Next-generation data centers are supposed to be hotbeds of automation with on-demand everything. In this ideal, system administrators are almost obsolete, and the data center runs “by itself.”

This vision brings to mind a sarcastic T-shirt seen recently on a system admin: “Go away or I will replace you with a small and efficient script.”

Reality rudely intrudes into this vision. For one thing, operating systems are composed of thousands of components with tight bindings and a maze of interdependencies. Each of these components, from the user-administration system to the mail server, may have a completely different management interface - some scripted, some GUI-based and so forth. In this kind of environment, it is very difficult to create an automated provisioning and deployment tool that works consistently.

Fortunately, all operating system vendors have identified provisioning as a key operational headache for their customers and have been building increasingly sophisticated tools to reduce the time and effort needed to deploy a server. Also, several vendors have further extended the basic operating system tools, turning them into sophisticated provisioning tools.

There are two general approaches to server provisioning: imaging and scripting. In the imaging approach, a server or application component backs up to a “gold” image, which can then be reproduced on a target server. Because the image is restored in one go, this process can be very fast, lasting only a few minutes. Scripting, by comparison, deploys the software using the standard installation process, with the administrator’s manual input automatically recorded and replayed for every installation. Since this process uses the generic installation program designed to provide a myriad different possible configurations, it is a much slower process. Furthermore, the resulting installation may require additional updates and patches. Though slower, the scripting method provides more fine-grained control over each configuration.

In practice, most provisioning systems often combine both methods into a hybrid approach, to gain the best attributes of each.

Yet a third approach is to extend the imaging method by storing a gold image and, in addition, “deltas” for each desired configuration. The deltas are the differences between the desired build and the base image. This approach allows administrators to have hundreds of virtual images (a single baseline and hundreds of small deltas) without having to store hundreds of large baselines.