Recently Chris Mellor wrote a piece entitled “Converged stacks will retard storage innovation” that everyone in the Storage Industry, both customers and IT wonks, should read. Chuck Hollis over at EMC wrote his response in a recent blog post “I Respectfully Disagree.”
Beyond the titles which plainly state the conclusions of the authors, Chris predicts a loss of innovation in storage while Chuck predicts a shift in where or who is doing the innovation. The undercurrent here is that if the “open” storage business closes, technology incubation start-ups will cease to exist, thus staunching the flow of innovation. After all, IBM, EMC, Dell, and HP have been reinvigorating their product lines by purchasing start-ups rather than innovating from within. So then what happens to innovation when the start-ups are gone?
At the end of the day, I claim customer perceived value of IT Systems will force innovation throughout the hardware and software stack. But where in the stack and does anyone care?
Well, if cars were built with standardized piece parts like engines, drive-trains, suspensions, electrical systems, and fuel systems, you would have an automotive industry that looks more like the IT Industry does today. You don’t buy computers today the way you did when the industry started. Instead, you buy CPU blades or servers, storage, networks, printers, clients, tape (or not), and piece them all together.
It may be quaint, but the term “computer” used to refer to the entire collection of devices that made any of the components useful, i.e. the system. Sure, some people let the Oracle guy, or the IBM guy, or the HP guy quote all the Oracle, IBM, or HP parts together, and some don’t. If you don’t like XIV because, well, it’s crap, then you can buy Pillar, EMC, or NetApp storage and gain performance, function, efficiency, and a clue.
But the threat is there, so each part sold by a system vendor needs to stand on its own. Many people take this for granted because they are too young to have experienced single-vendor-only options, like IBM in the early 60’s. Personally I was just starting kindergarten then, so my experience is limited, but I read a lot.
There are serious downsides to the functional component route – it is complex, and requires lots more thought than a complete solution purchase. Greenplum, Teradata, and Exadata are integrated data base appliance hardware and software stacks.
They all purportedly (and actually do, in the case of Exadata) yield great performance compared to a roll-your-own approach with standard components. There is no way HP can make an Exadata out of its stuff with Oracle Data Base; Exadata will kick Leo’s butt around the block before HP gets its pants on. From this example – what’s not to like if you are the customer?
Everything’s great until somebody corners the market and starts gouging the customers and/or locking others out of the business – think of pre consent decree IBM for example. Back to the car analogy, I like buying cars and I don’t want to spend the time building them out of pieces. I’d change my opinion to avoid paying $1M for a car that normal market forces would push into $100K … so we need to make sure that normal market forces exist.
Microsoft has often been cast as so big that normal market forces don’t affect it, so it can give everyone the shaft and no one can do anything about it. But then someone actually does something about it, like Linux, or Google Chrome, or a government. The corrections take a long time, sometimes a decade or more. The industry is littered with examples, like DEC, Oracle, Compaq, EMC, making things like VAX, relational data base, PCs, and mainframe and open storage. These companies offered solution pieces that eclipsed their all-in-one competitors. Nothing wrong with that, eh?
Mellor worries that if companies like IBM, EMC, HP, Dell, and Oracle all sell complete solutions, there will be no interoperability or open APIs. If customers can only buy equipment from their system supplier, there’s no opportunity for a new supplier to innovate and sell. The only possible competition would come from a completely different full solution stack from a different system vendor. Chuck’s simple argument is that the innovation isn’t gone; it just shifts elsewhere, like a Second Law of Thermodynamics for Innovation.
While Chuck could be correct, he misses Mellor’s point. Chris didn’t say that storage innovation would cease, but that the open technology benefits would fade away. At least that’s what I think Chris was saying, because nobody would claim that the automobile engine isn’t improving now or that IBM pre-consent decree storage was in stasis. For gosh sake, they invented the disk file and created many generations of the technology with virtually no competition in a totally closed environment.
What this vertical stack ownership would amount to is buying automobiles not pieces which is working just fine for me, but the automobiles still must be competitive, hence massive innovation across the board. You just won’t be able to pluck a component from the “stack”, and substitute another vendor’s component for it. No Honda engine in a Ford automobile for example.
Do Customers care? I hate to say this, but the truth is, I think not. Customers want competitive, high performance solutions to business problems. If the Ford automobile is better than the comparable Honda, they’ll buy it. They are buying transportation, not an engine, or a stereo system.
So why then did the “open systems” business come into being? Well, there are a lot of opinions, but I think it grew out of an industry where one company like IBM dominated the market to the point that their margins enticed others into the business. When the non-IBM products were inherently cheaper, or better, or much less onerous to maintain (read expensive), customers started screaming when those better options were cut off by business practices invented to keep competitors at bay.
Some of those business practices caused governments to get involved (1969 suit by the US Gov against IBM). The US Government opposed the use of monopoly power by IBM, ATT, and Microsoft to stifle competition. Without going through the last 50 years of history, companies were forced to “open up” (or unbundle) their technology, resulting in the birth of many companies that could plug into these big guys’ stuff, and Bob’s your uncle.
The Pendulum Swings
With vertical stacks, the products can be faster, and more efficient. The downside is that you are locked into those solutions because each stack owner tries hard to well, lock you in. They can do it by making better stuff which is fine, but they often perpetuate the lock-in by employing “voiding-the-warranty” tricks, proprietary formats, and other such chicanery. Lock-in tricks can lead to single vendor domination starting a new cycle.
Well, for right now big IT companies make both open and proprietary solutions. Exadata and its competitors are proprietary. They tend to be expensive, but they offer huge performance gains over open systems approaches. However, there is a limit to this specialty approach.
After all, there are thousands of applications and nobody wants a thousand “appliances” in their IT shop. Thus utility products that serve the general computing needs of business should continue to flourish. Most applications will derive only marginal advantage from a vertical stack, making utility computing efficient and the best general solution.
So, I think Chris and Chuck are both right, and wrong. I think the dynamics of this industry will be profoundly affected by vertical stacks and it should be a very interesting game to watch unfold. But rest assured, there will be innovation everywhere, because there’s money in it. If EMC can’t buy companies like Data Domain, or Isilon, or Data General, they will have to do some of this stuff on their own or Dell will kick their butt. No resting on laurels, or business practice shenanigans, else you will cease to exist.
Once you cease to exist, it is really hard to find the light switch.