Possibilities of Modeling

Help
1

1. Is there any solution at the level of logical modeling that allows automatic generation of MemSQL code?

2. Are line segment metadata generated and loaded into leaves or aggregators nodes?

3. How to query data and metadata of nodes specific? I want to see how the distribution was to understand the consequences of each change.

4. In terms of DDL, from what I observed, the possibilities of modeling variation would be:

  • Rowstores vs. Columnstore
  • Fragmented vs. Reference Tables
  • Shard key vs. Random fragmentation (keyless sharding)
  • Less amount vs. higher in columnstore_segment_rows
  • Is there anything else you would need to observe at this stage?
2

The best I can recommend today for 1 is to use a modeling tool that supports MySQL and then modify the output as needed.

Not sure what you mean by 2.

Regarding 3, logical metadata is the same across nodes. Physical metadata may be different. E.g. information_schema.columnar_segments has per-segment information about columnstore segments, including the partition where they are placed. See also IND_CS_PARTITION_ROW_SEGMENT_GROUP, DISTRIBUTED_PARTITIONS, SHOW PARTITIONS, as well as TABLE_STATISTICS for other use physical schema information.