Question

I have a very complex SQL View. My Problem is, that its very slow and need a lot of memory When i want to make a report.
What the View do:
It queries a total value and multiply it with a factor
Obviously it’s because it contains a lot subqueries… (9 if you count it)

The query runs on a system with Millions of datasets.

My question. Where are points to optimize it? Would it be faster, if i make prepared procedures oder create sql views for the subqueries?

SELECT 
    CASE BID WHEN 1 THEN 
        b.TOTAL 
    ELSE 
        (b.TOTAL * - 1) 
    END 
    * 
    CASE WHEN
        (SELECT 
            (CAST(DATEDIFF(day, 
                ISNULL
                ((SELECT dbo.TableA.TimestampFrom + 1 AS start FROM dbo.TableB INNER JOIN
                    dbo.TableA ON dbo.TableB.SID = dbo.TableA.SID
                    WHERE   (dbo.TableB.PID = dbo.TableC.PID) AND
                        (dbo.TableB.MString = 'S' OR dbo.TableB.MString = 'ST') AND
                        (dbo.TableB.TimestampCreated <
                            (SELECT TimestampCreated
                                FROM dbo.TableB AS TableB_2
                            WHERE (RID = dbo.TableC.RID)))
                        AND
                            (dbo.TableB.VPFrom =
                                    (SELECT VPFrom FROM dbo.TableB AS TableB_1
                                            WHERE (RID = dbo.TableC.RID))
                            )), VPFrom), VPTo)
                            AS Float) + 1)
            / 
            (CAST(DATEDIFF(day, start, Vertragsende) AS Float) + 1) AS Faktor

            FROM dbo.TableC
                WHERE (LaufNr = b.LaufNr))
                    > 1 
        THEN 
            1 
        ELSE
            (SELECT (CAST(DATEDIFF(day, 
                ISNULL
                    ((SELECT dbo.TableA.TimestampFrom + 1 AS start FROM dbo.TableB INNER JOIN
                            dbo.TableA ON dbo.TableB.SID =  dbo.TableA.SID
                            WHERE (dbo.TableB.PID = dbo.TableC.PID) AND
                                (dbo.TableB.MString = 'S' OR dbo.TableB.MString = 'ST') AND
                                (dbo.TableB.TimestampCreated <
                                    (SELECT TimestampCreated FROM dbo.TableB AS TableB_2
                                     WHERE (RID = dbo.TableC.RID)))
                                AND (dbo.TableB.VPFrom =
                                    (SELECT VPFrom FROM dbo.TableB AS TableB_1
                                    WHERE (RID = dbo.TableC.RID)))
                        ), VPFrom),VPTo) AS Float) + 1)
                / (CAST(DATEDIFF(day, start, Vertragsende) AS Float) + 1) AS Faktor 
                FROM dbo.TableC
                WHERE (LaufNr = b.LaufNr)) 
        END AS GrossPremium
    ,RID, PID, ONR, OPosLaufNr, LaufNr, bdate
FROM dbo.TableC AS b

Answer 1

The rules of thumb for indexing are to

• first, read the execution plan,

then think about adding an index to

• every column used in a WHERE clause,
• every column used in a JOIN condition,
• every column used in an ORDER BY.

In some cases, one multi-column index will be faster in SELECT statements than several single-column indexes.

Then, make sure your WHERE clauses are sargable.

Finally, it does make sense to try converting some of the subqueries into views for testing. My guess is that the root cause of the problem is that you’re executing different subqueries depending on the return value of several CASE statements.