If only all computers were as energy-frugal as the $100 laptop, the brainchild of the ‘One Laptop Per Child’ organisation (OLPC), which aims to put affordable computing power on the desks of children thoughout the developing world.
While children in Libya or Ecuador are busy hand cranking their machines or putting them out in the sun to charge the battery, data servers all over the ‘civilised’ world are soaking up electricity as if there’s no tomorrow. Whether there will actually be a tomorrow is a topic for another conversation.
An article in the Economist highlights the massive worldwide growth in computer data and in the technical infrastructure required to store it. 4% of all electricity consumed in America is used to power the “big servers that big firms use to store big data”. The number of servers in the world has doubled since 2000. Each new generation of servers consumes more power, generates more heat, requiring more air-conditioning. As much as 70% of the power used by a typical firm in a service industry is due to its servers, according to Tom Kelly of Logicalis.
While IT R&D slaves away to develop hardware that will consume less power, users have washed their hands of their responsibilities to manage their mushroming data effectively, the article claims. Once created, data is stored away on a disk on a server in a remote location, and quite probably never accessed again. In the sunset world of the large mainframe, Hierarchical Storage Management (HSM) has been the norm for decades. Tools such as IBM’s DFSMShsm offer automated policy-driven management for files, deleting after a set period of inactivity, migrating to a cheaper medium for long term storage, and so on. It seems the fight to extend that frugality and rigour to the server environment has been lost.
The placement of a PC on every desk was acclaimed as an act of liberation for users and their data, but came at the price of decentralising responsibility for data management. That duty was never accepted, with IT personnel doing just enough fire-fighting to keep servers running by plugging in more disks. Unless HSM can be implemented across networked servers with the same degree of control as on mainframes, server growth will continue to soar. It may well be too late.