For most of cloud computing's history, businesses have struggled with the question of whether to implement cloud themselves or acquire public cloud services. Mostly, and in particular for larger enterprises, the answer has been to build an elastic mixture of the two, so private cloud deployment has always been important.
Private cloud remains important today, but the landscape has changed. Cloud planners must know how they're getting to the private cloud; where they see applications evolving; and perhaps most crucially, where public cloud services are headed so they can balance issues of private cloud planning, orchestration and integration.
Users typically approach a private cloud deployment from one of three perspectives:
- Attempting to replicate public cloud services to facilitate hybridization to keep critical applications on-premises
- Expanding virtualization commitments by improving the efficiency of their resource pool
- Working toward a new cloud-friendly or even cloud-centric vision of IT
Each of these perspectives influences the opening salvo of a private cloud plan.
Using private cloud for hybridization, virtualization extension
Companies that want to use private clouds to replicate their public cloud services in-house must look for cloud-stack architectures compatible with the management interfaces of the public cloud services. Eucalyptus is the most common Amazon Web Services (AWS) private cloud partner architecture. Users of other cloud services should look for a similar private complement, such as Microsoft's own tools for Windows Azure or OpenStack for Hewlett-Packard or IBM.
If a company aims to use private cloud to build on virtualization, the operative question is which hypervisor tools are already in use. Most companies will use either VMware or the KVM hypervisor. For the former, VMware's own vCloud approach is likely the best way to extend virtualization because it preserves all existing management tools and migration features. For KVM or other hypervisors, both OpenStack and Apache's CloudStack will support and extend virtualization. The best choice will then depend on the other factors.
For both the public-cloud-extension or virtualization-extension users, the biggest cloud planning issue is likely orchestration. To reduce operations effort and errors, it's important to pick appropriate deployment tools. The quality of orchestration tools is so important to cloud success that it's sometimes justified to make a cloud software selection based on orchestration support. That's particularly true for users who plan to use hybrid clouds or multiple public cloud providers. Some public cloud orchestration tools will provide aid in private cloud deployment, and there are DevOps-as-a-Service tools that can orchestrate across multiple clouds, public or private.
Fitting private cloud into the future cloud-native landscape
As familiar as the first two models of the private cloud are, they probably don't represent the way of the future. Both models presume applications will be ported to the cloud largely unchanged, and it's becoming clear the optimum cloud application is one designed for the cloud. Such an application has an interesting implication for private cloud planning -- it might eliminate the need for it.
Modern apps designed by software-savvy companies, such as IBM, Microsoft, Oracle and SAP, are highly component-ized and designed to be orchestrated. Private cloud software is most useful when the applications are not designed for sharing and orchestration, and migrating to software practices that build components instead of applications will allow businesses to deploy on standard multitasking OSes (Linux, Windows, etc.) and still enjoy high levels of flexibility and resource elasticity. Even today, cloud bursting and failover applications are as likely to use standard IT platforms to connect with public cloud services as they are to require private clouds.
Software companies also provide horizontal and vertical platform services to developers, and in some cases, custom appliances -- Oracle Database Appliance and SAP HANA, for example. These services can be hosted inside a data center with zero private cloud commitment, and used both by traditional data center and public cloud applications. Platform services provided by the public cloud providers -- AWS in particular -- are also designed to be integrated with traditional data center applications as easily as with private cloud apps, and so they reduce the need to deploy private cloud technology at all.
These last two points mean that companies like Oracle, which has been criticized for being late to the cloud, may in fact be right on schedule. If apps are designed for deployment and orchestration in standard OS platforms without the need for hypervisors or cloud stacks, the pendulum of resource optimization will swing back to software companies that can use their platform software and application tools to quickly migrate users to these new app design principles. Other private cloud players could then have to catch up themselves.
Companies should not assume they must adopt private cloud technology to adopt the public cloud or to hybridize with it. The private cloud option is most likely to evolve from either current data center virtualization commitments or from a desire to move legacy applications or components between the data center and the public cloud. In either of these cases, the goal is to pick private cloud technology that facilitates the migration -- and in many cases it will be orchestration/integration or so-called DevOps tools that will determine what the best option is for you.
In the long run, application design will accommodate hybrid cloud use and even things like cloud bursting and failover without private cloud deployment. However, these advances will not reduce the need for integration and orchestration tools. The more components make up an application, the more dynamic the component relationships have to be in order to support each worker optimally. And the more dynamic the resources needs are, the more automation is needed to deploy and sustain the application. This means that DevOps or orchestration could very well be the element of private cloud technology that has the most, and most enduring, value to users.
About the author:
Tom Nolle is president of CIMI Corp., a strategic consulting firm specializing in telecommunications and data communications since 1982.