Datacentre checklist

Regional uptake of datacentres continues apace, but remains complex. ACN highlights some of the key areas to focus on in the market.

For all bar the most dedicated of IT professionals, datacentres remain a dense and often uncharted area, filled with complex terminology and vested interests.

To complicate matters further, current interest in datacentres has driven evolution of the sector into overdrive, with new products and vendors entering the market every month.

Here, ACN runs down some of the key areas to look at when planning a datacentre deployment or upgrade, and highlights a few of the latest developments or major players in each sector.

Cabling and racks It's not glamorous, but it is essential - and as much as people might like to dismiss the claims of cabling companies as so much marketing waffle, the choices organisations make about cabling today will be with them for many years - or decades - to come.

Most enterprises will now be familiar with the issues surrounding 10G cabling - and certainly in the limited environment of a datacentre, will be happier to invest in the most current cabling systems, which they might not choose to deploy to individual desks.

But datacentre managers also need to look at issues around managing cabling - especially in more complex datacentres. While cable management sounds like a mundane detail, the impact an effective management system - and procedure - can have on the day-to-day maintenance of a datacentre is dramatic.

Many companies offer up methods of dealing with an unruly mass of cables - cables which are growing increasingly thick, as speed demands dictate more shielding and higher-gauge copper cores.

US infrastructure giant Leviton alone offers horizontal, rack and frame and hook and loop cable management, a fibre raceway system, and its trade-marked Versi-Duct management system - and other cabling players such as Systimax, Nexans and Panduit all pitch equivalent offerings.

The boring-but-vital label also applies equally to racks - and again, a moment's thought would suggest that investigating the units that will house the vast majority of equipment for the coming years would be a sensible move.

Racks - or enclosures, as vendors prefer to call them - can affect everything from cabling management to thermal efficiency, power supply to ease of maintenance.

Not collapsing under the weight of half a dozen servers is also a key requirement for racks - and, levity aside, is a strong argument for thoroughly evaluating a potential purchase in advance.

Many of the same cabling vendors also offer enclosures designed specifically for modern enterprise datacentres - but the key element here is planning exactly what will be going into a datacentre, and how the centre is likely to expand in the future. This leads us neatly on to...

Servers

Think datacentres, think servers. Lots of servers.

One of the most enduring images of a modern datacentre is a rack - sorry, enclosure - filled with shiny new blade servers, probably being hot-swapped at will by cheerful-looking technicians. Hopefully the technicians are cheerful, but the image of blades-in-everything is more marketing dream than enterprise reality.

Perhaps one of the most surprising points about the modern server market is the persistence of the mainframe. Although sales have declined - rapidly - in recent years, there is still a significant market for traditional UNIX servers, as well as the cheaper and more common x86 varieties that have gained ground in recent years.

Mainframe stalwart IBM is certainly talking up its mainframe offerings, not only making them an integral part of its new - and fashionable - cloud-computing initiatives, but also releasing powerful new models, such as the Power 595, based around 32 5GHz Power 6 dual-core processors, against which eight-core offerings from x86 vendors seem weedy in comparison.

Big Blue has also rationalised its server offerings, bringing both its System p and System i lines together under the ‘Power' moniker - hopefully making the process of choosing a server somewhat more straight-forward.

Seeing as IBM had around 1000 different options for its server ranges - somewhat more than other vendors, but not necessarily dramatically more - server buying for enterprises is never going to be a wholly simple process.

What is now certain is that the CPU arms race between AMD and Intel has led to dramatically improved energy efficiency and performance from processors, and this movement is now extending to the rest of the server.

Buyers should be looking for the most modern available chipsets and server architectures - and be paying careful attention to overall power consumption figures and performance per watt.

But once the servers are in place, they still need to communicate with everything else: bring on...