I read a whitepaper recently from Teradata entitled “Exadata is still Oracle” http://exadata.teradatanow.com. This is no revelation, but it is does send an important message to potential buyers about being sure and evaluating other choices before buying one of Oracle’s proposed “magical” machines.

It does take into account the persistent question when considering Oracle - is the shared everything architecture the right choice for business intelligence and analytics? Recently Oracle has been highlighting wins of Exadata against TD and that is only because they can’t cite any technical wins against NZ. Teradata understood this challenge and realized that MPP was the key component to providing any kind of performance for analytics against large volumes of data. However in order to produce an appliance, the complexity of MPP needed to be hidden and, more importantly, the human intervention needed to be nearly eliminated to earn the name “appliance”. Teradata got it half right, but still requires very skilled personnel to implement and support the operational aspects of a Teradata implementation. Teradata were innovators for realizing that accessing and analyzing collected transactional and sub-transactional data would require a different approach that focused on parallelization and data scanning, but they have been unable to eliminate the complexity and peripheral costs associated with implementing and managing this sort of solution.

Now Oracle, with all its dominance and shared everything approach, has architected itself into an operational database corner. Starting with the wrong architecture has meant adding all kinds of objects and configuration tasks that created a monster of complexity. Is it even reasonable to think that one database can be multi-purposed without extreme complexity? If Oracle is serious and legitimately wants to compete in the business intelligence and analytics space they had better use a different product because one product cannot be optimal for both OLTP and ad-hoc, iterative analytics.

The innovation by Netezza revolutionized the method for I/O, which has always been the bottleneck of using a traditional data warehouse infrastructure. The use of an FPGA acting on the data as the disk is spinning has removed the latency created by moving data in blocks from storage - to server - to database memory before being able to respond. When you think about what makes Netezza superior to all other data warehouse systems, in the end it is simply a matter of moving the problem to the data and not moving the data to the problem.

All of that being said, I will conclude by saying that Oracle just does not get it! Teradata knew that MPP was critical. Netezza innovated even further with streaming I/O. When dealing with terabytes and now petabytes of data when it comes to moving data - less is always more. If you put a Netezza appliance on your data center floor to compare against either of these traditional data warehouse systems it will win hands-down. I encourage you to put aside the politics and make the right decision to provide your business with superior capabilities that truly do change the game of business intelligence.

- First Liberated