Posted by Reilly Atkinson on September 19, 2001 at 21:29:47:

In Reply to: Datawarehouse performance issues posted by Patterson on September 18, 2001 at 10:34:51:

Patterson -- While I tend to agree with the comments of Messrs, there is a great deal of experience with data warehouses(DW) that transcends the recent "aha" of the business community. Mr or Dr. Inmon is hardly the father of data warehouses -- unless he somehow was alive in BC Egypt, or in BC Mexico, or even in the late 16th century with Tycho Brahe.(We are talking astronomical data here)

My colleagues and I, and many others, built data warehouse in the 1970s, for forecasting projects, and, in addition, to support building simulation models of urban growth and change, and to develop a quantitative approach to national housing policy. We never heard of Inmon.

There are and have been huge data warehouses in all the sciences.

All these DWs were built for broad but well defined purposes -- for example, multiple years of tract level census data along with census long-form detailed data, along with data on housing permits, starts and completions.

Contrary to business practice, these other DWs are defined and developed by those who will use the data. Who else knows better? The programmers are often highly educated with Masters or Doctorate degrees. Clearly the business approach is rather different, and suffers as a consequence.

For me, the most frustrating aspect of corporate DW is the generally lengthy process of getting data -- and then to find all too often that the data have all sorts of problems. So, other metrics should measure speed of getting data,
and how much work is needed to clean the extracted data. My worst case, a few years ago, was an eight month wait for data to do sales forecasting for a food products company.

If an analyst can get the data he or she needs, yesterday, today, and tomorrow, in a timely fashion then you are well ahead of the game. No matter what hard-wired queries are appropriate, the analysts must have the ability to do ad-hoc analyses.

So, for me, the primary metric is the ease with which analysts can do their job. If they have problems, then other metrics should be derived from analysis of the problems. When in doubt, be empirical. And, it really does not need to be very complicated. Because we are talking business, a financial/cost/ROI metric or two would be appropriate.

Enough.
Regards,
Reilly Atkinson

---

Follow Ups:

• Re: Datawarehouse performance issues Martyn R Jones 05:19:46 09/20/01 (2)
  • Re: Datawarehouse performance issues Martyn R Jones 05:41:23 09/20/01 (1)
    • Re: Datawarehouse performance issues Reilly Atkinson 16:52:23 09/22/01 (0)

---

Click Here to Post Follow Up in New Forums

    Knowledge Management Think Tank (New)

  

 Subject: 

Message: