FROM tbl) T
WHERE RN = 1
UNION ALL
SELECT R.VarcharField,
R.DX,
R.ROWID,
R.QC,
R.Total
FROM (SELECT T.*,
T.QC + Total AS Total,
rn = ROW_NUMBER() OVER (ORDER BY T.DX)
FROM tbl T
JOIN RecursiveCTE R
ON R.VarcharField = T.VarcharField
AND R.DX < T.DX) R
WHERE R.rn = 1)
SELECT VarcharField,
DX,
ROWID,
Total
FROM RecursiveCTE
OPTION (MAXRECURSION 0);
这种计算运行总计的方法不能扩展到100万行。它将遇到与OP相同的问题。问题是1000000个组平均每个组将合计500000个值。总计有5000亿个值被求和。我没有尝试过,但可能删除where子句并将该条件置于join中有助于提高速度。我认为表示为子选择的求和比分组查询更直接的方法要慢。这就是有效的方法。我添加了上述索引。在tblVarcharField上创建索引ix,DX INCLUDE ROWID,QC选择T1.VarcharField,T1.DX,T1.ROWID,Total=从tbl t2中选择Sumt2.QC作为总计,其中t2.VarcharField=T1.VarcharField和t2.DX@kdigman添加t2.ROWID=T1.ROWID将极大地改变语义。