Unique identifier as primary key

What is better? Using GUIDs or Integer values? This is been an age long (religious) debate and there are advocates in both camps stressing on the disadvantages of the other. Both implementations have their advantages and disadvantages. In this blogpost, we’ll see that the ultimate SQL Server answer is also valid for this debate: it all depends!

Pro’s

– Pre SQL2012

A 128-bit GUID (unique identifier) key is of course 4x larger than a 32-bit int key. However, there are a some advantages:
– No “IDENTITY INSERT” issue when merging content
– If you use a COMB value instead of NEWSEQUENTIALID(), you get a “free” INSERT timestamp. You can even SELECT from the primary key based on a date/time range.
– They are globally unique.
– There is no need to track returned inserts, so DAL layers can assign the value rather than SQL Server, eliminating the step of SELECT scope_identity() to get the primary key after an insert.
– Using GUIDs makes it safer to expose keys in URLs, etc. without exposing yourself to “guess-the-ID” attacks.
– Can be fairly efficient with point seeks
– Is very efficient in complex  replication or multi server/environment scenario’s as every key is guaranteed to be unique

– Post SQL 2012

– Using sequences, you can also eliminate the step of scope_identity() to get a new key

Cons

– If you use them as Clustered Primary key, they do cause fragmentation because the insertion point of a new record is dictated by the value of the index key. As the key value is completely random, a random key insert will cause page splits if the index is full.
– You have to use a lower fillfactor to cope with the random inserts.
– Has the same dramatic effect in non-clustered indexes as well
– Both the lower fillfactor, and the fragmentation lead to wasted space.

– Will often turn seeks into scans because of the fragmentation and wasted space, this is actually a very bad thing.

How to solve this

– If you don’t need the key features of the unique identifier, use a numeric value instead
o Key features are:

•  Uniqueness over machines (merge/replication purposes)
•  Your application Can not use sequences or scope_identity
•  Is used to make a FK constraint with items not contained on the same DB server
• If you have/want to use unique identifiers, try using NEWSEQUENTIALID instead of NEWID ()
• If you cannot use a NewsequentialID, store a COMB value instead of the uniqueid
• Include the unique identifier where possible instead of using it as a key column
• Definitely do not place a clustered index on a column containing newid() values
• Be careful with FK lookups on uniqueid, it’s better to do the lookups on integer values and store the uniqueid as a property

Demo – INT

USE test
GO
SET NOCOUNT ON;
go
CREATE TABLE DemoUnique (
c1 INT IDENTITY(1,1),
c2 DATETIME DEFAULT GETDATE (),
c3 VARCHAR(max) DEFAULT ‘Somevalue’,

)
GO
CREATE CLUSTERED INDEX CIX ON DemoUnique (c1) WITH (FILLFACTOR=0);
GO
INSERT INTO DemoUnique DEFAULT VALUES;
GO 50000

SELECT
OBJECT_NAME (ips.[object_id]) AS ‘Object Name’,
si.name AS ‘Index Name’,
ROUND (ips.avg_fragmentation_in_percent, 2) AS ‘Fragmentation’,
ips.page_count AS ‘Pages’,
ROUND (ips.avg_page_space_used_in_percent, 2) AS ‘Page Density’
FROM sys.dm_db_index_physical_stats (DB_ID (‘test’), NULL, NULL, NULL, ‘DETAILED’) ips
CROSS APPLY sys.indexes si
WHERE
si.object_id = ips.object_id
AND si.index_id = ips.index_id
AND ips.index_level = 0;
GO
DROP TABLE DemoUnique
go
USE master
go
DROP DATABASE test

Results

Demo – Clustered

USE test
go
CREATE TABLE DemoUnique (
c1 UNIQUEIDENTIFIER DEFAULT NEWID(),
c2 DATETIME DEFAULT GETDATE (),
c3 VARCHAR(max) DEFAULT ‘Somevalue’,

)
GO
CREATE CLUSTERED INDEX CIX ON DemoUnique (c1) WITH (FILLFACTOR=95);
GO
INSERT INTO DemoUnique DEFAULT VALUES;
GO 50000

SELECT
OBJECT_NAME (ips.[object_id]) AS ‘Object Name’,
si.name AS ‘Index Name’,
ROUND (ips.avg_fragmentation_in_percent, 2) AS ‘Fragmentation’,
ips.page_count AS ‘Pages’,
ROUND (ips.avg_page_space_used_in_percent, 2) AS ‘Page Density’
FROM sys.dm_db_index_physical_stats (DB_ID (‘test’), NULL, NULL, NULL, ‘DETAILED’) ips
CROSS APPLY sys.indexes si
WHERE
si.object_id = ips.object_id
AND si.index_id = ips.index_id
AND ips.index_level = 0;
GO
DROP TABLE DemoUnique
go
USE master
go
DROP DATABASE test

Results

Demo – SequentialID

USE test
GO
SET NOCOUNT ON;
go
CREATE TABLE DemoUnique (
c1 UNIQUEIDENTIFIER DEFAULT NEWSEQUENTIALID(),
c2 DATETIME DEFAULT GETDATE (),
c3 VARCHAR(max) DEFAULT ‘Somevalue’,

)
GO
CREATE CLUSTERED INDEX CIX ON DemoUnique (c1) WITH (FILLFACTOR=100);
GO
INSERT INTO DemoUnique DEFAULT VALUES;
GO 50000

SELECT
OBJECT_NAME (ips.[object_id]) AS ‘Object Name’,
si.name AS ‘Index Name’,
ROUND (ips.avg_fragmentation_in_percent, 2) AS ‘Fragmentation’,
ips.page_count AS ‘Pages’,
ROUND (ips.avg_page_space_used_in_percent, 2) AS ‘Page Density’
FROM sys.dm_db_index_physical_stats (DB_ID (‘test’), NULL, NULL, NULL, ‘DETAILED’) ips
CROSS APPLY sys.indexes si
WHERE
si.object_id = ips.object_id
AND si.index_id = ips.index_id
AND ips.index_level = 0;
GO
DROP TABLE DemoUnique
go
USE master
go
DROP DATABASE test

Results

Demo – Includes

CREATE DATABASE Test
GO
USE test
go
CREATE TABLE DemoUnique (
c1 UNIQUEIDENTIFIER DEFAULT NEWID () ROWGUIDCOL,
c2 DATETIME DEFAULT GETDATE (),
c3 VARCHAR(max) DEFAULT ‘Somevalue’);
CREATE NONCLUSTERED INDEX Demo_W_Unique_NCL ON DemoUnique (C2,C1);
CREATE NONCLUSTERED INDEX Demo_W_Incl_Unqiue ON DemoUnique (C2) INCLUDE (C1);
CREATE CLUSTERED INDEX CIX ON DemoUnique (c2) WITH (FILLFACTOR=95);
GO
INSERT INTO DemoUnique DEFAULT VALUES;
GO 50000

SELECT
OBJECT_NAME (ips.[object_id]) AS ‘Object Name’,
si.name AS ‘Index Name’,
ROUND (ips.avg_fragmentation_in_percent, 2) AS ‘Fragmentation’,
ips.page_count AS ‘Pages’,
ROUND (ips.avg_page_space_used_in_percent, 2) AS ‘Page Density’
FROM sys.dm_db_index_physical_stats (DB_ID (‘test’), NULL, NULL, NULL, ‘DETAILED’) ips
CROSS APPLY sys.indexes si
WHERE
si.object_id = ips.object_id
AND si.index_id = ips.index_id
AND ips.index_level = 0;
GO
DROP TABLE DemoUnique
go
USE master
go
DROP DATABASE test

Results

Demo Comb Values

CREATE DATABASE Test
GO
USE test
go
CREATE TABLE DemoUnique (
c2 DATETIME DEFAULT GETDATE (),
c3 VARCHAR(max) DEFAULT ‘Somevalue’,
c4 CHAR(64) DEFAULT CONVERT(VARCHAR(24),GETDATE(),126) +’|’+ CAST(NEWID() AS VARCHAR(40)) — Comb value
)
GO
CREATE NONCLUSTERED INDEX Demo_W_Comb_NCL ON DemoUnique (c4);
CREATE CLUSTERED INDEX CIX ON DemoUnique (c2) WITH (FILLFACTOR=95);
GO
INSERT INTO DemoUnique DEFAULT VALUES;
GO 50000

SELECT
OBJECT_NAME (ips.[object_id]) AS ‘Object Name’,
si.name AS ‘Index Name’,
ROUND (ips.avg_fragmentation_in_percent, 2) AS ‘Fragmentation’,
ips.page_count AS ‘Pages’,
ROUND (ips.avg_page_space_used_in_percent, 2) AS ‘Page Density’
FROM sys.dm_db_index_physical_stats (DB_ID (‘test’), NULL, NULL, NULL, ‘DETAILED’) ips
CROSS APPLY sys.indexes si
WHERE
si.object_id = ips.object_id
AND si.index_id = ips.index_id
AND ips.index_level = 0;
GO
DROP TABLE DemoUnique
go
USE master
go
DROP DATABASE test

Results

Conclusion

If you don’t have a specific reason to use GUIDs, use (BIG)INTs.
Try not to use a GUID as an (first) index key
If you need a GUID as clustered key, make sure you address the randomness issue: use sequential guids

If you need a GUID as nonclustered key, try to add them as an include if they are nonsequential