Question

The following query is being used to create a pie within a business intelligence dashboard. The query runs and returns the correct data, however it is extremely slow (about 36 seconds). Can anyone see a way to optimize this query?

SELECT 
Name,
Date,
Severity,
PMissing


FROM
(
SELECT
Name,
Date,
Severity,
PMissing,

DENSE_RANK() OVER (PARTITION BY Name ORDER BY ate DESC) AS Rnk

FROM PctbyP

WHERE (Name like '%front-%'
OR Name like '%back-%'
OR Name like '%ap-%'
OR Name like '%mps-%'
OR Name like '%mg-%')
)t
WHERE rnk=1
AND PMissing != '0'

Answer 1

Create an indexed view:

CREATE VIEW
        V_PctbyP_Name
WITH SCHEMABINDING
AS
SELECT  Name, COUNT_BIG(*) AS Cnt
FROM    dbo.PctbyP
GROUP BY
        Name

CREATE UNIQUE CLUSTERED INDEX
        UX_V_PctbyP_Name
ON      V_PctbyP_Name (Name)

, then create an index on the table:

CREATE INDEX
        IX_PctbyP_Name_Date
ON      PctbyP (Name, [Date])

, then use this query:

SELECT  pa.*
FROM    V_PctbyP_Name vp
CROSS APPLY
        (
        SELECT  TOP 1 WITH TIES
                Name, [Date], Severity, PMissing
        FROM    PctbyP p
        WHERE   p.Name = vp.Name
        ORDER BY
                [Date] DESC
        ) pa
WHERE   (
        vp.Name LIKE '%front-%'
        OR
        vp.Name LIKE '%back-%'
        OR
        vp.Name LIKE '%ap-%'
        OR
        vp.Name LIKE '%mps-%'
        OR
        vp.Name LIKE '%mg-%'
        )
        AND pa.PMissing <> '0'