mercredi 8 mai 2013

Not Your Grandmother’s Server: Why Faster Upgrade Cycles Make Sense

A lire sur:

May 6 2013
In technology, nothing stays the same, or even looks the same, for long. Today’s laptops are thinner than paperbacks of short novels and come with huge screens and tiny ports compared to the models of 10 years ago. Desktop PCs, when you can find them, have small cases and flat screens compared to the big beige boxes and monster CRT monitors of yesteryear. Data center servers are no exception. Where once they stood on central plinths they now stand side by side and tightly packed in racks, occupying a fraction of the floor space they once hogged and often designed as blades with not much more than motherboard, connectors and cooling.
The changes in servers are not only cosmetic either. Servers today have been completely remodelled and redesigned to address the new realities of data center computing where more and more performance is needed but speed can’t come at the price of high energy consumption or heat dissipation.
Of course, servers today are faster and, despite previous expressions of concern that the rate of improvement could not be maintained, they continue to follow and usually exceed Moore’s Law, multiplying in performance through new silicon designs and incredible levels of integration that allow multiple CPU cores to reside in one package, together with closely coupled graphics and memory.
Where once ‘bang for buck’ (that is, performance to price) was the only metric, today, performance per watt is a critical KPI. Server designers and component makers do everything possible to make hardware that sips at power and hence keeps energy bills low and reduces heat dissipation. If they don’t, buyers face a nasty cocktail of high power charges, falling reliability and additional costs for systems to cool data centers… as well as a nasty red cross in the corporate social responsibility or ‘green’ checkbox.
Another game changer has been virtualisation software. Where previously it might have been common to see servers running at 10 or 20 per cent utilisation levels, today they can be run nearer their maximum ceiling and workloads can be consolidated onto fewer boxes. At the same time, virtualised servers can be used to run test-bed environments and to trial services in safe environments. This adds up to less data center space being taken and simpler power management and IT administration.
The new servers are very different to the old servers then, and they offer a very different buying proposition and encourage and a faster procurement cadence. Where once a sensible IT director might have been content to ‘sweat assets’ by running servers for years and years today it makes sense in many cases to upgrade servers on a regular basis to take advantage of greater compute capacity, power efficiency and server consolidation.
It’s worth taking a careful look at the new generation of servers… and it might very often work out that spending now means saving in the future.

Aucun commentaire:

Enregistrer un commentaire