Tables in the Object File System have limitations and different performance characteristics compared to tables in Block Storage on the primary cluster.
The following lists the limitations:
- No MERGE statement
- No primary index
- No secondary indexes
- No multi-table join indexes
- No load utilities such as FastLoad, MultiLoad, TPump, TPT Load
The following data types are supported only on the primary cluster:
- DATASET - AVRO
- DATASET - CSV
- XML
- JSON - UNICODE and LATIN
- UDT Structured/Distinct/Variant
The following data types are not supported:
- PERIOD(TIME [(n)])
- PERIOD(TIMESTAMP [(n)])
- MBR (minimum bounding rectangle)
- MBB (minimum bounding box)
See VantageCloud Lake on Microsoft Azure for Azure limitations.
For the Object File System, the following have default values that you cannot change:
- BLOCKCOMPRESSION
- CHECKSUM
- DATABLOCKSIZE
- FREESPACE
- MERGEBLOCKRATIO
INSERT-SELECT is the recommended way to load data into an Object File System table.
For Object File System, the data is stored in objects. The largest object size supported is 16 MB. Because there is a high amount of latency to read and write the object, it’s best if the INSERT-SELECT generates at least 16 MB of data per AMP involved in the INSERT.
Processing of LOBs is now supported on compute clusters. However, the transfer of LOBs between the compute and primary clusters not supported yet.