Xpand breaks each representation (primary key + table or other index) into smaller, more manageable segments called “slices”, each of which is assigned a portion of the representation’s rows. There are multiple copies of each slice, called replicas.

• Slices are distributed throughout the cluster to facilitate evenly distributed query processing.
• The number of slices specified for the table itself applies to the table’s data as well as to its primary key.
• A different number of slices may be specified for each representation.
• The number of slices for a given representation should never be less than the number of nodes in the cluster. (Tables distributed to ALLNODES are an exception.)
• When a slice becomes too large and exceeds rebalancer_split_threshold_kb, the Rebalancer will split the slice into new slices and distribute the original slice's rows among them. The larger a slice becomes, the more expensive it is to move or copy it across the system. 
  • The maximum slice size for an individual table or index can be specified by including split_threshold_kb in the representation's DDL.
• A slice must physically fit in its entirety on the storage device to which it is assigned. One slice may not span multiple devices. 

To modify the number of slices for an existing table or index, follow this syntax:

ALTER TABLE tbl_name   [SLICES = n]  [ , INDEX index_name  [SLICES = n] [split_threshold_kb = n]]

Global Variable Definitions                 

The following global variables impact Xpand slicing.

Best Practices

The default slice size of 8 GB is optimal for most use cases. If a very large table results in more slices than available cores, performance might be impacted and increasing the max slice size may be recommended.

Tables and indexes should have a minimum number of slices equal to the number of nodes, with ALLNODES tables being an exception. Use the following query to identify tables that contain fewer slices than the current number of nodes:

sql> SELECT   fd.name  `Database`,
              f.name   `Table`,
              Count(*) Slices
     FROM     system.slices
         JOIN system.representations fp USING (representation)
         JOIN system.relations f
                ON ( relation = `table` )
         JOIN system.DATABASES fd USING (db)
     GROUP BY `Database`,
              `Table`,
              fp.name
     HAVING   fd.name NOT IN ( 'system', 'clustrix_statd','clustrix_ui','_replication' )
        AND   slices < (SELECT Count(*)
                        FROM   system.membership
                        WHERE  status = 'quorum') 
        AND  (fp.name LIKE '%__PRIMARY%' OR fp.name LIKE '__base%')
     ORDER BY `Database`, `Table`, Slices;

Pre-slicing Tables 

During normal operation, relations are resliced on demand, however it can be advantageous to pre-slice tables for which large data growth is anticipated. Creating or altering a representation to have a slice count commensurate with the expected size will allow the cluster to add data to the representation at maximum speed as slice splitting will be unnecessary. For additional information, see Loading Data onto Xpand.

Use the following equation to determine the optimal number of slices for a table: (expected table size + 10%) / rebalancer_split_threshold_kb)

Managing Slice Splitting

The Rebalancer automatically splits a table slice or index slice when it reaches the threshold set by the global rebalancer_split_threshold_kb. However, some tables might benefit from more slices (for increased parallelism) or fewer slices (to reduce overhead of slices on the system).