@Olaf.Stephan asked:
In order to avoid shipping costs for non-collocated joins are replicated MQTs required.
What's about replicated MQTs in a BLU DPF environment if the base table a column-organized table?
@pbird replied:
I would suggest to the customer to treat BLU in MPP as a new technology and evaluate its native performance in MPP first before introducing additional objects to aid in performance. The "raw" performance may be sufficient for their performance needs without the need to introduce replicated MQT tables. In other words, I am saying to not assume that replicated tables are still needed with distributed column organized tables just because they were needed for row-organized tables.
We did introduce support for replicated MQTs on top of column-organized base tables with certain limitations: they must be defined as REFRESH DEFERRED and only the NONINCREMENTAL option of REFRESH TABLE is supported. In other words, you need to use the REFRESH statement to update the data and you have to replace the whole table when you do this. Also, another consideration is that the dictionary for a replicated table is only created once the threshold for automatic dictionary creation is reached (and will only affect future rows).
Please keep in mind that this is an early alpha build and so it does not have the full performance we anticipate in the final release.
If you have any more questions please feel free to reply to this topic.