Embedded vs. In-Memory vs. Server Databases

Software developers often interact with server oriented databases. However, there are at least two other types of databases, embedded and in-memory, which are less known to the software developer community. Here are the three types I would like to broadly classify the databases into:

a) Server Centric: They usually require a separate installation of the database server component. These are the most common types of databases, for example, SQL Server, Oracle, MySQL, Postgres, DB2, and so on. The database engine of these databases are hosted in a separate process. The applications trying to access data in these databases need to first connect using a database connectivity technology such as ODBC, OLE DB, ADO.Net, etc., and then execute SQL commands. The actual execution of the SQL queries are performed by the database engine.

b) Embedded: These databases do not require installation of any database components; often they are called as server-less or embedded databases. Some examples are SQLite, Firebird, MS Access, Derby, etc. The applications making use of these databases can create and manipulate databases from within their application itself, without requiring any additional software to be installed. These databases support SQL types command. There advantage is that they are easy to setup and require very little maintenance.

c) In-memory: These are the fastest of the breed. The database engine becomes part of the application accessing the database. Few common in-memory databases are Berkley DB, Microsoft's ESENT, Perst, etc. In order to make the data access faster, these databases do not provide support for SQL type command. In any other database type, the support of SQL forces additional layer of parsing to perform data access, and hence make them slower as compared with in-memory datavase. In fact, applications are required to use special APIs to access and manipulate data.

The in-memory databases are hardest to implement, although they have the best performance. However, there are wrappers available to ease out the development process, but still they do require additional work on the developer's end.