Time if day in the time dimension

From: <bmeltz_at_mediaone.net>
Date: Mon, 21 Aug 2000 13:27:05 GMT
Message-ID: <8nraml$a5d$1_at_nnrp1.deja.com>


This did not get much response in the comp.databases.olap forum, so I thought I would try it here. Sorry for any duplication for folks who subscribe to both groups.

With my new organization, our time dimension will likely go to the hour
(if not minute) level of granularity. This is new to me, since
previously, the finest grain of time has always been the day. By the way, the warehouse will be dimensional in a relational database (your basic ROLAP).

I was just going to (and may likely still) add a level or two to your basic time dimension, minute and hour. They are just different levels of detail, after all. Also, since the data will span multiple time zones, my first inclination was two time keys, one GMT and one local
(actually, it was my second inclination, the first was to have each

time record combining GMT and local variations with a time zone attribute for each time zone, but this was just a fleeting thought that gave me a headache and was quickly discarded).

Looking for new ideas and approaches (and maybe even validation of my initial thoughts), I checked the usual suspects. One stop was Ralph Kimball's article list where I ran across 'Think Globally, Act Locally'
( http://www.intelligententerprise.com/9812/warehouse.shtml ). In it,
Ralph does articulate the two time key idea, but he suggests that time of day be expressed in the fact table. This seems odd to me, since from a basic conceptual level, time is a 'by', not a 'what'. Also, time of day is, as previously stated, simply a lower level in the overall time hierarchy.

Does anyone have any comment on this article and area in general? What have folks used in the past and with what degrees of success?

Thanks,
Barry

Sent via Deja.com http://www.deja.com/
Before you buy. Received on Mon Aug 21 2000 - 15:27:05 CEST

Original text of this message