A beginner’s guide to natural and surrogate database keys
Types of primary keys
All database tables must have one primary key column. The primary key uniquely identifies a row within a table therefore it’s bound by the following constraints:
- NOT NULL
When choosing a primary key we must take into consideration the following aspects:
- the primary key may be used for joining other tables through a foreign key relationship
- the primary key usually has an associated default index, so the more compact the data type the less space the index will take
- a simple key performs better than a compound one
- the primary key assignment must ensure uniqueness even in highly concurrent environments
When choosing a primary key generator strategy the options are:
- natural keys, using a column combination that guarantees individual rows uniqueness
- surrogate keys, that are generated independently of the current row data
Natural keys’ uniqueness is enforced by external factors (e.g. person unique identifiers, social security numbers, vehicle identification numbers).
Natural keys are convenient because they have an outside world equivalent and they don’t require any extra database processing. We can, therefore, know the primary key even before inserting the actual row into the database, which simplifies batch inserts.
If the natural key is a single numeric value the performance is comparable to that of surrogate keys.
For compound keys we must be aware of possible performance penalties:
- compound key joins are slower than single key ones
- compound key indexes require more space than their single key counterparts
Non-numerical keys are less efficient than numeric ones (integer, bigint), for both indexing and joining. A CHAR(17) natural key (e.g. vehicle identification number) occupies 17 bytes as opposed to 4 bytes (32 bit integer) or 8 bytes (64 bit bigint).
The initial schema design uniqueness assumptions may not forever hold true. Let’s say we’d used one specific country citizen numeric code for identifying all application users. If we now need to support other countries that don’t have such citizen numeric code or the code clashed with existing entries, then we can conclude that the schema evolution is possibly hindered.
If the natural key uniqueness constraints change it’s going to be very difficult to update both the primary keys (if we manage to drop the primary key constraints anyway) and all associated foreign key relationships.
Surrogate keys are generated independently of the current row data, so the other column constraints may freely evolve according to the application business requirements.
The database system may manage the surrogate key generation and most often the key is of a numeric type (e.g. integer or bigint), is incremented whenever there is a need for a new key.
If we want to control the surrogate key generation we can employ a 128-bit GUID or UUID. This simplifies batching and may improve the insert performance since the additional database key generation processing is no longer required. Even if this strategy is not so widely adopted it’s worth considering when designing the database model.
When the database identifier generation responsibility falls to the database system, there are several strategies for auto incrementing surrogate keys:
|Database engine||Auto incrementing strategy|
|Oracle||SEQUENCE, IDENTITY (Oracle 12c)|
|MSSQL||IDENTITY, SEQUENCE (MSSQL 2012)|
|PostgreSQL||SEQUENCE, SERIAL TYPE|
Because sequences may be called concurrently from different transactions they are usually transaction-less.
Both the IDENTITY type and the SEQUENCE generator are defined by the SQL:2003 standard, so they’ve become the standard primary key generator strategies.
Some database engines allow you to choose between IDENTITY and SEQUENCE so you have to decide which one better suits your current schema requirements.
Hibernate disables JDBC insert batching when using the IDENTITY generator strategy.
Download free ebook sample
If you subscribe to my newsletter, you'll get:
- A free sample of my Video Course about running Integration tests at warp-speed using Docker and tmpfs
- 3 chapters from my book, High-Performance Java Persistence,
- a 10% discount coupon for my book.