Monthly Archives: September 2010

I’ve recently been involved in a rousing exchange out on Linked-In with one of Oracle’s employees – on the Exadata machine and how it compares to Netezza. The conversation simply had to stop when the Oracle employee implied that I might be biased. While I am a big fan of Netezza (and unashamedly so) it continually rests on Netezza’s ability to deliver, not because I am an employee (because I am not). I otherwise receive no financial consideration – whatsoever – for my presence here.

So the following dissertation is for Enzees on the subject, because much of the impetus for Exadata’s very existence is Oracle’s marketplace response to Netezza’s products, rather than truly offering up something original.

Objectivity on the matter requires us to spin back in time to 2007 when Oracle’s senior management publicly recognized the appliances (Netezza in particular) as a threat to their market share. I was in attendance at the 2007 Netezza Conference in Boston when the head of Business Objects global strategy group noted that the Netezza appliance had officially relegated many “mainstay” data warehouse technologies to “secondhand” status for data warehousing. The first on his list was Oracle. When this hit the big-screen, it elicited a standing ovation, because the vast majority of those folks had moved away from Oracle to Netezza, and the Oracle management had finally recognized it as a serious threat – one that then (and continually) affected Oracle’s public image outside of some rag-tag gaggle of naysaying defectors. The standing-O was an affirmation that all of them had done the right thing by moving to Netezza.

It was on.

So some of us watched with anticipation as Oracle, the industry giant, made a move. Some thought they might actually tender an offer to Netezza (if you can’t beat ‘em, buy ‘em). Then the rumors arose that Oracle was building its own mousetrap, partnering with other power-hitters to make it happen, and with all of Oracle’s experience and exposure to the data warehouse domain, and its laundry list of evident shortcomings that its users had streamed its way nonstop for over a decade. Surely, surely Oracle would be able to circumscribe the problem and offer a solution. Anticipation was high.

But nobody anticipated that Oracle would miss the mark so badly. Not just once, but twice. Putting the transactional Oracle engine on top of “parallel” hardware could not – and has not – solved the problem. In fact, it’s almost as if (and I’m speculating here) that Oracle’s internal discussions about the solution had resolved the wrong problem completely. On the one hand, Oracle had the opportunity to create a data warehousing engine that would resolve the banes and shortcomings that had earned it a “secondhand” status for RAC. A logical and reasonable question would have been “why are we secondhand?” But nobody at Oracle was asking that question. Were they?

Oddly, Oracle still maintains the debunked notion that performance is in the software, because Oracle is a software company. Alas, performance is not found in software. And that is the summary of how Oracle missed the mark. (Okay, argue with me that performance is in the software, and I will ask you to run RAC on a laptop, and see how well that goes).

The Exadata machine looks for all the world like a late-1990′s “system-slam”. That is – take some specially configured commodity hardware and strap some commodity software to it, and call it a day. Recall the old days when they would roll up their sleeves and do a TPC-D test on a carefully tuned, optimized commodity hardware configuration as a one-hit wonder for the sake of scoring high on the test? This looks (again speculation) like someone had the epiphany to put all that specially configured hardware into a machine to bypass the need to purchase separate assets. This was close, but then they spoiled the ride. Adding RAC on top of it. Whoosh is the sound of the fly-by as the missed opportunity rushes past.

This solution still carries the baggage of being an integrated solution rather than a purpose-built solution. Its evident purpose was to compete with Netezza, to protect Oracle’s existing user base from bolting, but not to actually solve data warehouse problems (because it didn’t do that). In short, Oracle’s priority seemed to be targeted towards stemming the hemmorage of lost market share, not provide an innovative technical solution. Time to market was critical for this to take place, so there wasn’t time to actually put some thought into a real data warehouse engine.

Oracle evidently embraced the RAC engine as the core of the solution with

(a) the fear of rolling out a brand new data-warehouse-centric engine or

(b) use the existing Oracle engine because it is mature and fully debugged.

Of course, going with (a) assumes that Oracle has a cadre of engineers that understand data warehousing and can build this sort of engine. The Exadata offering is a clear indication that their engineers either don’t understand warehousing, or don’t have a voice in the architectural decisions. Of course, going with (b) makes the rather bold assumption that potential data warehouse users actually want and like the Oracle engine for data warehousing, when the industry in general has bolted for Netezza precisely because they don’t like the Oracle engine for data warehousing. Someone at Oracle missed that memo, too.

So we are left with

(a) Oracle’s own data warehousing engineers are voices in the wilderness, don’t have the necessary experience, or they don’t exist at all,

(b) someone dropped the ball in the solution requirements analysis, or there simply was no analysis at all – and finally

(c) Oracle really hasn’t been listening all these years to all those complaints about its shortcomings in the data warehousing world, because Exadata sends this message loud and clear.

Pernicious hanging questions include: Why would management deliberately embrace the secondhand engine unless it was unaware that it is secondhand, or it is in denial of its secondhand status? Would the outcome have been different had Oracle recognized that the solution is found in the hardware, with a carefully, specifically constructed (purpose-built) software engine that embraces this? Is recognizing the critical role of hardware
even possible for a software company? Can Dudley Do-Right save Marigold from the evil villian before the commercial break?

The answer to these questions are: It doesn’t matter. The outcome is the same. By publishing Exadata as its very best shot at the solution, Oracle has publicly stated, in no uncertain terms, “we don’t get it.”

So the moniker of “secondhand” technology remains firmly affixed to Oracle’s data warehousing lapel. At this point the data warehousing marketplace has made-the-call, something akin to calling time-of-death for the patient,

because the Exadata offering – so far – remains SOA (Secondhand On Arrival).