Users should start accommodating "unreliable" technologies to deliver better computing performance, analyst Gartner has advised
Explaining the idea behind "unreliable computing", Martin Reynolds, vice-president at Gartner Dataquest, said, "This seemingly paradoxical term refers to the ability to assemble devices that do not always function correctly into entirely reliable systems, through redundan- cy and self-correcting designs.
"This philosophy will extend from the smallest devices to the largest systems."
The idea is to build IT systems that can cope with minor failures. One example is the design used by HP in its cross-bar architecture to accommodate faults in the circuits.
Although HP believes the technology will not be available in commercial products for at least 10 years, Gartner said HP's approach to computing would become pervasive in the coming decade.
Reynolds said the value of this approach lay in manufacturing systems' ability to produce 10 or even 100 times as many unreliable devices as reliable ones.
He said the increasing reliance on smaller devices was causing performance problems, because such devices are more affected by external influences such as radiation and electrical noise.
Unreliable computing also applies to large systems. Google, for instance, used low-cost components to assemble a very large computing infrastructure.
Gartner estimated that this unreliable computing approach allowed Google to deliver resources at one-tenth the cost of a typical server infrastructure.
The key role of excess capacity
Gartner has predicted that unreliable computing will become pervasive throughout IT architectures in the coming decade.
It said businesses should plan to "scale out" architectures - increase IT capacity to a level above that necessary - wherever possible. Scaled out server architectures represent an evolutionary step toward making effective use of unreliable computing, it said.
Companies should also explore virtualisation technologies as a way to transport applications between systems. These will be critical to adapting applications to the new infrastructure, Gartner said.