Space: The data-centre frontier?
Published: 16 Jul 2007 10:33 BST
The data centre is an inevitable consequence of our use of IT to gain competitive advantage in business. Use IT properly and your business will grow; when it does, you'll need more equipment to cope with the demand. Repeat this process and you end up with a large room full of IT equipment supporting your entire business. And, if you're in a city, where office space is in short supply, you have a problem.
While servers have got smaller and more powerful, we're now doing far more with them than we ever used to. The days of file-and-print serving are long gone — now we have ERP, CRM, messaging, intranets and all sorts of other server-based business systems, not to mention the databases that most of these applications rely on. This has meant a proliferation of server hardware and other supporting infrastructure, such as networking, power protection and air conditioning.
The net result is that your IT infrastructure not only costs you money, both for the equipment itself and its running costs, but also consumes a valuable resource — floor space. For companies that need to be located in city centres to be competitive, the shortage of suitable office space makes having a physically large data centre as much of a liability as a benefit.
Thus, if you're in charge of a data centre, you're faced with a dilemma: your users are demanding more from you, whether it's capabilities or capacity, but you don't have any more space to achieve this. Until Doctor Who lends you his Tardis to put your data centre in — or some other fairy story comes true — what can you do?
Tackling the issue of floor space
Hardware vendors were quick to realise that this problem needed addressing, and that they could make money from solving it. Slimline servers, blade servers and other compact machines appeared on the market, cramming more processing power into a small space, allowing businesses to expand data-centre capabilities without expanding their footprint.
Multicore processors can run several processes at once while still only taking up a single socket on the server's motherboard. Before this happened, servers — even rack-mounted servers — were based indirectly on the design of desktop PCs, and were laid out to make assembly easier, with the space they took up not being an issue.
Blade chassis vary in size and configuration, but typically allow you to pack a couple of hundred processor cores into a single rack, not counting any storage. Not every data-processing job is suited to running on blade servers — databases are probably still best served by a full-sized machine — but any job that needs more processing power than it does disk or network access is going to work well on blade servers.
Making better use of the hardware can reduce costs wherever a data centre is located
The problem with this approach is that it brings its own set of issues with it. Processing data uses power and, in the absence of superconducting chips and wires, produces heat. If you have more processors, and processors with more cores, in the same space, you have a greater heat density — more energy in the same amount of space. There's also less air in your racks to absorb and carry away all this heat, which means you need better ventilation and cooling — more equipment taking up space in your data centre. It also increases your power consumption as bigger, faster fans are needed to pump air past components that need cooling.
Different server applications are normally run on different pieces of hardware not for capacity reasons but for security. Although modern server operating systems do a good job of preventing one rogue process interfering with any other jobs that may be running, they're not perfect. Runaway processes can still hog all of a machine's resources, so that an email system could slow down an ERP system if they were running on the same machine, even if most of the time they could co-exist quite happily. It would be better if machines could be split in half, and not allow any one job to take more than its fair share of resources — and that's precisely what virtualisation allows you to do.
Making better use of the hardware can reduce costs wherever a data centre is located. The advent of virtualisation breaks the tie between one application and one physical box. Multiple virtual servers running on the same hardware allows each application to run on in its own operating system instance while avoiding having infrastructure lying idle because that system isn't being heavily used. It can also help with other data-centre issues, such as power consumption and disaster recovery, since a "machine" can now be moved as easily as a file can be copied, albeit a large one.
Depending on how well hardware requirements are managed, virtualisation can be the single biggest factor in saving space in a data centre. VMware, one of the leaders in virtualisation technology, says its customers...






