IBM Cloud Q&A: Kubernetes takes center stage

Jim Comfort is the right guy to talk to about IBM Cloud, not just because he’s general manager of IBM GTS Cloud Services. By his own account, Comfort was responsible for the 2013 acquisition of SoftLayer, whose 13 cloud datacenters immediately got IBM in the public cloud game.

At the time, IBM seemed ready to battle it out with Amazon Web Services, aggressively expanding points of presence around the world, atop which IBM’s promising Bluemix PaaS—and Watson analytics capabilities—would mount a formidable challenge. But Bluemix was slow to propagate across SoftLayer infrastructure and the prospect of IBM as a vigorous public cloud competitor to Amazon, Microsoft, and Google faded.

I spoke with Comfort in advance of the IBM Think conference in San Francisco this week, where the main announcement is the availability of containerized Watson AI services ready to be deployed on any cloud. The conference plays out against a backdrop of gigantic news—IBM’s acquisition of Red Hat—about which Comfort could say little, because the deal hasn’t closed yet. Too bad, because as InfoWorld’s Matt Asay says, the merger raises the possibility that IBM might finally succeed in becoming a cloud powerhouse. 

Yet Comfort still made clear that IBM was committed to playing at the container layer where Red Hat shines, rationalizing Kubernetes management and orchestration and offering services that will differentiate IBM on public, hybrid, and private clouds. The following interview with Comfort has been edited for brevity and clarity.

Eric Knorr: I know you can’t talk specifics, but in broad strokes, how do you think the Red Hat acquisition will fit into your cloud strategy?

Jim Comfort: We see it as a pure amplifier. We’ve already been very focused on open standards and open source, active in the community for years. We’re very active in Kubernetes. We’re very active in Docker and especially on the orchestration and management sides. Our container-based technologies for hybrid cloud rest on OpenStack, [Red Hat] OpenShift in that case, and the massive developer communities they facilitate.

Knorr: You anticipate no conflict between Red Hat’s OpenShift PaaS and IBM’s Cloud Foundry PaaS, for example?

Comfort: No. Multiple things will exist in the market. The larger theme, independent of Red Hat, is the evolution around containers and Kubernetes. For 80 to 90 percent of the benefits that everyone has historically had, the only option was public cloud. The pieces associated with it can now be done in containers anywhere and give you the flexibility to run on essentially any and every public cloud that runs a Kubernetes service. It makes design-once-to-play-anywhere possible in a way that was never really possible before.

Knorr: Hybrid cloud has always been central to IBM’s cloud strategy…

Comfort: Hybrid and IBM, some people viewed it as a defensive mechanism, and that’s absolutely not the case. It is a statement on the reality of enterprise transformation.

Knorr: How would it be defensive?

Comfort: If you say hybrid some people think you’re reducing your interest in the public side of hybrid.

Knorr: I see. OK.

Comfort: Let me put that in perspective. By 2020, about $600 billion will be in traditional IT, $450 billion in private clouds, and $609 in public, either IaaS or dedicated off-prem. That is the market reality. We want to play in all those markets. The hybrid cloud, meaning you have some things on-prem and some off, is central. Second, 94 percent of enterprises are using multiple clouds and about 70 percent are using more than one public cloud provider. Hybrid is whatever combination makes sense for the enterprise. Up until now the market has defined multicloud as: You use native services from providers A, B, and C. There’s a multicloud that will emerge where containers and Kubernetes become a common substrate and you have much more flexibility and portability and reuse. I don’t believe in fanciful bursts where things fly all over the place. That’s not the reality. It’s more about design once, manage centrally, and deploy anywhere.

Knorr: That’s been a promise for a long time.

Comfort: We actually practice this. We initially deployed that in multiple on-prem data centers and then it’s been deployed in one on-prem and two, three, or four IBM Cloud sites. It can be done on AWS or Azure, any cloud site. We actually practice what we preach. It is not possible with what everyone is calling the native services model. It is possible with what we see in the container world.

Knorr: One of the requirements for hybrid cloud is that you’ve got to have a private cloud. If you’re operating at the application/container/Kubernetes layer, then it’s not as important what sort of virtualization infrastructure you’re running, whether it’s OpenStack or VMware or whatever, right?

Comfort: It certainly becomes less relevant. You don’t have to complicate it with all the deployment model choices. Just focus on getting really good at containers. We see people doing that on-prem and we see people doing that when they move to a dedicated public cloud.

Knorr: What percentage of your customers is that far into containers at this point?

Comfort: Nobody is that far. All of them are working with it. The vast majority see this as the core of their strategy going forward. Most are still pretty early in the days of actually hardening and deploying. The largest and biggest have done quite a bit. They may be up to a few thousand apps. Most are still in a couple of projects at scale.

Knorr: The big announcement at IBM Think is that Watson capabilities will be containerized and portable to any cloud. 

Comfort: That’s AI and machine learning all being portable as containers. It can go anywhere and you can bring consistent methodologies and models to wherever the data sources are. Very powerful.

Knorr: Deploying across your SoftLayer infrastructure, across AWS and Google and Azure and anywhere, is the whole intent here?

Comfort: Take it from a client point of view. If I have to work in three different clouds, plus on-prem, and have four completely different methodologies for data analysis, data aggregation, model learning, training and deployment…it’s hard enough to do it well once. How are you going to do it four different ways? And how are you going to know the provenance of all those models and sustain all that? It’s more about being consistent so that the client can choose to have one methodology that works anywhere. When viewed from the client standpoint it’s a lot more about speed, consistency, efficiency of investment, because you can go everywhere.

Knorr: What’s on the roadmap for IBM Cloud that you can talk about?

Comfort: What we are investing very heavily in is making sure that we complete our focus around containers. Red Hat will be an amplifier in the second half of the year, for sure. One of the largest requests from clients is for a consistent way to manage this complex environment and we’re putting a lot of investment around the elements of a cloud management platform. If clients try and buy their own pieces they’ll fall into best-of-breed procurement hell. They’ll get tons of parts and they’ll spend forever integrating and getting limited value. The unsolved problem in the multi-cloud industry in my mind is truly consistent management. We think there’s a huge opportunity for us to create a more integrated platform.