When the first murmurs of "converged" infrastructure -- or what we refer to at ESG as integrated compute platforms, or ICP -- reached my pessimistic ears, I immediately thought, "Here we go again." What was old is now new -- just another in a long line of examples of the IT industry rebuilding the mainframe.
That's what VMware amounted to: "We can make one physical machine look like 13!" Isn't that exactly what the mainframe did all those hundreds of years ago? As a matter of fact, the word "virtual" nearly always amounts to this same idea: Make something look like more -- but actually execute on less. Less physical kit, that is.
Why? Because less is better. Less stuff to run equals less stuff to break, less power to consume, less space to buy or lease, less, less, less. Less makes sense. IT people love less. And yes, having less vendors to deal with would be welcomed.
An ICP is theoretically designed to give you less of the things you don't want (headaches), and is optimized to execute more of the things you do want (transactions, revenue, time savings, etc.).
And so I've been watching. I have watched the likes of Hewlett-Packard push the notion of converged infrastructure. Initially, HP wanted a way to sell more of its own stuff when selling other pieces of its own stuff. Eventually, this worked, because it provided more value to the customer -- easier-to-implement kit, with one throat to choke when something goes wrong. And the customer thinks, "Now that I've got a standard bill of materials, I can start to optimize for certain application environments or cloud-type functions." What's not to like?
In a similar vein, VCE, the joint venture between EMC, Cisco and VMware, has been building massive consolidated systems for equally massive production environments. That gives customers the ability to collapse down a ton of infrastructure silos into large, powerful integrated systems -- preconfigured and optimized for VMware environments.
Then there's NetApp and Cisco in the midmarket with FlexPod, and EMC's VSPEX.
IBM is in the game, as is Dell. Several startups, including Nutanix and Simplivity, are building purpose-built collapsed stacks.
And guess what: It's taking hold.
Customers are clearly seeing the benefit of "collapsed kit." (I'm going to use as many phrases for the same concept as I can until one sticks!) As a matter of fact, I'd say the market has now moved from "vendor push" to "customer pull." That means the market isn't happening just because of all those vendors and their nifty pitches any longer -- customers have started to demand more and more converged kit.
Now, there are three ways to play in the ICP world: do it yourself (buy up all the pieces and put it together on your own), leverage a reference architecture, or buy it turnkey.
DIY is old school. It's not what folks are doing unless they have to, or are seriously bored and just want to. Reference architectures are very popular, and "semi-configured" kit has been successfully marketed by the likes of EMC (the VSPEX initiative) and NetApp/Cisco (FlexPod). Many customers are relying on their distribution/channel resellers to build them a stack that is relatively easy to support. Compared with traditional methods of dragging in a thousand components and putting it all together, this has been an improvement welcomed by buyers of IT stuff. Judging by the numbers, demand is increasing.
But turnkey is turning back time. Steve Jobs was the first of our generation to recognize that in order to control the user experience, you had to do it all -- soup to nuts. The philosophy has served Apple well. It was the same thing that IBM had done with the mainframe, controlling the whole enchilada. Eventually, of course, we all reacted against IBM's mega-control in a big way, because of the fear of a single supplier, higher cost, etc. They were all valid concerns -- then.
Today, the soup-to-nuts stack is selling like mad because the benefits of tight integration and streamlined support are creating operational efficiencies so great that they far outweigh the potential negatives -- at least to a huge percentage of the market. We've come full circle.
Take Oracle, for example. Oracle is the only systems vendor building a truly 100% Oracle super-stack -- named Exadata. Oracle OS, Oracle applications, Oracle database, and all-Oracle kit. Oracle sales, support and services, too.
Best of breed? Nope. But it hasn't mattered all that much. When you have an effective monopoly on the one piece that is critical and expensive as all get out (the database), the other pieces don't have to be perfect -- just good enough. The integrated benefits outweigh the risks or shortcomings in the mind of the majority of that market.
A software company that had never sold hardware before a year ago or so is selling lots of it now. "Hardware, the software delivery system" might as well be Oracle's new tag line. Oracle customers, who are happy to tell anyone who'll listen how much they hate Oracle, are gobbling up these stacks. Why? Because it's easier. It's not cheaper -- it's easier.
I look at SAP and just know it has to be driving someone there mad that it gets boxed out of deals where it might have a better point solution but Oracle has the "appliance" factor. I don't know how much it's costing SAP, but I guarantee you it's costing -- these deals are big, for many millions of dollars. And SAP is missing out on those deals because a lot of SAP customers want easier too. They want someone to give them the whole enchilada.
I think that SAP will have to give you Mexican food, my friends. Enchiladas for everyone. I say this because SAP has proved to be a smart, resourceful company. As such, it's reasonable to think that SAP will have to find a dance partner in order to stem a growing tide. Yes, I'm talking about a growing tide of enchiladas -- that is, ICP.
What might SAP do? Well, a ton of SAP customers run EMC big iron already, so a deal with VCE would make sense. Smaller shops have the same desires for the Microsoft application suite, and everyone seems to have the taste for VMware.
Further, VCE has one of the most interesting attributes of all of the converged offerings (that for some reason it doesn't tout very loudly): It does all the heavy lifting when it comes to patches and regression testing. This may not sound like much, but it's a huge deal. There are dozens or even hundreds of components in a system like this, and that means there are dozens or hundreds of patches that come out each year. Each of those patches has the potential to cause total chaos with other components of the system, thus rendering the whole thing useless at the flick of a switch. VCE does all of the regression testing on all of the patches of your system back at the factory -- releasing them (and normally even being responsible for installing them) in logical, bundled ways, and in logical time frames. It will roll up all the patches that pass its regression testing into one update, and then roll it out to your system. That removes vast amounts of time, money and risk from your operation, even if everything were to work perfectly all the time -- which it never does.
Regardless of the mix, it has become apparent that this consolidation trend will continue -- at least until some new management/virtualization paradigm takes hold, rendering all hardware as disposable componentry. This has been my "data center OS" concept for the last decade. That's a long way away. In the meantime, I see this trend accelerating. So to all those vendors without dance partners, you best find your partners or look forward to a lonely time. Your customers are going to demand it.