Archives of the TeradataForum
Message Posted: Tue, 03 Jul 2001 @ 14:38:04 GMT
| Subj: | | Re: Data Refresh in Teradata |
| |
| From: | | James Judge |
Depends on your business need. I assume there has been a business problem/opportunity that this 1st iteration of data warehousing is
addressing. The business requirements should drive the technical solution; based on the costs and benefits of multiple approaches. In the
case of data maintenance, (and since you said refresh are you completely refreshing the warehouse each time?), the availability of source
data and the update cycle of that source(s) data will be a determination obviously.
If this initial implementation would address, say, "fraud detection" then data currency would be critical. Other the other hand if you
are trying to profile customer/prospects then if your marketing campaign can't act/react more than on a weekly/monthly/quarterly basis that
should influence the update cycle.
But as you can see, as your warehouse applications increase in number the data refresh requirements will be different for different data.
The important part to this would be the availability of the "metadata" about the refresh cycles to the warehouse users. As more information
gets incorporated into your warehouse schema the differences in refresh cycles can cause reporting problems if the refresh cycles are not
known and data is joined together; which should be a goal of any warehouse.
Taking a wild guess in that your current data is rated call detail (or some variation of that) from a billing provider who rates the raw
call detail, I would say this data arrives once a day so your refresh cycle cannot be shorter than that. But if you do completely reload
the table(s) then I would think daily would be prohibitive, obviously depends on current performance of volume/configuration. But to get to
a proactive warehouse that can profile "high-value" customers and detect changes in their habits you would need to start planning for a much
more active data update cycle.
| 
