Hi Alexandre-san
Thank you for summarizing the strategy. I agreed your design principle.
Per your phase, I feel we are on the phase 0 or 1, we can first commit the basic architecture.
> Note:
> I had different expectations, the first query uses two columns and filters over a 1-year period, the second query
usesonly one column and filters over.
> My first expectation was that if only one column is used there is less data being accessed.
> In my experience most queries will care about a timestamp column, it would be nice to be able to cluster the data by
certaincolumns (in my case the created) column
> Maybe even clustering by multiple columns independently mediated by CRID permutations. If you think that this is easy
enoughwe could insert this before phase 3.
The number of rows accessed by the first query and the number of
rows in the second query differ significantly.
This isn't only because one query uses a single column. Therefore,
I understood the advice to be that even when using VCI, it would be
beneficial to enable clustering on another column,
such as the created column, right?
Best regards,
Aya Iwata
