Tuesday, April 17, 2012

Oracle's Mendelsohn Talks Big Data Trash

It's truly fascinating to observe how behemoth incumbents behave in disruptive markets. 

Like all disruptive markets, they don't take hold immediately.  They begin as Noise, and with help from industry analysts who cover them, often experience a "hype cycle" of overheated expectations.

It is the potential to wreak havoc on the incumbents of established, highly profitable and mature markets that creates the frothiness.  Since the 1970s, technology incumbents have behaved according to the stage of market adoption.  And acted in predictable ways to stave off competitive threats from new, disruptive innovators.  Oracle is certainly no exception, particularly in this era of Big Data. 

Case in point is Oracle's Andy Mendelsohn, Senior Vice President of Oracle Database Server Technologies, who last week conducted a special webcast with Kash Rangan, a research analyst at Bank of America Merrill Lynch.  The topic was Big Data and why this new era is a good thing for Oracle and its stable of technologies. 

Stage 1: Combating the Noise
Picking up where Oracle left off from Kash Rangan's 2011 event when Mendelsohn completely disavowed the NoSQL/NewSQL movement beyond an interesting academic discussion, Mendelsohn began last week's talk by gushing about the vibrancy of the relational database market and citing Gartner's recently published relational database market report for 2011 that showed a very healthy 16.3% growth rate. 

Stage 2: Combating the Hope
Mendelsohn's opening comments suggest incumbent Oracle has moved to the next chapter of its playbook to align with the second stage to which Steve Duplessie refers as the Hope Stage.  This is the classic 'We know the industry and selective customers are buying into the Big Data Noise, and Hope springs eternal, but transactional, structured information is still the crown jewels of most organizations which everyone knows is best stored in a relational database' Oracle type of posturing. 

Within the call's first five minutes, Mendelsohn essentially teed up for his subsequent series of misleading statements and lies, holding firm to his strong relational database orthodoxy for the remainder of the webcastHere's a perfect example in reference to one of his comments about the variety and value of Big Data: 
"And so the real key isn't that relational databases can't store this data.  Relational databases can store all this data.  The real key is that most of this data is worthless to the business.  You don't want to store it.  You want to capture and store the really valuable information, and that's what you want to store in your high performance relational databases (aka Oracle)."

This is Oracle incumbent speak for telling the market it has no intention of supplanting its core Oracle relational database that runs on expensive processor cores with some niche NoSQL database... besides, we already have such a product, and it's called Berkeley DB.  Together with our flagship Oracle database, we meet all your structured and unstructured data requirements for the next five to ten years.

Stage 3: Combating the Legitimate (Market) 
Of 'course no Big Data discussion is complete without talking Hadoop.  And since last year's Merrill Lynch webcast, Oracle has announced a partnership with Hadoop distribution vendor, Cloudera.  It's bundling Cloudera's CDH with the Big Data Appliance, Oracle's new engineered system that's the ultimate ETL processing machine to move data between Oracle and Hadoop really fast (but, by no means, cheap).  Regarding these technologies, Mendelsohn had the following to say:
"So the Big Data Appliance is our engineered system for running Hadoop and the Oracle NoSQL database at very high performance and very good time-to-value.  So if you're an Oracle customer today and your business wants to get into the Hadoop space, you can just call us up, and we will give you an - sell you a rack, which is the Big Data Appliance, a rack which is essentially an 18-node Hadoop cluster-in-a-box.  It's got lots of cores, lots of memory, lots of storage...And the key thing in the Hadoop world is people don't want just 18-node clusters, some people want to ride in 100-node or 200-node clusters...this Big Data Appliance is a building block."

This is Oracle incumbent speak for saying we have a Big Data product, too.  Oracle has all the functionality that you, Mr. Customer, said you need, and now we have it.  All you have to do is call us up, and write us a check for a cool million bucks in hardware, software and annual maintenance fees per 18-node cluster that guarantees Oracle a healthy revenue stream of maintenance for years to come.

After plowing through Oracle's Acquire and Organize information lifecycle strategy, Mendelsohn tackled the Analytics piece.  In Oracleland, this is where Exadata and Exalytics are positioned to compete with purpose-built systems designed for complex, analytical workloads.  

This is a Teradata mainstay market that's invited fierce competition, not only from Oracle, but from IBM's Netezza, HP's Vertica, EMC's Greenplum, Microsoft's PDW, and SAP's HANA in-memory and it's acquired Sybase IQ columnar database technologies. 

Massively Parallel Processing (MPP) shared nothing architectures, in-memory for both transactional and analytical processing, and column rather than traditional row-based storage is all the rage today, and Oracle does a lot of bolt-ons to its core relational database to provide similar benefits of these newer, more performant technologies. 

So how does Mendelsohn address these topics in the Q&A session?  Very predictably of an incumbent, he positions each as just another feature of the relational database -- Conveniently, the very product that incumbent, Oracle, enjoys a 48.8% share of a $24 billion market. 

I particularly liked what Mendelsohn had to say about in-memory databases
"So in-memory databases are an interesting technology... Going back to my car analogy.  Cars have evolved a lot over the years, and one of the things people have sort of figured out is that you can get really good performance at low cost with these things called turbochargers.  You stick them on the engine, the engine runs a lot faster... they're expensive [just contradicts himself] so you don't put them in all the cars, but you put them in some of the cars.  And you think of in-memory database technology as the same sort of thing.  It's another feature of the relational database... But pure in-memory databases where you say, 'Okay, all that data in my database has to sit in Memory?'  That's really expensive.  It's like saying in the car industry, everybody's going to buy Ferraris... We all know everybody doesn't buy Ferraris."

This is Oracle incumbent speak for telling the market these newfangled turbo engines need to be used sparingly because they cost a lot of money (aka it's an expensive, niche market), and we can provide you the same benefits of speed with our flash-enabled Oracle Database Machine. 

In other words, you can blissfully ignore these new innovations because not everyone is going to use the stuff anyway, and we implement a similar technology that's better and cheaper.

So there you have it.  A classic incumbent response to a disruptive market force played at a high level.  It's clear Oracle has taken notice of this nascent market and has entered it with much fanfare and bravado.  

Not only is Oracle competing to win, but they're playing to ensure no stinkin' startup or an 'SAP on drugs' shifts major value and money out of the relational database racket they've been playing for 30 years -- and very profitably, I'd say.

Saturday, March 3, 2012

Miles' Law and the Evolving Hadoop Ecosystem

So you're likely asking what is Miles' Law and what the heck does it have to do with Hadoop?"

Miles' Law says: "Where you stand depends on where you sit". The concept is probably as old as Plato, but the maxim originated from Rufus E. Miles, an assistant secretary who served under three U.S. Presidents, Eisenhower, Kennedy, and Johnson.

Although Miles used the phrase in reference to a bureau employee who had left to join another government agency in the 1940s, it certainly speaks to what we're seeing in 2012 within the Hadoop movement.

The various vendors are jockeying for position whether they're a legacy incumbent with expensive, proprietary kit or an emergent commercial open source company with inexpensive tools to build an early adopter sales pipeline.

This week's Strata Conference in Santa Clara was a sold-out show, and since Strata 2010, Hadoop has gained real enterprise traction. It's being used to solve high-value business problems in nearly every industry segment today. This rapid adoption is gaining so much steam that the industry stalwarts like IBM, Oracle, Microsoft, and EMC are getting in on the action, establishing partnerships with Hadoop distribution vendors Cloudera, Hortonworks and MapR. And all are carefully casting Hadoop in a somewhat subservient role to their existing portfolio of products so as not to canabalize them.

Playing nice to drive interoperability while preaching the traditional holistic, enterprise top-down mantra is what these big incumbent vendors will stand by. After all, they're sitting on huge cash cow businesses representative of the old, monolithic, centralized paradigm where expensive processor core licenses, long-tail maintenance fees and SAN storage prevail.

Oracle and EMC probably have the most to lose as the Hadoop ecosystems evolves and matures. That's because the best use cases for Hadoop are all about off-loading computational intensive tasks away from expensive Oracle cores and storing huge amounts of raw data away from expensive SAN storage.

Big, new markets just like new roles in government drive change in strategy, allegiances, attitudes, and perspectives. It will be interesting to see how the Hadoop/Big Data movement evolves and what strategies the various players develop to capitalize on their respective positions of strength to successfully compete.

In any ecosystem, where you stand depends on where you sit.