What
everyday IT
can learn
from
hyperscale
What do turbochargers have to do with data centers?
Turbochargers became popular in the early 1980’s with the Saab 900 and the Buick Grand National (remember that beast?) and now are expected to be in over 40% of light American cars, and 85% of European cars by 2016. In the 1990’s anti-lock brakes (ABS) became a common car feature, and after 2007 became mandatory in new cars in the EU.

Where did these now essential technologies come from? Airplanes.
The forced induction of turbochargers was necessary at high altitude, because the reduced air pressure resulted in significant power loss. When applied to cars, turbochargers provided sufficient power enhancement to enable smaller displacement engines to replace larger engines, saving on weight and fuel consumption.
Anti-lock brakes were developed because threshold braking was nearly impossible for planes. When applied to cars, ABS has proven to significantly improve safety for most on-road situations.

When it comes to IT infrastructure, one may wonder what can be learned from environments with thousands of servers and hard drives. Well, much like the capabilities derived from 600,000 pound airliners, hyperscale data centers have much to teach.
Keeping applications running in large-scale hardware environments forces new behavior and new technology. With the same number, and often fewer administrators, there is no time to stop everything to replace failed drives, servers, or even racks of servers. Application clients have to be rerouted immediately, and hardware replacement relegated to a regular maintenance task (you deal with them when you can, potentially on your next scooter run — see video for details).
True data resiliency, even in the case of rack and site failure, can’t be a complex dance of local RAID, asynchronous replication, offsite backups, manual steps and as much as 400% overhead. Data must be identifiable at a much more granular level than half-empty volumes and data protection must incorporate notions of local and remote natively. This enables better protection than the independent mechanisms of today, with much greater efficiency.
Applications can’t be thought of in isolation. Instead of buying a separate control plane, data controller, and persistence layer for every new app or expansion, the storage infrastructure must continue to scale capacity and performance linearly as a single platform.
Much like turbochargers have improved the basic way cars go, and ABS has enhanced the basic way cars stop, technologies from hyperscale environments, evinced by software defined storage products, can drastically improve many IT infrastructures. Don’t be fooled by the heritage.