Please start any new threads on our new
site at https://forums.sqlteam.com. We've got lots of great SQL Server
experts to answer whatever question you can come up with.
| Author |
Topic |
|
coolerbob
Aged Yak Warrior
841 Posts |
 Posted - 2007-10-19 : 05:40:57
|
| Has anyone here been involved in a successful commercial Data Warehouse project?If so, what was your role? How big was the project? Did you do it all or was their one guy data modelling, one for ETL, one for Client-side, one project managing, etc...? What tools did you use? Was it in SQL2000? 2005? Did you use DTS/SSIS for your ETL or Sprocs? Did you use SSAS? Were you writing lots of MDX to meet client-side needs? Or did you use a completely different toolset for all or some of this?Did you use the Kimball approach? Or were you in the other camp?Was your latency nightly or did you have some type of "real-time" capability?I would prefer to hear from people who've actually been involved in successful projects (or failed projects if you would like to share with us why you think they failed). I'm not too interested in hearing from people that have never done it but have a theory as to how it should work (like me). Many Thanks,Coolerbob(Or is there some other website forum I should be asking this question on? (like mdxteam, olapteam, datawarehouseteam or something) |
|
|
Kristen
Test
22859 Posts |
Posted - 2007-10-19 : 05:43:29
|
"I'm not too interested in hearing from people that have never done it but have a theory as to how it should work (like me). "I'll lurk on this one ...  |
 |
|
|
coolerbob
Aged Yak Warrior
841 Posts |
Posted - 2007-10-23 : 08:42:56
|
| Gutted, I know people are using these tools out there. But there's just no-one out there who knows about this thread or cares to comment on it. That's a real shame I think. I don't quite know why the "culture" around Warehousing is so different to that of RDMS. |
|
|
|
rockmoose
SQL Natt Alfen
3279 Posts |
Posted - 2007-10-23 : 11:15:19
|
Top 1 priority for a successful DW project is data quality.Followed by accessibility.I have been involved in a couple, SQL7 -> migrated to 2000, SQL2000 and SQL2005.Crystal Reports(SQL7), Analysis Services + ProClarity(2000), Reporting Services(2005).Loading method has been mainly sprocs.For the 2000 project we started to use DTS, but as the project proceeded we used more&more BULK INSERT,bcp&sprocs, and many DTS packages were migrated to sproc format.Approach has been to have a normalized data store, then build dimensions or aggregations from the data store.Philosophy has been that all the denormalized/aggregated tables can be dropped and recreated if need be.The roles for the projects have been "User Requirement Gathering", "Data modeler", ETL, "Cube Designer", "Report Designer".There has not been any real time requirements, mostly nightly or hourly loads. The hourly ones in order to make several small batches i.o.f one big.Things that fail a project;Slow reports, Bad/Wrong dataHeadaches;Aggregated or denormalized tables that need to be recalculatedUser requirements  Success;Easy to maintain the load processEasy to extract data (make new reports)Optimized (fast / efficient)rockmoose |
|
|
|
|
|
|
|
|
Success Stories