Microsoft took a two-pronged approach towards cloud computing with private cloud hardware and software displayed at its Ignite event.
Jason Zander, corporate vice president for Microsoft’s Azure Web service, showed three nearly identical half-rack servers from Dell, HPE and Lenovo at the event. They were working testbeds for Microsoft’s anticipated Azure Stack private cloud due for release in mid-2017.
The systems—when combined with code from Microsoft’s new partnership with Docker–are going to blind-side the rest of the market for private-cloud infrastructure with a nicely integrated suite.
Azure Stack will initially be a subset of Microsoft’s Azure services that customers will own and operate themselves as private cloud instances. Over time Microsoft will release a more complete set of Azure services into Azure Stack. The benefit to private cloud customers will be the ease of migrating applications on-demand from Azure Stack private clouds into Microsoft’s Azure public cloud.
Azure Stack targets industrial, global transportation and smart-city markets. They need local hardware either because they must continue to operate if they are unintentionally disconnected from network access or because the nature of their use is occasionally connected.
The Azure Stack systems from Dell, HPE and Lenovo have a lot in common as the table below shows.
Table 1: Configurations may change before servers ship. (Images: Tirias Research)
All of the servers implement dual-socket Intel Xeon processors. The three will compete detailed specs of the eight 2U systems. Final specifications of the servers will change over the next few months.
In the public cloud world, where services such as Google, Facebook, Alibaba and Baidu purchase at hyperscale, there is a concept of a stamp. A stamp is a pre-configured rack or perhaps several racks that are purchased at one time to serve a specific set of workloads.
These Azure Stack OEM systems are clearly intended to provide a default certified private cloud stamp that each vendor can then lightly customize. Their half-rack size means they can be deployed more easily at an enterprise friendly scale than full-height racks intended for web scale service providers.
Certification will be important because Azure Stack is not a general purpose server OS. Its first incarnation will be based on a minimum viable set of Microsoft’s Azure cloud stack services. There will be other services at general availability, but not the full set of services supported in the Azure public cloud.
Microsoft is working with the OEMs to ensure that they can differentiate from each other. The last thing that OEMs want is a race to a pricing floor with completely undifferentiated hardware.
These Azure Stack offerings may preempt much of what the Facebook-led Open Compute Project is trying to do with private cloud hardware. In addition, Microsoft’s partnership with Docker may preempt much of what OpenStack is trying to accomplish through enterprise-grade Linux distributors such as Red Hat and SUSE.
Customers who buy an Azure Stack license from Microsoft will also receive a Docker commercial support license. A large part of the appeal of running containers such as Docker is that a container virtualizes an application’s interface to what would be a guest OS in a typical hypervisor.
The virtualized OS interface eliminates redundancies farther down the software stack. Containers are architecturally simpler and somewhat more efficient than hypervisors running virtual machines; they incur less runtime overhead.
In addition, Microsoft and Docker are working to optimize how their products work together. That’s one more reason end users may choose the systems.
Figure 1: From the left, Lenovo, HPE and Dell half-rack servers at the Microsoft event looked similar on the outside.
*– Paul Teich is a principal analyst at Tirias Research, has 12 patents and is a senior member of both the ACM and the IEEE.
This article was first published by EE Times.*