A database is simply a collection of data. A database manager is the software that allows users to store and access data in a database. It achieves this function by using system resources, including CPU, memory, disk, and communications. In a partitioned database system, a single database manager and the collection of data and system resources that it manages are referred to collectively as a database partition server (node). A partitioned database system is the collection of all the database partition servers that you create to handle data requests.
In DB2 Enterprise - Extended Edition (DB2 EEE), the quality, functionality, reliability, and robustness of a the database manager is extended to the entire partitioned database system.
In a partitioned database system, multiple database partition servers can be assigned to a machine (or to multiple machines), and the database manager at each machine is responsible for a portion of a database's total data (each database partition server houses a portion of the entire database). This portion of the database is known as a database partition (node). The fact that databases are partitioned across database partition servers is transparent to users and applications.
A partitioned database system can maintain very large databases and open opportunities for new applications. DB2 EEE provides fast response time for both decision-support (DSS) and online transaction processing (OLTP) applications.
DB2 EEE can be configured to execute on a shared-nothing hardware architecture, in which machines do not compete for resources. Each machine has exclusive access to its own disks and memory, and the database partition servers that run on the machines communicate with each other through the use of messages. For a database system to exploit shared-nothing architecture, typically one database partition server is assigned to each machine. Another possible configuration is running multiple logical nodes, in which more than one database partition server runs on a machine. For more information, see Multiple Logical Nodes.