By Marten Mickos | March 23, 2011
We are barely into the beginning of cloud computing, so any prediction of what its future will be prone to error. Massive shifts in IT, such as the shift away from client/server into cloud architectures, are a function not only of winning technologies but also of users' behavioral patterns and of leading vendors' strategic decisions. That's one reason why prediction is so difficult. The biggest shifts in IT so far have been characterized by a technological break-through yielding a 10x improvement in efficiency or reduction in overall cost. Within this area of technology, one or very few dominant designs have emerged, and within each dominant design, one or very few vendors have risen to significant market prominence. Nearly without exception, those market leaders have come out of small companies growing big, rather than big companies entering new segments.
In the database world, the dominant design was the relational model and the SQL standard attached to it. In the first decade, this market had a handful of strong candidates for leadership: DB2, Ingres, Informix, Oracle, and Sybase. It took another decade for the ultimate leaders to shake out: Oracle, DB2 and SQL Server. Then with the internet came a new dominant design: open source scale-out databases (at a fraction of the cost), and MySQL became the dominant player by a long shot in that market. In middleware the evolution was similar, but with fewer players. It didn't take long for BEA to become the dominant leader of the first wave. And analogous to MySQL, JBoss came in as the modern open source alternative and dominated the new middleware space.
In cloud computing, what might constitute a dominant design? Is it the data layer with big data solutions like Hadoop? The various hypervisor designs (KVM, Xen, ESX)? The operating system (Linux, Linux and Linux)? The infrastructure layer with its API such as Amazon EC2 and S3? PaaS solutions like Force.com or Makara? Services like Salesforce and Netsuite? Perhaps it is not just one of these - perhaps the cloud market is or will be so vast that there is room for several dominant designs on the various layers of the elastic stack.
If we for argument's sake assume that all of the above categories will be home to a dominant design, what determines who gets to be the market leader in such a space? Not always the originator of such a design. IBM must be credited with having devised (most of) the relational model and the SQL language. But it was Oracle that ultimately achieved market dominance in that space. Xerox PARC invented the graphical user interface for an operating system, but it was Microsoft who rose to dominance in that space. And some will claim that it's now Apple's turn to claim that leadership. Bell Labs and others figured out the main architecture for Unix, but it required a Linux to fulfill the whole promise of that design.
It seems that Darwinism determines the winners. The players with the best execution and the fastest learning cycle can outrun the slower or more rigid ones. We can thank Yahoo for Hadoop, but we may end up thanking Cloudera for turning big data into a massive business. We can thank Amazon for the de facto IaaS API, and we may see other vendors too, such as RightScale, build huge businesses on that concept. Or perhaps it will be DeltaCloud which acts as a unifying force for IaaS APIs. We can thank VMware's founders for groundbreaking work on the virtualization layer, but we may end up running all our infrastructure on the open source disruptors KVM or Xen. We can thank various inventors for the fantastic programming languages they created, but we may end up seeing the corresponding business being conducted by PaaS vendors who build online frameworks around those languages.
And, naturally, we will see some players fail if they attach their business to a design that doesn't become dominant. I am sure Betamax video recorders were great. So were object-oriented databases and XML databases. But they faded into oblivion. In the cloud space, there are many new designs and APIs that are vying for a place in the sun. But we know from history that only very few will ultimately make it.
Whoever the winners are, I believe they will come out of the spheres of dominant designs that are now emerging all over the elastic stack of the cloud. Some of these evolutionary developments will be quick, with winners appearing within a few years. Other developments, typically the ones deeper down in the stack, will take several years to shake out. Candidates for dominant designs are paradigms that are so simple that they can scale massively, so open that they can be adopted broadly, and so central to the operation of the cloud that they rise to the top of everyone's buying agenda.