Microsoft’s cloud strategy has long focused on the kind of hybrid cloud deployments that allow enterprises to run workloads in a public cloud like Azure and in their own data centers. Azure Stack, its project for bringing the core Azure services into the corporate data center, is the logical conclusion of this. If developers can target a single platform for both the public and private cloud, the thinking goes, then hybrid deployments become almost trivial.
After a year of technical previews, Microsoft has now delivered the first release version of Azure Stack to its hardware partners so that they can finish their certification process. As Microsoft announced at its Inspire conference today, partners like Dell EMC, HPE and Lenovo will start shipping their integrated systems with Azure Stack in September. Other partners, including Huawei and Cisco, will launch their systems at some point in the near future.
“One of the key things we truly believe is that hybrid is a key differentiator for us and a steady state for our customers,” Mike Neil, Microsoft’s corporate vice president for Azure Infrastructure and Management, told me. “We don’t believe every customer will move to the public cloud.” The reasons for this are pretty varied, but for quite a few companies, it’s simply a matter of remaining in control of their data — often for regulatory and/or data sovereignty reasons. “Many of them have their roots in their environment today — in their data centers or their hosting partners’ data centers,” Neil said. “We wanted to provide a solution that met these different needs.”
During the technical preview period, the team also saw a number of users who wanted to use Azure Stack to power their applications at the edge of the network (or even completely disconnected from the wider internet). Those may be banks, for example, that want to do so for security reasons, or oil exploration companies that need to run their own mini-data centers in extreme conditions where connectivity can’t be guaranteed. Another example Microsoft likes to highlight is Carnival Cruise Line, which is deploying Azure Stack on some of its ships to power many of the day-to-day operations of running a massive cruise ship.
Microsoft is obviously no stranger to building applications and services for the data center. With Windows Server and Systems Center, it has long had a presence there. Unsurprisingly, Windows Server forms the basis of Azure Stack, too, but overall, the company is taking a completely different approach to how it delivers Azure Stack to its users.
Pay as you go
Take how Microsoft plans to charge for Azure Stack, for example. While enterprises can buy a standard, outright license and service contract for the product, Microsoft actually expects most companies to choose another option: pay for Azure Stack in their own data center just like they do for resources in the cloud — by how much they utilize the service. If you spin up a new virtual machine, you pay Microsoft. If you don’t use it, you don’t.
The price for this will be significantly lower than in the Azure cloud, of course (enterprises are using their own hardware, after all), but it’s an interesting model because it aligns Microsoft and the enterprise’s interests: if Azure Stack doesn’t work as expected and if developers don’t use it, then Microsoft doesn’t earn anything. If it works well, both the enterprise and Microsoft are better off.
“If people can’t get it deployed, we get no money,” said Jeffrey Snover, one of Microsoft’s technical fellows and the chief architect for the Azure Infrastructure and Management group (which includes Azure Stack). “I have this very strong incentive to make sure my customers are always successful.” Neil agreed and added that this move means that Microsoft can now be an honest broker because its sales teams can have an open conversation about public vs. private cloud with potential customers, knowing that they will get the same bonuses no matter which option the customer chooses.
Snover didn’t hold back with his criticism of competing projects like OpenStack, which started out with a mission of bringing an AWS-compatible platform to the private cloud and has since grown to be one of the largest open source communities in the enterprise space. It’s no secret that OpenStack can be hard to deploy and update (though the community has made some advances here in the last few years). Snover argues that it still takes too much work to deploy these open source cloud platforms. “It was a path that was not succeeding for people — or the utilization of their cloud wasn’t very successful,” Snover said. Though, to be fair, the OpenStack team can also point at plenty of successful deployments.
Because the team wanted to ensure that customers won’t have any issues with deploying Azure Stack, the team decided to deliver the service with an appliance experience. That means Dell EMC, Lenovo and HPE will offer integrated systems that have been certified to run Azure Stack. This is meant to avoid both scenarios where companies struggle to deploy the platform but also issues down the line. Many companies don’t update Windows Server or even Windows on the desktop, after all, because they rely on an old piece of peripheral hardware that has reached its end of life, with no compatible driver for a modern OS in sight. “We don’t want our customers to have any drama,” Snover said and stressed that its partners will deliver Azure Stack on industry-standard servers (starting with four-node systems) and not on some one-off machines with gold-plated components.
It’s also important to note that the Azure Stack release schedule will closely follow that of Azure itself. Users will be able to defer updates for up to six months, but after that, they will have to update or go out of support. Thanks to the Azure’s versioned APIs, that shouldn’t be an issue. New changes will always come to Azure first, though, and then move downstream to Azure Stack.