|Partitioning Tables / Creating Partitioned Tables|
As a best pratice, use the SYS.CREATE_ALL_BUCKETS('table-name') procedure to create the buckets for a partitioned table before you begin loading data.
SYS.CREATE_ALL_BUCKETS immediately creates the necessary buckets on datastores for partitioned tables. (Normally GemFire XD assigns buckets to servers lazily, as inserts are executed against the table.) Pre-allocating the buckets ensures that partitioned table data is evenly distributed in the cluster. If you do not use SYS.CREATE_ALL_BUCKETS('table-name'), data can become skewed if you rapidly load table data using concurrent processes.
If a partitioned table becomes skewed, use SYS.REBALANCE_ALL_BUCKETS to rebalance all partitioned table data in the distributed system.