Re: Sales db design...
Date: Thu, 23 Feb 2006 16:56:02 -0500
Message-ID: <dtlb1o$5k08$1_at_news3.infoave.net>
Volker Hetzer wrote:
> A few billions of rows aren't supposed to be a problem for a real database.
I'm beginning to understand that after reading more on the efficiency of indexing. I was surprised to learn that a "real" database (and I'd think that MS SQL Server should qualify) only had to divide a million rows 20 times at most to find any given row, which is immeasurably faster than plowing straight across a million rows, as I'd imagined would be necessary. This would make digging data out of massive tables far faster than I ever imagined possible.
> What's often done (if the underlaying database permits it)
> is to partition the table so that you have a small one with the open
> invoices
> and a large partition for archival purposes. Those partitions are then
> placed
> on different discs.
> That way you can do fast oltp on one partition, and the warehousing and
> reports
> on both. Or rather, that's what the database does. You just run queries.
Excellent information. Thanks so much! :-) Received on Thu Feb 23 2006 - 22:56:02 CET