Filegroup confusion

Where did I put my Sales table?

Filegroups? What can I say! Sometimes I love them and sometimes I just hate them. In theory this should be quite simple and give you benefits – but in my years as a consultant, I have seen so many bad file group designs where the simplicity is traded in for an over complex layout that looks good on paper, but is a nightmare to maintain.

A scenario that I often come across is misplaced tables and indexes. This happens mostly if the person that created the database layout is not responsible for creating objects afterwards, or if the end-users does not get the proper instructions on how to do things the right way. When this is the case, things can end up messy. I will show you how to query the Meta data, so that you can check if your data is placed in the right filegroup.

Here is the database layout that I’m about to create:

In addition, here is the code used to create the database

CREATE DATABASE [Inspari]  
 ON  PRIMARY   
(   
    NAME = N'System01',   
    FILENAME = N'C:FusionIO_MPSystem01.mdf' ,   
    SIZE = 100Mb,   
    FILEGROWTH = 100Mb  
),  
(   
    NAME = N'System02',   
    FILENAME = N'C:FusionIO_MPSystem02.ndf' ,   
    SIZE = 100Mb,   
    FILEGROWTH = 100Mb  
),   
 FILEGROUP [Data]   
(   
    NAME = N'Data01',   
    FILENAME = N'C:FusionIO_MPData01.ndf' ,   
    SIZE = 100Mb,   
    FILEGROWTH = 100Mb  
),   
(   
    NAME = N'Data02',   
    FILENAME = N'C:FusionIO_MPData02.ndf' ,   
    SIZE = 100Mb,   
    FILEGROWTH = 100Mb  
),   
 FILEGROUP [Index]   
(   
    NAME = N'Index01',   
    FILENAME = N'C:FusionIO_MPIndex01.ndf' ,   
    SIZE = 100Mb,   
    FILEGROWTH = 100Mb  
),   
(   
    NAME = N'Index02',   
    FILENAME = N'C:FusionIO_MPIndex02.ndf' ,   
    SIZE = 100Mb,   
    FILEGROWTH = 100Mb  
)  
 LOG ON   
(   
    NAME = N'Inspari_log',   
    FILENAME = N'C:FusionIO_MPInspari_log.ldf' ,   
    SIZE = 100Mb,   
    FILEGROWTH = 100Mb  
)  
GO  

Let us create a table, pay attention to the ON [Filegroup] keyword

CREATE TABLE RandomPeople  
(  
    Id INT IDENTITY(1,1) PRIMARY KEY,  
    DateCreated DATETIME2 DEFAULT SYSDATETIME(),  
    Name VARCHAR(100),  
    Salary BIGINT not null  
)   
ON [Data]  
GO

Let us create an index on the table again pay attention to the ON [Filegroup] keyword.

CREATE INDEX idx_RandomIndex ON RandomPeople  
(  
    Name, Salary  
)  
ON [Index]  
GO

Now imagine that you have a massive database where multiple people have created tables and indexes, the odds for everything being placed in the right file group is close to not existing. Some object will eventually end up in the wrong file group. Here is a T-SQL script that queries the meta data to show You which objects are stored in which file groups.

SELECT   
  OBJECT_NAME(t1.object_id) as tablename,  
  t3.name AS indexName,  
  t3.type_desc,   
  t1.index_id,   
  t4.name AS FileGroupName,  
  t1.rows AS RowsInObject,  
  (total_pages * 8) / 1024 AS SpaceUsedInMB  
FROM sys.partitions t1  
INNER JOIN sys.allocation_units t2 ON (t2.container_id = t1.hobt_id)  
INNER JOIN sys.indexes t3 on (t1.object_id = t3.object_id) AND (t1.index_id = t3.index_id)  
INNER JOIN sys.filegroups t4 ON (t4.data_space_id = t2.data_space_id)  
WHERE t1.object_id in   
(  
    SELECT object_id  
    FROM sys.objects  
    WHERE is_ms_shipped = 0  
)  
ORDER BY 1

Is your data stored correct? If that is the case, then congratulations – good job. If not, then there is work to do. The good news is that you might discover that your Sales table is located on slow disks, and by moving it to the faster disks, you’ll improve system performance immediately.