The idea behind HP's OpenView Automation Manager is a lot simpler than the technology that underpins it, but you might not guess that from HP's jargon-heavy announcements about 'adaptive enterprises' and 'linking IT to business processes'. So we asked John Schneider, the company's VP and general manager for change and configuration software, to explain how it actually works.

"It's like pooling, or dynamic provisioning," he says. "Typically people build their systems with headroom to cope with workload peaks, so a lot of their hardware is partially utilised, it's the same for your backup and test systems. This means you can run your servers hotter, keep fewer spares and move them around as needed."

For all the talk about business agility, what it boils down to is doing the same for a bunch of servers that a multi-tasking operating system does for a single server, i.e. allocating processing resource among different applications according to their priorities and needs.

This can be done on a scheduled basis, for example if you have systems assigned to a task that's only needed from 9-5, the bulk of them could be reallocated overnight for use as backup servers. Or it can be automatic, so OpenView might spot that processor utilisation has passed some predefined trigger level - say 75 percent - and step in to provide that application with more servers, taking resource away from a lower priority application if necessary.

The reality is lots more complicated though - just think how long it takes a competent sysadmin to build and install a server, put it on the network and allocate it to the load balancer, even using scripts and standard software images. Now imagine doing that automatically, in response to an immediate requirement for more power.

"Change management is the part in the middle that takes the triggers from BPI [Business Process Insight, an OpenView component], analyses them, kicks off the workflow to figure out the new desired configuration, then SIM [Systems Insight Manager, also part of OpenView] deploys that. Then we keep checking that the configuration stays right," Schneider says.

Some of these tasks can be done with scripts, of course, but HP argues that these are usually not good at handling errors, which makes them tough to truly automate. In addition, they need updating to allow for service packs and suchlike.

The key component of Automation Manager allowing it to provision automatically is Radia, which HP acquired when it bought software developer Novadigm earlier this year. This lets the configuration manager define a desired system state and then implement it. In future it will link to SRM software too, so it can dynamically allocate storage as well.

Schneider makes the point that Radia does not simply write a whole new system image - it looks first at what's there already and only adds or removes software as needed. It can also merge states, for example to deploy two applications to one server (with or without virtual machines), and it can maintain desired states, rolling back a damaged system to the way it should be.

Defining those states is just an extension of what sysadmins do today, Schneider says. "One way or another you are going to define configurations - in your head or in a script. We take a reference architecture - a fully patched Linux system, say - and use that as a model."

Nor does it have to be an HP system, he stresses. "For example, you can use our solutions to work on an IBM server running Linux - perhaps you might have a mixed environment as a result of acquisitions. We want to win customers from having better products, not by taking them captive."

The other thing that needs modelling is the business, so the right triggers can be identified. Schneider suggests this will already be underway within many organisations as they deploy IT service management schemes and define SLAs for their key services, or carry out business process re-engineering. "They're going through modelling as part of their evolution," he says.