I was watching the news today and a commercial caught my eye with some nifty graphics representing a mainframe server and a cluster of 1 RU (Rack Unit measuring 1.75" high in a 19" mounting rack) pushing a stream of data. It was an advertisement for Oracle grid technology that ended with the slogan "Runs faster. Costs less. And never breaks." While I have nothing against Oracle since they are among the pioneers in the database business and their technology successfully runs some of the largest databases in the world, I do have a problem with any vendor blatantly overstating their technology. It reminds me of a few years ago when Oracle CEO Larry Ellison announced that Oracle was "unbreakable" and soon after someone announced that they found a flaw in Oracle databases. Of course, that minor little detail didn't stop Oracle from using "unbreakable" in their ensuing ad campaign. What surprises me is that no one demanded their money back from Oracle when their Oracle database did break. With these three new advertising claims, let's see how well they stand up under close scrutiny.
Oracle's first claim is that grid technology "runs faster." Before we examine this claim, we need to first clarify that Oracle's description of a grid is more appropriately described as a cluster. Oracle is not talking about a mega global grid spread across the Internet which has some serious fundamental issues that makes it extremely limited in the types of applications that can be run. Oracle's current grid technology is, in fact, the successor of Oracle RAC (Real Application Cluster). Now, let's look at the merits of the "runs faster" claim. While benchmarks aren't perfect, they are the only standardized method that can measure dissimilar products in as fair a way as possible. One of the most prominent standards in the database market is the TPC-C benchmark, which provides a generic measurement of OLTP (On Line Transaction Performance). Oracle did take second place honors behind an IBM DB2 implementation, but ironically it wasn't a clustered implementation of Oracle's 10g database. The clustered version of Oracle 10g came in third place on an HP/Linux platform. While that's nothing to be ashamed of, it certainly doesn't take top honors.
The moral of the story here is that 9 out of the top 10 TPC-C performers were not clustered implementations. The fundamental reason for this phenomenon is that you can only take clustering so far because of the overhead involved in clustering database applications even while using high-speed gigabit interconnects between the nodes of the cluster. This is why super grids are limited to specialty applications like SETI that have minimal I/O requirements and only need raw processing power. Technically speaking, there is nothing wrong with the statement of "runs faster" because it isn't clarified what Oracle grid technology is faster than. The bottom line is that buyers need to be aware of what "grid" or "clustering" can and cannot do and that it isn't what it's hyped up to be.
Oracle's second claim is that Oracle grid technology "costs less." Again, this begs the question: Costs less than what? Any CIO would probably chuckle if you use the words "costs less" and "Oracle" in the same sentence. Again, I have no problem with Oracle's technology or with anyone making money or even lots of it, but it did raise my eyebrows when I heard the words "costs less" in that advertisement. To quantifiably define the price/performance ratio of database technologies, TPC-C also provides the top price/performance leaders. Oracle only managed to get seventh place far behind Microsoft SQL-based solutions and IBM DB2 solutions. Does that merit a "costs less" slogan for Oracle grid technology? Probably not by most reasonable standards, but Oracle is probably not alone in exaggerating their value proposition. It's just that the TPC-C price/performance numbers don't agree with Oracle.
Oracle's last claim of "never breaks" is probably the most controversial. "Never breaks" essentially means the same thing to me as "unbreakable" and is equally dubious. As an engineer who works on the front lines in IT, I know first hand that nothing is unbreakable -- especially at the complexity level of a full-fledged database application. While I'm not saying that Oracle's technology is any less or any more reliable than anyone else's database, I was offended by the use of the term "unbreakable" back then and I'm just as offended by the use of the phrase "never breaks." If there are any doubts, just ask the poor database administrator who has to stay up late at night and over the weekends trouble-shooting database issues or applying the mega-quarterly patches that Oracle issues. Again, Oracle isn't the only one with bugs and patches to deal with, but Oracle is the only one claiming "unbreakable" and "never breaks" and it's shocking that they can get away with saying it.