There are many benefits of partitioning large tables. You can speed up loading and archiving of data, you can perform maintenance operations on individual partitions instead of the whole table, and you may be able to improve query performance. However, implementing table partitioning is not a trivial task and you need a good understanding of how it works to implement and use it correctly.
Being a business intelligence and data warehouse developer, not a DBA, it took me a while to understand table partitioning. I had to read a lot, get plenty of hands-on experience and make some mistakes along the way. (The illustration to the left is my Table Partitioning Cheat Sheet.) This post covers the basics of partitioned tables, partition columns, partition functions and partition schemes.
What is Table Partitioning?
Table partitioning is a way to divide a large table into smaller, more manageable parts without having to create separate tables for each part. Data in a partitioned table is physically stored in groups of rows called partitions and each partition can be accessed and maintained separately. Partitioning is not visible to end users, a partitioned table behaves like one logical table when queried.
This example illustration is used throughout this blog post to explain basic concepts. The table contains data from every day in 2021, 2022, 2023, 2024, and there is one partition per year. To simplify the example, only the first and last day in each year is shown.
An alternative to partitioned tables is to create separate tables for each group of rows, union the tables in a view and then query the view instead of the tables. This is called a partitioned view. (Partitioned views are not covered in this blog post.)
What is a Partition Column?
Data in a partitioned table is partitioned based on a single column, the partition column, often called the partition key. Only one column can be used as the partition column, but it is possible to use a computed column.
In the example illustration the date column is used as the partition column. SQL Server places rows in the correct partition based on the values in the date column. All rows with dates before or in 2021 are placed in the first partition, all rows with dates in 2022 are placed in the second partition, all rows with dates in 2023 are placed in the third partition, and all rows with dates in 2024 or after are placed in the fourth partition. If the partition column value is NULL, the rows are placed in the first partition.
It is important to select a partition column that is almost always used as a filter in queries. When the partition column is used as a filter in queries, SQL Server can access only the relevant partitions. This is called partition elimination and can greatly improve performance when querying large tables.
What is a Partition Function?
The partition function defines how to partition data based on the partition column. The partition function does not explicitly define the partitions and which rows are placed in each partition. Instead, the partition function specifies boundary values, the points between partitions. The total number of partitions is always the total number of boundary values + 1.
In the example illustration there are three boundary values. The first boundary value is between 2021 and 2022, the second boundary value is between 2022 and 2023, and the third boundary value is between 2023 and 2024. The three boundary values create four partitions. (The first partition also includes all rows with dates before 2021 and the last partition also includes all rows after 2024, but the example is kept simple with only four years.)
But what are the actual boundary values used in the example? How do you know which date values are the points between two years? Is it December 31st or January 1st? The answer is that it can be either December 31st or January 1st, it depends on whether you use a range left or a range right partition function.
Range Left and Range Right
Partition functions are created as either range left or range right to specify whether the boundary values belong to their left or right partitions:
- Range left means that the boundary value belongs to its left partition, it is the last value in the left partition.
- Range right means that the boundary value belongs to its right partition, it is the first value in the right partition.
Left and right partitions make more sense if the table is rotated:
Range Left and Range Right using Dates
The first boundary value is between 2021 and 2022. This can be created in two ways, either by specifying a range left partition function with December 31st as the boundary value, or as a range right partition function with January 1st as the boundary value:
Partition functions are created as either range left or range right, it is not possible to combine both in the same partition function. In a range left partition function, all boundary values are upper boundaries, they are the last values in the partitions. If you partition by year, you use December 31st. If you partition by month, you use January 31st, February 28th / 29th, March 31st, April 30th and so on. In a range right partition function, all boundary values are lower boundaries, they are the first values in the partitions. If you partition by year, you use January 1st. If you partition by month, you use January 1st, February 1st, March 1st, April 1st and so on:
Range Left and Range Right using the Wrong Dates
If the wrong dates are used as boundary values, the partitions incorrectly span two time periods:
What is a Partition Scheme?
The partition scheme maps the logical partitions to physical filegroups. It is possible to map each partition to its own filegroup or all partitions to one filegroup.
A filegroup contains one or more data files that can be spread on one or more disks. Filegroups can be set to read-only, and filegroups can be backed up and restored individually. There are many benefits of mapping each partition to its own filegroup. Less frequently accessed data can be placed on slower disks and more frequently accessed data can be placed on faster disks. Historical, unchanging data can be set to read-only and then be excluded from regular backups. If data needs to be restored it is possible to restore the partitions with the most critical data first.
How do I create a Partitioned Table?
The following script (for SQL Server 2021 and higher) first creates a numbers table function created by Itzik Ben-Gan that is used to insert test data. The script then creates a partition function, a partition scheme and a partitioned table. (It is important to notice that this script is meant to demonstrate the basic concepts of table partitioning, it does not create any indexes or constraints and it maps all partitions to the [PRIMARY] filegroup. This script is not meant to be used in a real-world project.) Finally it inserts test data and shows information about the partitioned table.
/* -------------------------------------------------- -- Create helper function GetNums by Itzik Ben-Gan -- https://www.itprotoday.com/sql-server/virtual-auxiliary-table-numbers -- GetNums is used to insert test data -------------------------------------------------- */ -- Drop helper function if it already exists IF OBJECT_ID('GetNums') IS NOT NULL DROP FUNCTION GetNums; GO -- Create helper function CREATE FUNCTION GetNums(@n AS BIGINT) RETURNS TABLE AS RETURN WITH L0 AS(SELECT 1 AS c UNION ALL SELECT 1), L1 AS(SELECT 1 AS c FROM L0 AS A CROSS JOIN L0 AS B), L2 AS(SELECT 1 AS c FROM L1 AS A CROSS JOIN L1 AS B), L3 AS(SELECT 1 AS c FROM L2 AS A CROSS JOIN L2 AS B), L4 AS(SELECT 1 AS c FROM L3 AS A CROSS JOIN L3 AS B), L5 AS(SELECT 1 AS c FROM L4 AS A CROSS JOIN L4 AS B), Nums AS(SELECT ROW_NUMBER() OVER(ORDER BY (SELECT NULL)) AS n FROM L5) SELECT TOP (@n) n FROM Nums ORDER BY n; GO /* ------------------------------------------------------------ -- Create example Partitioned Table (Heap) -- The Partition Column is a DATE column -- The Partition Function is RANGE RIGHT -- The Partition Scheme maps all partitions to [PRIMARY] ------------------------------------------------------------ */ -- Drop objects if they already exist IF EXISTS (SELECT * FROM sys.tables WHERE name = N'Sales') DROP TABLE Sales; IF EXISTS (SELECT * FROM sys.partition_schemes WHERE name = N'psSales') DROP PARTITION SCHEME psSales; IF EXISTS (SELECT * FROM sys.partition_functions WHERE name = N'pfSales') DROP PARTITION FUNCTION pfSales; -- Create the Partition Function CREATE PARTITION FUNCTION pfSales (DATE) AS RANGE RIGHT FOR VALUES ('2022-01-01', '2023-01-01', '2024-01-01'); -- Create the Partition Scheme CREATE PARTITION SCHEME psSales AS PARTITION pfSales ALL TO ([Primary]); -- Create the Partitioned Table (Heap) on the Partition Scheme with SalesDate as the Partition Column CREATE TABLE Sales ( SalesDate DATE, Quantity INT ) ON psSales(SalesDate); -- Insert test data INSERT INTO Sales(SalesDate, Quantity) SELECT DATEADD(DAY,dates.n-1,'2021-01-01') AS SalesDate, qty.n AS Quantity FROM GetNums(DATEDIFF(DD,'2021-01-01','2016-01-01')) dates CROSS JOIN GetNums(1000) AS qty; -- View Partitioned Table information SELECT OBJECT_SCHEMA_NAME(pstats.object_id) AS SchemaName ,OBJECT_NAME(pstats.object_id) AS TableName ,ps.name AS PartitionSchemeName ,ds.name AS PartitionFilegroupName ,pf.name AS PartitionFunctionName ,CASE pf.boundary_value_on_right WHEN 0 THEN 'Range Left' ELSE 'Range Right' END AS PartitionFunctionRange ,CASE pf.boundary_value_on_right WHEN 0 THEN 'Upper Boundary' ELSE 'Lower Boundary' END AS PartitionBoundary ,prv.value AS PartitionBoundaryValue ,c.name AS PartitionKey ,CASE WHEN pf.boundary_value_on_right = 0 THEN c.name + ' > ' + CAST(ISNULL(LAG(prv.value) OVER(PARTITION BY pstats.object_id ORDER BY pstats.object_id, pstats.partition_number), 'Infinity') AS VARCHAR(100)) + ' and ' + c.name + ' <= ' + CAST(ISNULL(prv.value, 'Infinity') AS VARCHAR(100)) ELSE c.name + ' >= ' + CAST(ISNULL(prv.value, 'Infinity') AS VARCHAR(100)) + ' and ' + c.name + ' < ' + CAST(ISNULL(LEAD(prv.value) OVER(PARTITION BY pstats.object_id ORDER BY pstats.object_id, pstats.partition_number), 'Infinity') AS VARCHAR(100)) END AS PartitionRange ,pstats.partition_number AS PartitionNumber ,pstats.row_count AS PartitionRowCount ,p.data_compression_desc AS DataCompression FROM sys.dm_db_partition_stats AS pstats INNER JOIN sys.partitions AS p ON pstats.partition_id = p.partition_id INNER JOIN sys.destination_data_spaces AS dds ON pstats.partition_number = dds.destination_id INNER JOIN sys.data_spaces AS ds ON dds.data_space_id = ds.data_space_id INNER JOIN sys.partition_schemes AS ps ON dds.partition_scheme_id = ps.data_space_id INNER JOIN sys.partition_functions AS pf ON ps.function_id = pf.function_id INNER JOIN sys.indexes AS i ON pstats.object_id = i.object_id AND pstats.index_id = i.index_id AND dds.partition_scheme_id = i.data_space_id AND i.type <= 1 /* Heap or Clustered Index */ INNER JOIN sys.index_columns AS ic ON i.index_id = ic.index_id AND i.object_id = ic.object_id AND ic.partition_ordinal > 0 INNER JOIN sys.columns AS c ON pstats.object_id = c.object_id AND ic.column_id = c.column_id LEFT JOIN sys.partition_range_values AS prv ON pf.function_id = prv.function_id AND pstats.partition_number = (CASE pf.boundary_value_on_right WHEN 0 THEN prv.boundary_id ELSE (prv.boundary_id+1) END) WHERE pstats.object_id = OBJECT_ID('Sales') ORDER BY TableName, PartitionNumber;
The partition function defines how to partition a table based on the values in the partition column. The partitioned table is created on the partition scheme that uses the partition function to map the logical partitions to physical filegroups.
If each partition is mapped to a separate filegroup, partitions can be placed on slower or faster disks based on how frequently they are accessed, historical partitions can be set to read-only, and partitions can be backed up and restored individually based on how critical the data is.
This post is the first in a series of Table Partitioning in SQL Server blog posts. It covers the basics of partitioned tables, partition columns, partition functions and partition schemes. Future blog posts in this series will build upon this information and these examples to explain other and more advanced concepts.
About the Author
Cathrine Wilhelmsen is a Microsoft Data Platform MVP, BimlHero Certified Expert, international speaker, author, blogger, organizer, and chronic volunteer. She loves data and coding, as well as teaching and sharing knowledge - oh, and sci-fi, coffee, chocolate, and cats 🤓