it has come to my attention that the “industry” believes (incorrectly) that getting data out of a data vault is difficult, if not impossible. this is so far off the truth it isn’t even funny. individuals can find the right information in the book in chapter 6, and in depth discussions in the data vault 2.0 certification class (cdvp2).
this is a short entry (really short) that discusses the simplicity of getting data out of the raw data vault, and virtualizing your downstream layers.
1) you can build business vault structures – consistent with master data, or pre-aggregated data, pre-computed data, cleansed data, etc… read about the business data vault for more information. the bdv (when built properly) reduces the total amount of tables and data that the queries need to use to retrieve their data sets.
2) best option: read about point-in-time and bridge tables. eat, breathe, sleep, and dream about these. these are the fundamental methods for making sub-second queries fast, they are the way to virtualize your downstream mart objects (most of them).
it seems pit and bridge tables are often forgotten or overlooked, unfortunately in data vault 1.0 they aren’t even really defined very well. data vault 2.0 relies on these objects. when built properly, the benefits include:
- table elimination
- row elimination (pre-join)
- equi-join conditions (to raw vault and business vault tables)
- co-location layout (mpp solution)
- time buffering
- real-time alerting
- star-join optimization
- partitioning opportunities
when built properly they can and often do contain:
- indexable calculated fields (no function based indexing needed)
- replicated (some specific / sparse) satellite fields for like clauses (for example)
and more. unfortunately i don’t have time to go in to the lecture of how and why they work – you will just have to read the book, or go to training…
http://scalefree.com (usa and europe)
http://performanceg2.com (usa and canada)
http://certussolutions.com (australia and new zealand)
hope this helps,
(c) copyright 2017 dan linstedt, all rights reserved.