It is easy to imagine IT managers bursting into hysterical giggles at the suggestion that datacentres should be switched off at night – many companies have only just got their server rooms working properly, and would be unlikely to risk any unnecessary reconfiguration.

Also, most large datacentres are owned by multinationals that operate 24/7 and need to stay "on" at all times. And given that many senior executives tend to work as close to 24/7 as physiology will allow, it's crucial to ensure the email servers that let them access and respond to messages, anywhere, anytime, remain in constant operation.

But in some ways, powering down datacentre servers makes sense, and not just from an environmental point of view.

For a start, servers that are not running cannot be hacked into, even if the hacker gets past the firewalls and other network defences. Nor can data left in memory get corrupted by some rogue transistor cell deciding to self-destruct out of spite.

But the potential management benefits come unstuck when you consider that many datacentres contain hundreds if not thousands of servers, and the idea of restarting that lot every morning would be any sane IT manager's idea of hell.

New technology can help, though. Virtualisation, coupled with a comprehensive infrastructure management system, can create a framework that gives IT chiefs the option to power-down certain systems.

Despite the advantages that virtualisation offers, many enterprises still run infrastructures built on the inefficient one-server-per-application model. However, a more flexible approach is being advocated by a number of vendors. HP, for example, is promoting its NonStop high-availability server as a "state database engine" that can underpin an almost limitless range of cheap Linux servers. The NonStop engine holds data about the current state of every server so that, if one fails, its state can be reloaded onto a spare machine, and the status quo is maintained.

And then there are application virtualisation tools such as Softricity's SoftGrid technology, which pushes just the application code needed to run any one task off the server and out onto the desktop PC. That way, there is no need to load the whole application in either location. Though SoftGrid is targeted at Microsoft desktop PC environments for now, the technology should in theory scale to a state engine/racks of servers model in the future.

Combined, these technologies could enable IT departments to save money on both energy and hardware, while providing them with far greater operational flexibility.