I know hybrid cloud has been a big area of focus for Microsoft. Why has that been such an important part of the strategy?
Part of this is just our experience we’ve built over the last couple of decades of being an enterprise company. We expect that we will have a hybrid landscape for the foreseeable future. Clearly a lot is going to move into public cloud, we’re definitely seeing that. At the same time there are scenarios that people just need to be able to meet. Imagine your factory floor: If someone cuts the fiber cable between you and your cloud you’re going to be losing $15,000 a second if your production line is down. You can’t have that.
Hybrid, to us, is a first-class part of the solution that you’re going to need to have, so we’ve invested in that from day one. The best example of this is our work in Azure Stack, the ability to have a consistent cloud environment for on-premises as well as the public cloud and the ability to figure out how do I leverage both and create the best possible combination of the two. We’ve also put product truth into things like SQL Server and other environments. It’s not so much that we’re just tacking things together as much as we’ve actually made very deep R&D investments to make sure that that the hybrid world is something you can actually materialize and get a lot of value out of on day one.
Let’s talk about Azure Stack, Microsoft’s private cloud appliance. It was recently announced that the product is going to be delayed. So when will it be available? Also, what will Azure Stack look like on a customer premise? Is it going to require specialized hardware that customers need to buy?
We’ve got great relationships with three close partners, Dell, HPE and Lenovo in the preview phase right now. Basically they’re all working on a converged appliance, which will include the hardware plus our software that will go together. What we’ve tried to do is figure out what are the things that make private cloud implementations fail. If people have tried using other software solutions or building their own, why have they failed?
We found there were two key elements to it: One was the software. We try to make sure you have software that can really deliver a cloud pattern, not just VM automation, but an actual cloud pattern with containers and PaaS and all the rest of the things you expect in a cloud. The other piece is hardware and getting that right, making sure it can handle the new cloud patterns. We put those two things together and that is what Azure Stack represents. You’ve also seen from us Project Olympus and some of the work we’ve done with the Open Compute Project. Those are also part of that same strategy to make sure that that kind of converged hardware/software solution that we’re making is something that we can both share with the industry and also drive to scale.
What’s the timing on Azure Stack? Is it delayed from when you initially were hoping to get it to market?
We’ve pushed what we call Technical Preview 2, which we announced at our Ignite Conference a few months ago. That’s out there right now. The hardware along with that is also out there being tested and we have a lot of early adopters that are giving us the beta feedback on top of that. Basically our goal is to get those things in place, take that feedback and continue on the Technical Preview track to make sure we’re meeting all those requirements. We’re targeting this upcoming calendar year.
It seems that some companies are coming to a realization that the public cloud is not a good fit for the workloads they have migrated there. I wrote a story recently about GitLab, which announced it is exiting the public cloud because they feel like they can run their database application more efficiently on infrastructure that they control. Is this a trend? Have you seen customers go to the public cloud then pull the plug on it for cost or performance reasons in favor of running apps in their own data center?
I would say historically for people looking at that, if they’re getting to a very, very large volume then I’m certainly aware of companies that like to really do the customization. What I would say is we doubled the number of compute configurations or VM types that we’ve had just in 2016 alone and that was in direct response to requests – people say ‘Hey, I’d like to have something with more CPU and less RAM, that would be great for gaming.’ Some workloads just can’t use a generic VM.
The N-series with GPUs (Graphic Processing Units) is the most recent example that we’ve brought out. The FPGA work that we’ve announced is another. You’re starting to see more and more types of workloads that need something specialized and I think the public cloud will definitely handle it. That’s not to say that there won’t always be somebody who’s really just hyper-optimizing.
I think the thing you typically look at is: what percentage of my overall portfolio is running infrastructure and is that really the thing that I actually want to be doing? Am I getting the kind of scale wins that come out of a supply chain where I’m spending billions of dollars and buying millions of things at a time? Are you really going to be able to compete in that kind of environment? Are there cases of that? Sure, I’ve seen it, but I do think that as the public cloud evolves it will handle those scenarios. I think the number of companies that are in a position to run their own is pretty small.