Many cloud providers increase ROI and ensure business success by practicing complete in-house development and management of all platforms. Unfortunately, for all but the largest enterprise IT environments, this is often too great an undertaking. But IT departments can still follow this successful cloud business practice by taking advantage of commercial management tools and keeping an eye on long-term costs.
Amazon built its own provisioning system for
Owning the whole stack allows cloud providers to control everything and to keep knowledge within the organization. Large-scale private cloud business success relies on this control and knowledge retention, and enterprise IT could benefit from the practice as well.
Building a large-scale infrastructure
Read more lessons from cloud IT
What enterprise IT can learn from cloud business models
Approaching infrastructure design like a cloud provider
When enterprise IT builds a large-scale infrastructure, it also wants end-to-end control of every element. The total project cost consists of not only buying hardware and building the platform, but also of operating the platform for its lifetime. Years of TCO studies have revealed that platform operational costs are a multiple of the initial hardware purchase cost, a fact that many IT departments neglect to consider. When you make purchasing decisions, especially for automation tools, keeping long term costs in mind benefits your data center more than an initial 5% discount on the hardware.
Cloud providers build more than just virtual machines: They must also build and configure servers, networking and storage. All infrastructure components link together, which means the build processes must also be linked together to expose and minimize any bottlenecks. Cloud providers must practice proper provisioning, and the same is true for any large IT organization.
The bigger picture
Project-based procurement and deployment doesn't cut it for public cloud providers. Projects will end before the lifetime of the system they deliver, meaning the project objectives have the potential to run against the system objectives. For example, tying the build process to a specific model of physical server is a design decision that makes the project or management tool easier to deploy, but harder to operate.
Looking beyond the build process to the life of the platform shows you that the platform should be designed to accommodate different server models and, ideally, automatically manage the differences without any operational processes. Enterprise IT needs to think beyond the current project and make the best decisions for long-term business practices.
Tools at your disposal
Fortunately, you can now choose from commercial products with the scope and range required to deploy and manage an enterprise infrastructure. Unlike Amazon and RackSpace, most data centers will benefit from purchasing these products rather than building them in-house. Most cloud providers, on the other hand, will customize open source products, such as OpenStack, to fit their needs. Buying an automation platform from a vendor is similar to project-based delivery. Vendor interests may not match the platform objectives.
If your platform isn't large enough to warrant in-house development, then you should measure each automation tool against your platform to find the closest match. You should also budget for modifications and to manage the lifetime of those modifications with your own developers and operations teams.
As enterprise IT deployed virtualization, it broke down some of the barriers between silos. The server team works more closely than ever with the storage and network teams. Data centers with teams that do not work so closely together fail to reap the full benefits of virtualization. The next level of this team collaboration would be a single unified tool for provisioning and management.
This was first published in July 2013