Oracle isn’t normally one to expose its fetishes, but the company’s fixation on Amazon Web Services has become glaringly, disturbingly obvious. Whether at OpenWorld, in analyst briefings, or on Twitter, Oracle can’t seem to talk about anything else. Two years ago Oracle peddled the myth that “AWS is 20 years behind Oracle,” even as AWS boasts 80 times more IaaS revenue and 10 times more PaaS revenue than Oracle, according to Gartner data. More recently, Oracle chairman Larry Ellison, in a “pot calling the kettle black” moment, dissed Amazon Redshift as “inflexible and expensive.”
Even more recently, Oracle’s database chief, Andy Mendelsohn, pilloried AWS databases as underpowered for enterprise workloads. If your only ambition is to run “small, departmental databases with decent performance,” Mendelsohn crowed, AWS is adequate. But “If you want to run the biggest, baddest enterprise workloads, they can’t run on Amazon.” This is ridiculously false; AWS has had scores of enterprises on the record embracing AWS databases for their most mission-critical needs. But perhaps Oracle still desperately wants to believe it.
Indeed, Oracle needs this to be true because, as Gartner analyst Merv Adrian told me, rival databases are starting to put the hurt on Oracle, with AWS the biggest threat by far. But it’s not true.
Oracle’s database business is deteriorating
Originally the biggest threat to Oracle’s database dominance seemed to come from the NoSQL crowd, given how data has changed over the past ten years. For decades, the traditional relational database, with its assembly of data into neatly ordered rows and columns, served us well. As data volumes, variety, and velocity changed (the so-called three V’s of big data), the venerable RDBMS seemed outdated.
Perhaps it is, but that doesn’t mean enterprises can afford a rush to the exits in favor of the flexible schema that NoSQL offers. As Adrian puts it, “When someone has invested in the schema design, physical data placement, network architecture, etc. around a particular tool, that doesn’t get lifted and shifted easily, something that Gartner calls ‘entanglement.’” As nice as “rip and replace” may sound, once an enterprise has bolted dozens of systems to a particular database, replacing that database may also mean replacing some or many of those interconnected systems.
And doing that is simply too painful. “The greatest force in legacy DBMS is inertia,” Adrian concludes.
But for new applications NoSQL databases specifically and cloud databases in general are having a moment that keeps going and going and going. In 2011, the top five database vendors—Oracle, Microsoft, IBM, SAP, and Teradata—owned 91 percent of DBMS revenue. By 2016, that number was down to 86.9 percent. Although that doesn’t seem like a precipitous drop, the database market is worth roughly $34 billion. A drop of a few percentage points is a very big deal, and it involves lots of cash.
Oracle, for its part, has shed market share points every year since 2013, Adrian notes. Yes, that share is still about 40 percent, which is roughly double that of second-place Microsoft. But the difference is that Microsoft’s share has grown every year during that same period. Oh, and AWS? AWS is “roaring up the charts,” Adrian noted, while “IBM is dropping precipitously.”
No, we’re not going to see Oracle’s database revenue fall off a cliff. But that might not be because its customers remain committed to the database leader. Instead, they may simply continue to pay for stuff they don’t actually use. As a recent Rimini Street survey showed, as much as 74 percent of Oracle customers are running unsupported, with half of Oracle’s customers not sure what they’re paying for. These customers are likely paying full-fat maintenance fees for no-fat support (meaning they get no updates, fixes, or security alerts for that money).
These aren’t behaviors of companies that are committed to the Oracle value proposition. They’re just conditioned to write that check.
Except, of course, for new applications.
For new applications, NoSQL and AWS are getting the action
For those new applications, non-cloudy NoSQL is taking a significant chunk of business, Adrian underlined. Nonrelational databases like MarkLogic and MongoDB now generate $268 million in revenue each year, a number that is “growing nicely” in the mid double-digits. If you add in Hadoop vendors, that nonrelational number jumps to $1.5 billion, or 4.5 percent of the DBMS market. Nonrelational databases, in other words, have “hit escape velocity,” Adrian argues.
This, however, is not enough to strike fear into Oracle. The largest, fastest-growing of the nonrelational vendors—Cloudera—could hit 40 percent growth each year for a few years and would still take years to get to $1 billion. That’s significant, but it’s not AWS—which, again, is “roaring up the charts.”
Which is why Oracle fears AWS, and rightly so.
Ovum analyst Tony Baer describes Oracle’s cloud strategy as “Cloud 2.0.” Despite being so late to the cloud party, Oracle now wants us to believe that it can learn from the mistakes of AWS, Microsoft Azure, and others to leapfrog them all. This is complete and utter nonsense. Not only is Oracle ill-suited to actually build a next-generation cloud database, because it has no experience running cloud applications at scale (unlike Amazon, Microsoft, and Google, which have that experience baked into their DNA), but Oracle’s volume and velocity of cloud investments lag AWS by dozens of datacenters and years, not months.
Meanwhile, AWS’s database products are its fastest-growing services. Most of this database adoption is for new applications (which are growing dramatically faster than old-school, Oracle-inclined applications). But AWS CEO Andy Jassy has also announced more than 50,000 database migrations, much of them from Oracle.
Oracle is a fantastic database for yesteryear’s enterprise applications, but it is a poor fit for modern, big data applications. For these, Amazon will continue to gobble Oracle’s market share, $1 billion at a time. This will lead Oracle to fixate even more on AWS, but that fixation doesn’t seem to be fixing the problem.