Just the other day I was talking with a tier-1 industry
analyst about the next generation of IT infrastructure and who he thought the
significant players would be. I was not surprised to hear his top 4 (in no
What I was surprised to hear however, was who he thought was a viable candidate for 5th place. I would have assumed BEA would easily have taken that spot but surprisingly, he put them at maybe 7th or 8th on the candidate list. His opinion was firmly that the most viable candidates for #3 would be (in no particular order):
SFdC is pretty easy one to swallow. The meme-ometer around SFdC after the Web 2.0 conference is off the charts this week. I swear if Benioff put out a solo album right now he'd rack up more first week sales than K-Fed. But when you dig deeper into the possibility of Amazon, Ebay or Google to take the 5th position, it gets a bit harder to see how a company such as Citigroup would rely on one of these players to run any significant portion of their IT infrastructure.
As I was pondering all of this, I noticed a recent announcement from Amazon hinting they were heading further in this direction. After taking a deeper look I must admit that I am starting to see how Amazon may actually be more of a significant player than I first expected. My jury is still out on Ebay and Google, although it would not be hard for them to replicate some of the moves that Amazon is playing now.
Let’s step back a second and understand why Amazon’s maneuvers are more than just mashup-mania. To some extent, all the infrastructure players of the future are really just web services platforms with some prebuilt applications on top to lure the customers to one platform or another. Oracle’s Fusion, SAP’s NetWeaver, the ever imminent Microsoft Dynamics, and IBM’s WebSphere (sans the applications) - it's all basically the same stuff. If you look at what Amazon has built, it’s not much of a stretch to place them in that club. And to some extent, with the available web services on Amazon’s Web Services hub, one could actually argue they are better positioned than WebSphere who has huge customer penetration but exists primarily as a “we want to build it all from scratch” option. Remember, IBM has no applications of their own to offer. They rely on SAP or PeopleSoft (et al) customers to use WebSphere as an alternative to the ISVs proprietary application server. If you don’t believe Amazon is serious about services or a services platform, read through this fantastic ACM Interview with Amazon's CTO, Werner Vogels. I have said before that Werner articulates more clearly in this article the value of an SOA approach to applications than all the commentary I have seen in the space combined.
But let’s be realistic: a services platform does not make a winner in the market. What Amazon is doing in addition is making a compelling argument to move a piece of the IT infrastructure puzzle to the cloud. Just like Salesforce.com articulated enough of a value proposition to move a piece of IT (CRM applications) to the cloud, Amazon is attempting to do that with their Elastic Cloud initiative. Essentially Amazon is arguing to move the basic stack from the hardware up to the application server over to Amazon’s data center. Baris Karadogan wrote a compelling piece in Venture Beat about why this is not complete folly for Amazon and how this move is heralding the inevitable abstraction of the data center. He recants a similar history in the semiconductor fab industry as a pretty solid analogy.
Also, it should be noted that I have some familiarity with
this issue. I sold my last company, Service Metrics, to Exodus. Exodus, who is no
longer in business, was at the time the world’s largest hosting company. We
served over ¼ of the bits flowing on the internet and had the most global data centers
(I seem to remember over 50 at one point in time). My position was VP of
Research and I was responsible for the future of services strategy in the data center
(e.g. storage networking, streaming media, content distribution, etc..). I
remember there was always a lurking feeling that we were an over inflated
real-estate play waiting for a market crash (which came around March of 2001). No
matter how hard we tried to change things (especially in my group), the reps
were most comfortable selling ping, power, and pipe. And this was inevitably our
demise. The cost of building out class-A data centers resulted in us selling 1
Mbps of bandwidth for $1200/mo (market price) and having it cost us $1500/mo
fully loaded with utilities, headcount, and debt service for the billions we
spent on building out nice facilities.
And now, out of the ashes of Exodus and good datacenter design is emerging a completely new model for actually monetizing these expensive data centers. And not surprisingly (hindsight is 20/20) its coming from applications companies (Yahoo, Google, Microsoft, Amazon) and not from infrastructure companies. Whether SAP, Oracle or IBM want to admit it, as they try to sell into the middle market, they will inevitably find a set of competitors they did not expect competing with a value proposition they may find it hard to compete with. Just like many companies have been comfortable relinquishing on-premise rights to customer data to SFdC to buy the application one user at a time, some companies will be comfortable getting the complete stack from storage up to application server if they can pay for usage in a similarly incremental way. I am looking forward to seeing this unfold.