Speed up table partitioning

Databases are constantly increasing in size, and more and more IBM i customers are approaching or exceeding the limits for a single DB2 table: 1.7 TB storage or 4.2 billion rows.

DB2 MultiSystem allows for a table to have up to 256 partitions instead of only one, thus raising the limit by a factor of 256. And IBM changed the price for DB2 MultiSystem to zero in summer 2022, so all customers can now use this feature for free.

Changing a table from being non-partitioned to a partitioned table is an extensive operation, since a new structure has to be created and all the data copied into this structure, according to the partition setup (range, hash).

However, the operation runs in a single thread so it is very slow!

The last time we had to partition a non-partitioned table, it took 26 hours(!) creating the new structure, copy the data and rebuild the primary indexes and constraints! This was on a 8 core Power9 with 1 TB memory attached via 3 FC adapters to a V7000 with FCM modules, but the hardware did almost nothing. The application was down all 26 hours because of this.

Please consider changing this operation to use multiple threads and exploit the available hardware. Other operations in the database already do this, e.g. index builds.

Many IBM i customers will experience the same issue as us, as their tables approach the limits, and we will also have to go through this again, when the next table has to be partitioned. And it is difficult for the business to understand, that the application will be down for more than 24 hours, because we have to change a table to allow for more data.