Buckets
For any cube that uses more than 512,000 records (by default), the engine organizes the cache into buckets. Each bucket corresponds to a large number of contiguous records in the fact table, as shown in the following figure:
The final bucket (or partial bucket) is the active bucket and is not represented in the cell cache.
By default, the fact table contains records in the same order as the source table. You can specify Initial build order for the cube to control the order in which the engine examines the source table records when it performs a full build of the cube; see Other Cube Options.
When you update a cube by synchronizing or rebuilding it, or when you explicitly invoke %SetCubeDSTime() after a manual update, the engine discards parts of the cache as appropriate. More specifically, the engine invalidates any buckets that use records from the affected part or parts of the fact table. Other buckets are left alone. When it executes a query, the engine uses cached data only for the valid buckets. For records that do not have valid cached results, the engine uses the bitmap indexes and recomputes the needed intermediate values. As the last phase of query execution, the engine consolidates the results. Thus the engine can provide results that come from a combination of cached data and new or changed data. Also, because some of the engine work can be split by bucket, the engine can (and does) perform some processing in parallel.
Default Bucket Size
By default, a bucket is 512,000 records. The bucket size is controlled by the bucketSize option, which expresses the bucket size as an integer number of groups of records, where a group is 64,000 contiguous records. The default bucketSize is 8, so that the default bucket is 8 x 64,000 records or 512,000 records. For information on bucketSize, see <cube>.
