“It doesn’t matter who delivers it, how it operates or how it’s provisioned – but it matters how it runs,” says Heptio’s Craig McLuckie.


AUSTIN — Craig McLuckie isn’t a fan of every buzzword flitting across the tech landscape. He admits that multi-cloud wasn’t one of his favorites, but he’s changing his mind.

“I’m starting to see a deep legitimacy to multi-cloud,”  says McLuckie,  co-founder of Kubernetes and now CEO of Heptio, “I used to nod and smile when people talked about it, but never really believed it. I’m starting to see it for reals now.”

Multi-cloud was one of the three main pillars that he sees for the future of Kubernetes — the other two are improving developer productivity and tackling enterprise — that he elaborated on during keynotes at KubeCon + CloudNativeCon North America 2017.  And while it’s worth noting that multi-cloud is really happening, “we really need strong conformance” to make it materialize, he says.

He gave a shout out to cloud providers who introduced Kubernetes-as-a-service offerings, especially those that are upstream friendly. McLuckie says all signs point to positive because “we’re getting into a situation where it doesn’t matter who delivers it, how it operates or how it’s provisioned – but it matters how it runs.”

Circling back to the stress conformance, he added: “for us to get to a high level of assurance that Kubernetes is working exactly as it should conformance really matters. We need to have the ability to attest that these clusters they’re semantically consistent that they have the same behavior.” To this end, the CNCF-certified Kubernetes program is important, he says.

“When you see that Kubernetes logo on someone’s offering you can have a relatively high degree of assurance that it’s working as it should..” This in turn makes hybrid cloud real. Operators can run something on-premise — something they construct or from distro providers — and something from a public cloud that’s delivered by their preferred container-as-a-service (or Kubernetes-as-a-servce.)

The next question is to ensure that everything is semantically consistent in these
environments. Making multi-cloud a reality hinges on the ability to run the same tests in both places and the ability to certify your on-premise piece. (Here he gave a quick plug to Heptio’s Sonobuoy, a diagnostic tool for running Kubernetes conformance tests.)

For McLuckie, multi-cloud is not about taking an application and running it in two clouds. There are always a few exceptions, he says — for example, banks running long stateless cycles or bitcoin mining. But “if you’re doing anything that has massive amounts of data, nothing creates gravity or inertia like data,” he says adding that when you start to pull in service dependencies they will hold you back, “you have to be judicious about it.”

He defines multi-cloud as the flexibility to pick which cloud to deploy your new service to — being able to run things in different regions, have the data in those regions and the ability to pick cloud providers for those regions as well.

How to get there? Provide not just a “tested stable narrative” but get to the point where you can start to extend governance, risk management, compliance, policy and security enforcement with a common abstraction across the top. “That’s why extensibility will become so important,” he says giving a nod to a Heptio’s recent collaboration with Microsoft to use the Seattle-based startup’s Ark, an open source utility for managing disaster recovery and backup of Kubernetes clusters.

This, he says, is where it really get’s exciting. “This technology allows you not only to take your cluster state and restore it there — but to potentially restore somewhere else…” He admits that it’s not possible yet on the persistent volumes “But you can see the promise of this as a way to start moving things around inside clouds directly — whether it’s Microsoft cloud or between clouds or from on-prem clouds…and Kubernetes is key to this whole thing.”