The Private Cloud Is Not What You Think
It was at the annual VMworld show and conference last fall that it dawned on me that VMware’s naming of its core product set was not, in fact, just another example of cloud washing. That, in turn, got me thinking about the myth of private cloud.
Cloud washing is the “us too” arm-waving practice of using the word “cloud” on or about your products to leverage the buzz around cloud for your own marketing. It starts to feel as though every technology product or service is a cloud product or service. This only leads to confusion about what cloud computing really is. If everything is cloud then nothing is.
VMware has been as guilty as anybody of cloud washing, including even in the marketing titles of its annual show: in 2010 it was “Virtual Roads, Actual Clouds,” and in 2011, “Your Cloud. Own It.” A couple of years ago VMware named its integrated virtual infrastructure stack – virtualized processing, networking, and storage – Cloud Foundation.
Cloud washing? Actually, no.
I think there is an important point in the naming. Virtual machines, software-defined networking, and storage are not a cloud. But they can be a valuable, even necessary, underpinning of a cloud – not a cloud but a cloud foundation.
Some might read the previous paragraph as stating the obvious. Of course a bunch of virtual machines aren’t a cloud. But this is not universally understood. The logic goes like this.
A: External public cloud is made of abstracted infrastructure components.
B: On-premises virtual infrastructure is made of abstracted infrastructure components.
Therefore, C: Virtual infrastructure is a private cloud.
Makes sense but wrong. Yet how often have you heard somebody refer to their virtualized infrastructure as their private cloud? Even beyond cloud washing, official sources can be misleading.
Take, for example, the regular “State of the Cloud Report” from RightScale, which is valuable for understanding the relative growth and use of cloud services. But the report has that flawed logic built right in. To get a sense of the usage and growth of public cloud RightScale counts the enterprises that have application workloads on the likes of Amazon AWS or Microsoft Azure. Among the “private cloud” metrics is the percentage using VMware vSphere.
A straight counting of virtual machines (VMs) hosted on different platforms is one of RightScale’s metrics for measuring the growth of cloud.
More Than a Bunch of Managed VMs
If a bunch of VMs managed on a hypervisor like vSphere is not a private cloud, then what is? Info-Tech’s Own the Cloud blueprint does a good job of starting with first principles. Though it seems tedious to pull out a cloud definition that is more than a decade old, we still see value in beginning with a common understanding of cloud. I’m referring to the definition put forward by the National Institute of Standards in Technology (NIST). We focus particularly on the five essential characteristics of a cloud service:
- On-demand self-service: The ability to access resources instantly without vendor interaction.
- Broad network access: All services delivered over the network.
- Resource pooling: Multi-tenant environment (shared).
- Rapid elasticity: Expand and retract capabilities as needed.
- Measured service: Transparent metering.
These are essential characteristics of all clouds: public, private, hybrid, or community. The difference between public and private is in who provisions and owns the underlying infrastructure – the cloud foundation, if you will. According to NIST, a private cloud it is “provisioned for the exclusive use by a single organization comprising multiple consumers (e.g., business units). It may be owned, managed, and operated by the organization, a third party, or some combination of them, and it may exist on or off premises.”
In order to have those essential attributes noted above, the private cloud cannot end at the provisioning of infrastructure. It requires additional layers of automation and orchestration, especially for on-demand self-service, elasticity, and metering. These tools exist but they are not used in every virtual environment.
When RightScale’s private cloud adoption list is considered through the lens of automation and orchestration, OpenStack, VMware vCloud, CloudStack, and Microsoft Azure Stack meet the bar. VMs being managed with VMware vCenter or Microsoft System Center do not.
Selling the Hybrid Cloud Short
Does it really matter how private cloud is defined? After all, it can be helpful to deflect pressure from higher offices to “get some of that cloud stuff” to point to the vSphere cluster and say, “Look, we already have a private cloud.” Isn’t instantiating a VM on AWS IaaS pretty much the same as putting it on a virtual server host in the data center?
The problem is that limits the potential of the cloud and especially the hybrid cloud. A hybrid is a unity of two or more clouds. You can drop a VM on cloud IaaS, but there are a range of higher-level platform and cloud-native services available on the public cloud. These can be more cost efficient than basic IaaS. If an organization is moving to exploit these services on the external cloud but not the “private cloud,” then the unity between internal and external is limited.
Take, for example, cloud database services. It can be more efficient for a cloud-based application to consume database-as-a-service. But doing so means that databases are architected differently on premises and in the cloud. In cloud migration organizations are faced with a decision to lift and shift virtualized database servers to the cloud or to refactor the data for database-as-a-service. In a true hybrid you wouldn’t play by one set of rules on premises and another off premises.
And Then There Were Containers
Counting VMs still makes sense for basic IaaS, but even here things are starting to change. In a multi-cloud world, the unifying lingua franca is not the VM but the code container. Use of containers-as-a-service on the public cloud has been exploding. Most use some iteration of Kubernetes, an orchestration platform for containers first developed by Google for its cloud service but now widely used open source.
This brings us back to VMworld 2018. In years past VMworld announcements amounted to stupid virtualization tricks – innovating on the VM, how to make it more portable, resilient, and performative. Not anymore. The focus now is operations in a multi-cloud world. There was a lot of discussion of containers and orchestration with VMware’s Pivotal Container Service (PKS – yeah, I know. The “K” is for Kubernetes).
Also on the agenda were items such as tying multiple clouds together with VMware’s software-defined network, NSX, and the fruits of VMware’s growing partnership with Amazon. A major announcement was Amazon Relational Database Service (RDS) for vSphere, which automates database management regardless of where the database resides.
VMware is certainly not the only virtual infrastructure provider to aim higher for private cloud capabilities and hybrid cloud. Rival Microsoft has Azure Stack to run your on-premises infrastructure as an Azure node. This includes support for containers and even serverless functions.
Recommendations
- Aim high in your cloud strategy. Except in a few circumstances, such as replication for cloud-based DR, understanding cloud migration as simply a lift and shift of VMs from your internal private cloud to an external public cloud is too limiting. Costs will be higher than expected for 24/7 operations and you will fail to leverage the true potential of cloud computing (from those essential characteristics listed above).
- Ask if you really need a private cloud. In past research, Info-Tech has segmented the virtualization journey into three stages – Consolidation, Management, and Internal Cloud.
(Source: Info-Tech Research Group, 2012)
These are sequential but not necessary phases. If your organization has a relatively static and stable virtual infrastructure that is performative, resilient, and resource efficient you likely don’t need to implement higher-level cloud automation and orchestration (even if you are 90% or more virtualized). - Get to know containers. Notwithstanding the above recommendation, container orchestration may still be in your future as containers become a more standard form of application code delivery. You can start small with a single server hosting Docker containers, but keep an eye on developments in full orchestration with Kubernetes or equivalent platform.
Bottom Line
If your definition of private cloud is a bunch of virtual servers in the data center, you are likely missing the boat. A real private cloud has an additional layer of automation and orchestration to give it cloud-like attributes. Further, the vaunted hybrid cloud will need uniform capabilities across multiple clouds (including private) that go beyond basic machine virtualization.
Want to Know More?
Public Cloud Services Market Trends and Buyer’s Guide
Own the Cloud: Strategy and Action Plan
Containers and the “End” of Server Virtualization
Containers Survival Guide for Infrastructure