Multiple forces are behind a $34 billion IBM-Red Hat deal that has the top commercial open source advocate joining...
up with the 100-plus-year-old force in enterprise computing.
Some top-level trends got the major attention. The principals were quick to point to such drivers as hybrid cloud computing, which matches public and private cloud processing and data, as well multi-cloud implementations in which unimpeded data moves between clouds as quickly and easily as just-dismissed high-school kids through subway turnstiles.
But just as important as these forces is something the late semiconductor impresario Andy Grove would call an inflection point -- a disruption of computer architecture that will have a ripple effect on methods of data management. It takes the form of new approaches based on flexible, swappable cloud containers, Kubernetes and microservices.
A 'serverful' of architecture
Both IBM and Red Hat had important roles in establishing the formerly dominant three-tier computer architecture. It was based around application servers. Let's call these monolithic, "serverful" architectures.
Both IBM and Red Hat now have to shift their enterprise software stacks fundamentally in order to prosper in the new world of serverless computing that microservices presaged.
Both players in the IBM-Red Hat deal are presently busy lifting out application servers and shifting in microservices, and they may find strength in the tighter coupling of services that comes with the acquisition. Behind talk of hybrid and multi-cloud are containers and Kubernetes.
"Cloud architecture is the architecture of the future, and it's already started," said Adrian Bowles, analyst at Boston-based Storm Insights, referring to forces driving the IBM-Red Hat elopement.
"People want to take advantage of scale-up and scale-down without having to change platforms," Bowles continued. "Increasingly, you will see containers that let you move from one cloud to another."
Reminiscent of ClouderaWorks
Adrian Bowlesanalyst at Storm Insights
In dollar terms, the IBM-Red Hat deal dwarves the Cloudera-Hortonworks merger, which now goes by the handle ClouderaWorks and was the last biggest shakeup in software. But there are similarities in underlying motive, with cloud-native platform enhancements for data handling being one.
Hadoop mainstays Cloudera and Hortonworks both faced significant software rewrites as they hastened to the call of the cloud. At least one reason they merged was to pool their resources to prepare for wider Kubernetes container adoption.
Both the IBM-Red Hat and Cloudera-Hortonworks deals "reflect the fact that we are going through a huge paradigm shift in the enterprise stack," said Jack Norris, senior vice president for data and applications at MapR, a Hadoop vendor based in Santa Clara, Calif., and Cloudera-Hortonworks competitor.
"This is about recognizing that execution of workloads is getting more distributed, and that, with AI and machine learning just around the corner, the types of analytics we are doing are changing," Norris said.
That is to say, Hadoop is now just one of many workloads being jockeyed around in the new microservices containers. Driven in part by enterprise stack disruption, the fallout from the big Hadoop merger has special meaning for MapR.
When the Cloudera-Hortonworks pairing is completed, MapR jumps from third to second in the independent Hadoop vendor parade. Norris said MapR, which never tied its fortunes to the Hadoop Distributed File System, is in a favorable position as containers gain traction.
Hark, the hybrid architecture sings
Both of the big deals were preceded by news that IBM, Red Hat and Hortonworks were set to work on an Open Hybrid Architecture Initiative to build out a new architecture for Kubernetes-managed container workloads on cloud infrastructure. This is just one of several alliances Red Hat and IBM have formed over the years, but it speaks directly to the most telling architecture trend of the day.
When it goes with IBM, Red Hat seemingly clears up one of the questions long hanging over its open source strategy. That is, as analyst Stephen O'Grady of RedMonk, based in Portland, Maine, posed on the RedMonk blog earlier this year: "Should Red Hat buy or build a database?"
Even within the IBM fold, Red Hat has plenty of leeway in terms of open source cloud databases it can support. But its focus will clearly be on harmonizing these data workloads with microservices orchestrations especially helpful to application developers.
Talk of hybrids and multi-cloud are important, of course. But the feeling here is the architectural shift underway -- the leap from monolithic application servers to Kubernetes-infused serverless servers -- is a major force driving IBM and Red Hat together.
Cloud is no longer about where data processing is done. It's about how.