Phil Factor (7/26/2010)
I'm not clear to me whether you are suggesting that this approach (restore followed by a rename) is a good way of updating the data marts from the hub. If so, then it would need some more explanation, as I'm still wondering whether you are suggesting it over more elegant solutions just because of huge volumes of data to be transferred. Apologies if I've misunderstood.
Unfortunately, the last part of the article is missed and this is created misunderstanding. Even for me is difficult to understand myself. I did sent email to the publisher but has no response. You are right. For the small data warehouses (up to 10G) it is one of the ways to replace data and make very short period of data inavalability. We preparing data in separate data warehouse, loading it to every server from HUB and then switching within 1 second or even less. We have 3-4 data warehouses and doing this for at least 2 years during low nightly traffic. Our data warehouses supporting many applications. What we got with this process 1. data updated on all servers practically at the same time. 2. Data availability 24*7 3. Ability load and verify data first before it get used 4. Data integrity at any time - there is no time when data inconsistent because of load process.