Database Schema in ScalarDB

ScalarDB has its own data model and schema, that maps to the implementation specific data model and schema. Also, it stores internal metadata for managing transaction logs and statuses. This document briefly explains the ScalarDB data model, how data types are mapped between ScalarDB and the other database implementations, internal metadata, and how ScalarDB database schema can be defined.

Data model

The data model of ScalarDB is a multi-dimensional map based on the key-value data model. A logical record is composed of partition-key, clustering-key and a set of values. The value is uniquely mapped by a primary key composed of partition-key, clustering-key and value-name as described in the following scheme.

(partition-key, clustering-key, value-name) -> value-content

For each database implementation, there is an adapter that converts the database specific data model into the ScalarDB data model, thus; users usually don't need to care about how it is converted and can design database schema on the basis of the ScalarDB data model.

Data type mapping between ScalarDB and the other databases

Here are the supported data types in ScalarDB and their mapping to the data types of other databases.

ScalarDB	Cassandra	Cosmos DB	DynamoDB	MySQL	PostgreSQL	Oracle	SQL Server
BOOLEAN	boolean	boolean (JSON)	BOOL	boolean	boolean	number(1)	bit
INT	int	number (JSON)	N	int	int	int	int
BIGINT	bigint	number (JSON)	N	bigint	bigint	number(19)	bigint
FLOAT	float	number (JSON)	N	double	float	binary_float	float(24)
DOUBLE	double	number (JSON)	N	double	double precision	binary_double	float
TEXT	text	string (JSON)	S	longtext	text	varchar2(4000)	varchar(8000)
BLOB	blob	string (JSON)	B	longblob	bytea	blob	varbinary(8000)

Internal metadata

ScalarDB manages metadata (e.g., transaction ID, record version, transaction status) stored along with the actual records to handle transactions properly. Thus, along with any required values by the application, additional values for the metadata need to be defined in the schema.

Schema creation

It is a little hard for application developers to care for the schema mapping and metadata for transactions, so we offer a tool called Schema Loader for creating schema without much knowledge about those.