Насчет избыточности имеются вполне определенные мнен

Bill Inmon’s approach to data warehousing is a holistic data management approach, not an approach for providing stovepipe solutions. In other words, a holistic data management approach recognizes the many different types of decision support requirements all organizations have, such as operational reporting, operational ad hoc querying, tactical management reporting and strategic trend analysis reporting. It also includes many different types of applications, an implementation strategy for the organization, and data standards to avoid uncontrolled redundancy. The "drawback" (although I would not call it that) is that it takes longer to build a holistic decision support environment than to build stovepipe data marts for different sets of requirements without any consideration of standardization or integration (which includes reducing redundancy) across the organization. The end effect of stovepipe solutions is that they add to the non-integrated spaghetti chart of systems (thus adding to the redundancy) that already exist in most organizations. And the large the spaghetti chart the harder it will be to manage your data as a true corporate asset. Кстати, дальше дискуссия как раз разивается в сторону, что считать избыточностью: "The single most dramatic way to affect performance in a large data warehouse is to provide a proper set of aggregate (summary) records ... in some cases speeding queries by a factor of 100 or even 1,000. No other means exist to harvest such spectacular gains." Those are Ralph Kimball's words from "Aggregate Navigation With (Almost) No Metadata" (DBMS magazine, August 1996). If the results are so spectacular - and I don't hear anyone arguing they aren't - why then is aggregation so underused? First, I believe that the answer lies in our relational database culture or folklore. We were all taught not to redundantly store what could be calculated. This, by the way, is a restriction that users of multidimensional databases have happily ignored. Second, many of us are not clear on what constitutes a good set of aggregates. One way to get more comfortable with aggregate tables, and to see how mandatory they are, is to think of them as indexes. We would not dream of implementing any significant OLTP or data warehouse database without indexes. These traditional indexes usually duplicate the information content of indexed columns, yet we don't disparage this duplication as "redundancy," because of the benefits.