Wednesday, October 11, 2006

Existing Oracle Customers Get More Confused: When should they use Sunopsis, Oracle Warehouse Builder, Informatica, or Ascential?

Oracle announced that it would be acquiring Sunopsis, a data integration vendor with solid products. Well, you can't fault Oracle for giving their customers choices! Existing Oracle customers already had THREE SEPARATE ETL tools that they had to use with Oracle products even BEFORE this acquisition:

  • existing ETL technologies of Oracle Warehouse Builder which they just made a big deal about how they had enhanced significantly in their latest release. Guess it couldn't have been THAT good.
  • existing OEM relationship they have with Informatica that they inherited from Siebel Analytics.
  • existing OEM relationship they have with Ascential that they inherited from PeopleSoft.

And now, Sunopsis. So for Oracle customers who owned PeopleSoft HCM, and Siebel CRM, and Oracle ERP, what should they use for their data integration needs? Contrast this with SAP, who has pursued an organic growth strategy and has one set of data integration tools for our applications across the entire enterprise. NetWeaver Business Intelligence includes sophisticated ETL capabilities that support analytic applications in every major horizontal: CRM, HCM, FIN, SCM, SRM, and PLM. With over 12,000 installations, and referenceable customers in every vertical we must be doing something right.

So, thank you, Oracle for continuing to give customers a reason to take the Safe Passage to SAP. We look forward to you acquiring even more companies that overlap directly with your existing assets and continue to confuse your installed base.

Does SAP really provide a Data Warehouse that requires no tuning or aggregates?

Mark Rittman, one of the sharpest minds in the Analytics and Business Intelligence space, had this to say about my previous post on SAP's Business Intelligence Accelerator: "Neshan Bardolliwalla (sic) talks about SAP's new in-memory BI accelerator that requires no tuning or need to build aggregates (as my son says, "yeah...right.")" It is hard to believe for those of us who have been in this industry for a while, but it is absolutely true. Here is an excerpt from my response to Mark which I posted as a comment on his website.

While I am not at liberty to disclose the detailed internals of the product, consider that a 64-bit computer can address 16 exibytes of memory. So, if you were take your ENTIRE data warehouse at its MOST granular level, without ANY aggregates whatsoever, put it in main memory, index that data, and then use clever on-the-fly search techniques to locate that data, you could most certainly eliminate the need for aggregates or any tuning whatsoever. I’ve seen it with my own eyes, and yes, I didn’t believe it when I heard it either.

With today’s technology and Moore’s law continuing to operate, the need for disk-based database systems will continue to become less relevant as memory-based systems become cheaper to exploit and algorithms continue to be developed that allow one to search the addressable space of memory that a 64 bit processor is capable of addressing.

To put this into perspective, to get the same performance that BIA provides from an RDBMS vendor would cost about TEN TIMES as much.