Data Partitioning (DPF) is technique that splits tables across multiple partitions (on the same DB server or on different DB servers). It was designed to induce query parallelism on SQL statements on data warehouse systems that read large portions of the table wach time. It does not work as well on OLTP systems where only a few rows are accessed via an index.
Even though DB2 DPF can be made to provide some semblance of failover protection and high availability, there are other DB2 facilities such as HADR (which is new in version 8.2) which are more appropriate to OLTP systems and high availability.
However, none of these DB2 features provide a good solution for high availability and scalability in an OLTP environment, such as might be better provided by Oracle RAC. This is because with DPF, the overhead for DB2 figuring out which partition has the data is too high for most OLTP systems, and the HADR solution requires that the backup server be a passive standby system.