Although I understand your argument and conceptually agree with it, I respectfully think that your Oracle example is accurate but do not take into considerations the de facto issue that most people in the data warehousing do not have the expertise necessary to make the changes that you allude too. Another issue that is not address in your articles is the need for high-level and predictive analytics to proactively provide guidance to business stakeholders so that they can quickly and efficiently implement changes in a rapidly changing business envirnoment where the ability to "turn around in a [yearly] quarter" is becoming crucial for companies. Specifically, Hadoop has the ability to do fast processing by creating similar (not equal) segments for purposes of processing, BUT NOT IMPORTANT (my apologies for using caps but I wanted to emphasize the point), those segments can also be used for predictive modeling and high-level analytic purposes. The key is that segmentation for analytic purposes is a very time consuming effort and Hadoop greatly accelerates and automates this process (although it needs implentation knowledge to modify those segments). As more industry specific sensors are developed by Intel and other companies the fast processing of new segments,with this additional data from sensors for analytics purposes, will become more important than ever. Lastly, I am not advocating that data warehousing is obsolete. I am advocating that big data and advanced automated analytics need to be considered in any data warehousing architecture, so that a system can quickly grow to consider new business conditions and data sources, including text and sensor data.