• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 1568
  • Last Modified:

Wide fact table vs. multiple fact tables

I am currently developing a data warehouse for a telecom company.
We collect several different measures (100+) from the network elements on an hourly basis.
I'm wondering what the optimal data warehouse design would be:
1) Since all measures bear the same granularity (1 hour), and originate from similar network elements, it would make sense to put them all in the same fact table. I would therefore end up with a fact table composed of about 5 dimensions and 100+ fact fields. Does a 100+ field table sound like an acceptable design?
2) I could create several fact tables with each about 5 dimensions and each only a subset of the 100+ facts, grouped in a way likely to be queried by the end users. I would therefore have several fact tables with maximum 25 fact fields (which seems more reasonable?), but I would loose some flexibility in the sense that users could no longer query the data in ways that were unplanned or unexpected.
I would greatly appreciate any advice or comment or field experience with the 2 options proposed (or a 3rd alternative...).
0
optima-sc
Asked:
optima-sc
  • 2
  • 2
1 Solution
 
frankyteeCommented:
would any user need to regularly access the 100 fields in a query?
fact tables should only contain related measures so i think option 2) is the way to go. group your measures by category into several fact tables.
just ensure that your fact tables can be joined together using a common id field etc or a "bridge"/mapping table
0
 
optima-scAuthor Commented:
Well, I would not say that any user would need to 'regularly' access any of the 100 fields.
However, it is quite possible that, at certain points in time, they will need this kind of flexibility for troubleshooting a particular network element or do root cause analysis.
I guess the idea of grouping the facts by likely categories sounds good, however, isn't that potentially going to lead to very slow (and heavy) cross-table queries whenever more flexibility or unexpected analysis is needed from the end user?
0
 
frankyteeCommented:
if you index your tables appropriately then you should get reasonable/good performance. and the only time you need to join the tables is when the user needs measures from both multiple fact tables which should be "infrequent" if the measures are not related. most data warehouse that i have seen have multiple fact tables not one huge one with 100 measures.
but dont take my opinion as gospel, create and test the 2 different models to see which one is better for you.
0
 
optima-scAuthor Commented:
Well, after re-discussing this topic thoroughly with my client, it appears that the 'unexpected' queries might not be so infrequent. Since the wide fact table also has the advantage of simpler applications no matter what the end user choice is, we decided to go for that solution.
Thank you for taking the time to give me your advice, anyway. I'll award you the points.
0

Featured Post

Industry Leaders: We Want Your Opinion!

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

  • 2
  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now