Microsoft's Azure Stack, which in essence allows enterprises to run a version of the Azure public cloud in their own on-premises data centers, has been delayed until mid-2017. When it does arrive, it will do so in a much different form than initially announced, and the decision has some would-be customers unhappy.
Initially, Microsoft said Azure Stack could be deployed on the hardware of customers' choice. But now it will initially be delivered as a turnkey appliance on hardware from HPE, Dell and Lenovo, the company announced. What's the rationale? Microsoft corporate VP Mike Neil laid it out in a blog post:
Over the past few months, we’ve consistently heard from customers and service providers that they see tremendous business value in the hybrid capabilities Azure and Azure Stack can uniquely deliver. We also heard a universal theme around infrastructure; customers don’t want to deal with complexity in deployment and operations, they prefer to focus on driving applications and innovation.
For us, these learnings reinforce the need to make Azure Stack easy to manage as we translate global-scale Azure infrastructure designs to enterprise-scale environments.
To best meet these requirements, we will prioritize delivering Azure Stack as turnkey integrated systems in the initial general availability (GA) release, combining software, hardware, support and services in one solution. Predictability of system infrastructure enables Microsoft to more-rapidly deliver innovation from Azure to Azure Stack. Based on our experience with Microsoft Cloud Platform System (CPS), we’ve seen customers go from order to operation much faster than they had with their previous processes. Microsoft and our partners can also ensure greater quality and system reliability across the full lifecycle.
Microsoft released a technical preview of Azure Stack earlier this year, which was limited to single-server deployments. A second technical preview is still slated for this year. Microsoft initially planned a general availability release for Azure Stack by the end of this year.
The company is advising customers who want to run an "Azure-consistent" enviroment in their data centers now to use CPS, as Azure Stack will be able to manage CPS environments.
Microsoft's decision received harsh criticism from many commenters on both Neil's and a related blog post.
"This is really disappointing to hear. There has been no mention ... that the initial general availability would force us to essentially buy a private cloud in a box from a vendor that we may or may not already have an existing relationship with," one wrote. "It was communicated to me that upon general availability we would be able to buy our own hardware (vendor of our choosing) and purchase the Azure Stack software/license and that would be it. How much extra are we going to need to spend on vanilla servers, support, licensing etc from one of these vendors who will jack up prices to do something that we are capable of doing ourselves?"
"I can understand it is a big challenge to support such a sensitive stack which requires continuous updates and requires watertight deployment and operational practices," another wrote. "However, MSFT should find a middle ground to help large enterprise customers and partners like us who are not too big nor small."
Anthony Spiteri, lead architect at IaaS provider ZettaGrid, wrote a detailed post reacting to Microsoft's move, going so far as to question whether Azure Stack is a "white elephant":
This is an interesting and in my opinion risky move by Microsoft. ... You only have to look at how much of a failure VMware’s EVO:Rail product was to understand the risks of tying a platform to vendor specific hardware and support. Effectively they are now creating a Converged Infrastructure Stack with Azure bolted on where as before there was absolute freedom in enterprises being able to deploy Azure Stack into existing hardware deployments allowing for a way to realise existing costs and extending that to provide private cloud services.
The point of private cloud is to drive operational efficiency in all areas. Having a fancy interface with all the technical trimmings isn’t what will make an on-premises stack gain mainstream adoption. Flexibility, cost and reduced complexity is what counts.
The Bottom Line
Microsoft has decided to err on the side of conservatism with Azure Stack, choosing to tightly control how it can initially be deployed by customers. On the one hand, it's not a completely outrageous idea to limit Azure Stack at first to carefully configured and managed turnkey systems, as this should help early deployments be more successful.
But Microsoft will have to contend with the valid arguments from customers over the additional—and potentially extra-pricey—hardware costs tied to that approach. One possible olive branch could be modest pricing for the Azure Stack software licensing components, but those details won't be known for some time. Microsoft also has given itself a year to deliver on an Azure Stack GA, during which it could decide to be more flexible in other ways, such as allowing a bring-your-own-hardware option for limited deployments.
Ultimately, the point of Azure Stack is to enable hybrid cloud scenarios, with customers able to move workloads between their premises and the Azure public cloud. Friction at the purchasing point for Azure Stack will do little to help that goal, so it will be interesting to see how Microsoft responds to the initial customer feedback over its decision.
24/7 Access to Constellation Insights
If you’d like unrestricted access to Constellation Insights, consider joining the Constellation Executive Network for analyst advice and analyses that you can use.
