Sometime SQL Server Select Query is too slow - sql

Question

I have a table like this which has more than 7 million records:
CREATE TABLE [dbo].[Test]
(
[Id] [bigint] IDENTITY(1,1) NOT NULL,
[UUID] [nvarchar](100) NOT NULL,
[FirstName] [nvarchar](50) NULL,
[LastName] [nvarchar](50) NULL,
[AddrLine1] [nvarchar](100) NULL,
[AddrLine2] [nvarchar](100) NULL,
[City] [nvarchar](50) NULL,
[Prov] [nvarchar](10) NULL,
[Postal] [nvarchar](10) NULL,
[DateAdded] [datetime] NULL,
CONSTRAINT [PK_Test]
PRIMARY KEY CLUSTERED ([Id] ASC)
WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF,
IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON,
ALLOW_PAGE_LOCKS = ON, OPTIMIZE_FOR_SEQUENTIAL_KEY = OFF) ON [PRIMARY]
) ON [PRIMARY]
Now, the system runs the following select query everyday during the afternoons. The funny thing is sometimes the same query is so slow which takes about 4 minutes. The second time or other times, same query is pretty fast which takes less than a second.
The query:
WITH testquery AS
(
SELECT TOP 1
'Matched' as location,Firstname, LastName,
AddrLine1, AddrLine2, City, Prov, Postal
FROM
[Test]
WHERE
UUID = 'BLABLABLABLABLABLABLABLABLA'
ORDER BY
DateAdded DESC
),
defaults AS
(
SELECT
'Rejected' AS location, NULL AS Firstname, NULL AS LastName,
NULL AS AddrLine1, NULL AS AddrLine2, NULL AS City, NULL AS Prov,
NULL AS Postal
)
SELECT *
FROM testquery
UNION ALL
SELECT *
FROM defaults
WHERE NOT EXISTS (SELECT * FROM testquery);
Can somebody help please?
Notes:
I have a service which adds around 1000 new records to the table everyday in the mornings.
[avg_fragmentation_in_percent] is 0.01
UUID can be duplicated if I have the same person with different addresses.
The table is not used somewhere else at the same time.
Database is not busy with other queries at the same time. I checked using "sys.dm_exec_requests"

Answer 1

You need a good index to service this query efficiently.
You say that you can't create one because of duplicate key errors: there is no need for an index to be unique.
So the one you're looking for will depend on what other queries you are running, but the following will suffice for this query:
CREATE NONCLUSTERED INDEX IX_Test_UuidDate ON
Test (UUID ASC, DateAdded DESC)
INCLUDE (Firstname, LastName, AddrLine1, AddrLine2, City, Prov, Postal)
GO
Furthermore, there is no need to query the table twice.
Start with a dummy VALUES table constructor so you always have a row, then LEFT JOIN the table and use CASE to deal with not having a row.
WITH testquery AS
(
SELECT TOP 1
*
FROM
[Test]
WHERE
UUID = 'BLABLABLABLABLABLABLABLABLA'
ORDER BY
DateAdded DESC
)
SELECT
CASE WHEN UUID IS NULL 'Rejected' ELSE 'Matched' END as location,
t.Firstname,
t.LastName,
t.AddrLine1,
t.AddrLine2,
t.City,
t.Prov,
t.Postal
FROM (VALUES(0)) AS v(dummy)
LEFT JOIN testquery AS t ON 1=1;

Answer 2

The usual explanation for this is a cold cache. In your case, I think the issue would be the ORDER BY in the first CTE.
To fix this problem, you want an index on test(UUID, DateAdded desc).
I'm not sure why this would speed up after the first execution. Perhaps the server's caches are working particularly well.