Building a Best-Fit Data Warehouse: Why Understanding Physical Database Structure Matters

by Hassib K
The article went on at length discussing the various uses of data warehouses and how they should be designed physically so that they can complete all necessary tasks quickly and efficiently.  There are 3 types of business intelligence that it went over and those included, operational BI, analytics, and management BI.  It mentioned that there are other scenarios as well, but it did not go into those.  O’Brien states that the data warehouse will perform at its peak when the logical data models, BI workload, and optimal physical database design are matched up; so there is a RIGHT way to physically design the database for each situation.  He discussed the difference between SMP, Clusters, MPP, and Grids.  BI workloads can be viewed at from a perspective of how much demand they place on the physical architecture.  The characteristics to look at include:  data volume, number of users, types of queries, frequency and timing of access, and latency requirements.  Some business intelligence activities will place more demand than others.  Operational BI for example requires time-sensitive real-time information usually and this will place a heavy demand on the physical architecture if many users are requesting the same information at the same time.  Management BI usually requires the “push-button” access to various reports and information and Historical Analytics usually requires incredibly large amounts of data that can possibly contain information from a client over the past 10 or 20 years.

It goes on to list various possible solutions for each BI group and their physical requirements, etc.  It is a very thorough article that contains alot of great information regarding physical database design.  O’Brien concludes with, “don’t design for the probable – design for the possible. ” (O’Brien 62)

I agree with most of the article.  I would actually have no reason not to agree.  I don’t know as much about this stuff as the author does, so i’m sure he knows what he’s talking about.  Also, all of his arguments are backed up by good logic so it makes sense.  I definitely understand the importance of physical database design even more now.  Designing your database for what could potentially come up is a very smart approach, however, it would probably cost more money and i’m sure thats why it’s not always designed that way.  However, even though it would potentially cost more money, I think the expense would be worth it and would pay off in the long run because having access to good information is crucial to any business so making sure that it is available quickly and reliably when you need it, is worth a few extra dollars.  There would definitely have to be some sort of cost-benefit analysis done before committing to any physical database design.

 

Sources:

OBrien, J. (2008, Building a best-fit data warehouse: Why understanding physical database structures matters. Business Intelligence Journal, 13(1), 51-62. Retrieved from http://search.proquest.com/docview/222595249?accountid=10357