maybe food for thought....

the random data is a bit simplistic...but relatively large.

not sure if this meets your requirements.

/*set up some simple test data*/

use tempdb

go

If exists(select * from information_schema.TABLES where TABLE_NAME = 'Target') drop table Target

If exists(select * from information_schema.TABLES where TABLE_NAME = 'Use_value') drop table Use_value

SELECT TOP 1000000

Something = '1' + CAST(Abs(Checksum(Newid()) % 90000000 ) AS varchar(50)),

SomethingElse =cast ( '0' AS varchar(50))

INTO Target

FROM sys.all_columns ac1

CROSS JOIN sys.all_columns ac2

CROSS JOIN sys.all_columns ac3

/*set 'somethingelse' to eqaul 'something' for testing*/

UPDATE Target

SET SomethingElse = Something

SELECT TOP 10000

number = IDENTITY(INT, 1, 1),

col1 = CHAR(Abs(Checksum(Newid())) % 26 + 65) + CHAR(Abs(Checksum(Newid())) % 26 + 65) + CHAR(Abs(Checksum(Newid())) % 26 + 65) + CHAR(Abs(Checksum(Newid())) % 26 + 65) + CHAR(Abs(Checksum(Newid())) % 26 + 65),

col2 = 1 + CAST(Abs(Checksum(Newid()) % 9 + 1) AS INT)

INTO Use_value

FROM sys.all_columns ac1

CROSS JOIN sys.all_columns ac2

CROSS JOIN sys.all_columns ac3

ALTER TABLE [dbo].[Use_value] ADD CONSTRAINT [PK_Use_value] PRIMARY KEY CLUSTERED ([number] ASC)

--SELECT TOP (100) Something,

--SomethingElse

--FROM Target

--SELECT TOP (100) number,

--col1,

--col2

--FROM Use_value

/*possible solution*/

SELECT TOP (100) Something,

SomethingElse

FROM Target

order by Something

/*create new table with random selection*/

If exists(select * from information_schema.TABLES where TABLE_NAME = 'T1') drop table T1

DECLARE @cnt AS INT

SELECT @cnt = count(*) FROM use_value

SELECT *,

randCol = (1 + CAST(Abs(Checksum(Newid()) % @cnt) AS INT))

INTO T1

FROM Target

DROP TABLE Target

SELECT T1.Something,

Use_value.col1 AS SomethingElse

INTO Target

FROM T1

INNER JOIN Use_value ON T1.randCol = Use_value.number

SELECT TOP (100) Something,

SomethingElse

FROM Target

order by Something