Automatic provisioning puts zip into data centres

Once the latest updates are loaded, one system after the other is configured, tested and switched online. No wonder manual server maintenance represents a major cost factor for IT departments and their budgets.

Costs are also driven by the complexity of IT infrastructures that are built around a variety of servers requiring specific administration know-how. Many an IT manager would be only too happy if he could cut costs at this point - and now rather than later.

Automatic provisioning products that operate like virtual IT administrators can be an alternative to this dreary manual work. This article sketches the current status of the technology, explains how server provisioning works and shows the extent to which these solutions are actually fit for everyday use.

Automatic provisioning replaces server sitting

Provisioning itself is not really new. Originally, the term referred to the service requisitioning process, starting with input of the request and ending with delivery of the service.

In IT it is normally used as a general term to describe the configuration of computer systems so that a service can be offered. In practice, this now mainly involves installing and activating the software that is needed on an existing server. It also covers the installation and configuration of servers, storage, networks and other components.

Server provisioning tools are now being used to automate these processes. The aim is to lower administration costs through remote installations and updates and to increase application availability - automation lessens the risk of human slip-ups in processes that are often complex and prone to error.

The distribution process to the systems can be rationalized down to a matter of minutes rather than days thanks to the time saved on work and shorter distribution times, which translate into direct cost savings. The virtual IT administrator always has full control, and spots unused resources immediately.

Since resources can be distributed quickly and efficiently to places where they are currently needed, automatic server provisioning optimizes utilization levels. This makes it possible to resolve the dilemma of accurate capacity planning: IT infrastructure extravagances that are geared to meeting any eventuality and a minimal infrastructure strategy are both, in their own ways, money pits.

Deploying resources more efficiently

Automation solutions currently on the market such as VERITAS OpForce 4.0, Think Dynamics from IBM, Microsoft ADS and SMS 2003, Novadigm from HP, Consera and Sun with Terraspring and CenterRun naturally vary in the services they provide.

Some operate in heterogeneous, some only in homogeneous environments, some have very user-friendly GUIs, and so on. Nevertheless, they all work in the same way and use the same core components.

Inventory, configuration and network management
Their first job is to discover the location of a server, how it is used, its configuration and current utilization. This information is the basis of every automatic server provisioning solution.

At the physical level, there are many provisioning products that can trace the precise position, i.e. the room and building, of a server on a company's premises. At the logical level, the solutions offer a higher-level concept, such as server inventory pools. Any server can belong logically to several inventories, acting as production, reserve or maintenance pools.

Other information shows which software release they have installed. Provisioning tools can also detect how a specific server is integrated in the network. Additional information reveals which storage resources are currently in use, above all in a SAN.

Imaging and scripting

There are two ways of distributing software to servers. One of these procedures is known as the golden image. It is based on imaging technology and is characterized by high speed. This method simply creates a snapshot of the entire software that is installed on a source server.

The snapshot is then distributed to the target server, which becomes a clone of the source. Server provisioning tools like VERITAS OpForce 4.0 build on Microsoft's Sysprep technology to create a universal and portable Windows server snapshot, producing a blueprint that can be used to match any server.

This unique capability further automates the provisioning of dissimilar servers and provides flexibility when faced with configuration and hardware changes that would otherwise force the recreation of images.

The second provisioning method is known as scripting. As the name suggests, it uses a script to automate the installation process. Ultimately, this is an automatic image of the person sitting at the server console and responding to reactions from the installation process - in other words a virtual IT administrator.

A fully automatic provisioning product must be able to handle imaging, scripting and a combination of the two. The reason for this lies in the advantages and disadvantages offered by the two technologies, which make it impossible to find a single optimal solution.

The main characteristic of server distribution that is based on imaging is its high speed, although images cannot be created for all applications. One such application is Microsoft Exchange, which does not allow a golden image to be generated because the server is too deeply embedded in the Active Directory environment.

In contrast, scripting can deal with a greater variety of hardware and software constellations but is much slower and therefore unsuited to large numbers of servers or highly dynamic distribution processes.

Whichever technology is used, an automatic server provisioning tool must manage the scripts and images it creates and distributes unaided. An inventory of the images and scripts showing what has been distributed when and, above all, where is absolutely indispensable. Another basic function it must include is bare-metal OS distribution, including setup and initialization of disk storage, for instance as RAID arrays.

Change management

Software is subject to continuing further development. Automatic server provisioning must always have the latest updates and versions available, and provision and configure them automatically; vice visa, it must also be possible to rectify unwanted results.

Automation guidelines

Guidelines are the crucial factors in all automation processes. They specify the events that will trigger automatic resource distribution. The automatic server provisioning product then defines the server that will be used and distributes the requisite software.

This does not mean that inventorying and capacity planning are no longer necessary. A basic prerequisite for speedy provisioning is that a server is actually available. This may be a standby or secondary server. In order for an automatic server provisioning tool to be put to optimal use, the program needs information beforehand on the available resources, in other words the current inventory and its configurations.

Additionally, the change management process must be transparent, meaning that it must be apparent where something needs to be placed in the network, what image or script source will be used and what patches will be deployed after distribution.

No patent recipe (yet)

This technology is still in its infancy, and many toolsets are still having teething problems. Virtually all of today's automation solutions produce excellent results for first-time installations and software provisioning, including setup for operating systems, middleware and the like. But these provisioning mechanisms are only part of the job; still not enough is done to fully integrate all systems and achieve end-to-end monitoring and management.

Additional manual work is necessary when it comes to describing network topologies, suitable portfolio structuring and other data that is not input in upstream tool routines.

Business enterprises also need to meet certain prerequisites if they want to deploy the solutions that are currently on the market.

They can, for example, theoretically automate server provisioning even if they run complex heterogeneous infrastructures. However, if no clear pattern or similarities between the systems can be detected, a great deal of time will have to be invested at the start in order to inventory and systematically describe the resources that are available for provisioning.

Software cloning normally also means that the systems involved must have identical or at least very similar hardware configurations. In actual fact, this is hardly ever the case, with the result that IT administrators have to handle different systems across several server rooms. Image management may therefore turn out to be a major challenge.

Any company wanting to take advantage of automatic server provisioning will either have to invest large sums of cash up front to homogenize its IT infrastructure or - and this is the cheaper way - wait for the next upgrade and, with it, the next episode in its natural replacement cycle.

Integration

There are already a number of tools on the market that handle scripting and server distribution, inventory monitoring and detection of performance problems. What is missing, though, is a solution that integrates all these tools, especially across vendor and product boundaries. There is hope that Web and Grid Services will improve this situation over time. But regardless of efforts like DCML and OGSA, many vendors still show too little interest in collaborating, with the result that progress is, at best, slow.

Summary

The ultimate goal of automated provisioning is to automate all maintenance routines in the data centres along with routine tasks. At the present stage of development, however, it tends to be more of a support tool. Deployment of automatic provisioning systems cannot replace IT administrators but can take considerable work off their shoulders.

Functions that work well are the automation of initial operating system and application installations as well as automatic upgrades and patch deployment. The programs are also able to configure storage systems and networks. To reap the full benefits of automatic provisioning, it is not only single components that should be provisioned: all resources throughout the enterprise that are not used efficiently at present need to be located and then redistributed.

Undoubtedly, the solutions on the market today are an important step in the right direction. Automatic server provisioning provides consistency and lessens the potential for server performance problems by quickly discovering, analyzing and reporting differences between any number of provisioned servers. Administrators can therefore detect and correct server configuration drifts extremely quickly. The result is more efficiency, dynamics and reliability in the data centre.

The bigger the IT infrastructure, the more obvious the cost benefits to be gained from smooth operations as a result of task automation. This also applies to growing companies, which need to be able to adjust their servers to enhanced requirements on an ongoing basis. Distributed server provisioning and application configuration management are elementary components of a utility computing strategy aimed at improved server utilization and greater flexibility in the server infrastructure at lower cost and with reduced complexity.

Automatic server provisioning is the ideal way for enterprises to harmonize their IT infrastructures with changing business needs.