Cloud Factories: Energy and Information

September 23, 2012

The New York Times has, today, published the first article in a series of articles about the power consumption in data centers and other cloud computing infrastructures.  The consumption numbers are rather huge and the efficiency numbers are rather small.  As you know, I’ve been thinking about the relationship between energy and information for some time now, and spoke about it in Cambridge in July.  One of the key points to be made is that information is physical: “bit is it” to quote a famous physicist who also symmetrically said “it is bit.”  As noted in the ISIT 2012 paper as well as in the ISIT 2008 paper, the tradition of separating the study of energy from the study of information goes back almost a century.  I wonder what effect the academic separation between power engineering and radio engineering has had on popular understanding.  Indeed, the NYTimes article points out that:

With no sense that data is physical or that storing it uses up space and energy, those consumers have developed the habit of sending huge data files back and forth, like videos and mass e-mails with photo attachments.

And so about three-quarters of data is created by ordinary consumers.  Going into the history of this some more, I think I’ll quote a bit from a book called Grammatical Man by Jeremy Campbell, which I had first read many years ago.  On p. 193, he says:

Norbert Wiener made it clear at an early stage, however, that there is a critical distinction between power engineering and communication engineering, and this distinction must be grasped if we are to begin to understand how the nervous system works.  A television transmitter, Wiener said, may need large amounts of power to do what it is supposed to do, but it is first and chiefly a device for sending messages.  A dentist’s drill, on the other hand, may use only a tiny fraction of the power needed to drive the transmitter, but the prime consideration in designing the drill is the energy it consumes.  Wiener, no shrinking violet where his own reputation was concerned, claimed credit in his memoirs for first alerting the scientific world to the importance of this distinction, and for showing that control devices, like the ones used for aiming antiaircraft guns at German planes, were as much a part of communications science as the telephone or the radio, even though their function might be to move an object as heavy as a large gun.

Going on p. 195 to say:

One especially important difference between energy and information is that the first is subject to the laws of conservation, while the second is not: information can be created or destroyed

and then on p. 270:

Even more surprising, Aristotle gives a hint of the peculiar asymmetric relationship between energy and information, a relationship which was brought to light in modern science only when the full implications of Maxwell’s demon were understood in the twentieth century.  The demon needs enormous quantities of information about the particles in the dark chamber of gas in order to reduce the entropy of the gas by even a small amount.  In other words, it is relatively easy, though not costless, to convert orderly energy into information, but difficult and expensive to transform matter into a more orderly state by the use of information.  Aristotle showed that he had a general grasp of this inequality.

Although not a central part of the talk in Cambridge, my mention of results from the thermodynamics of computation seemed to elicit quite a bit of discussion.  Perhaps counterintuitively, several of our past/present colleagues at IBM Research have established that mathematical work does not actually require energy, though there are all kinds of caveats.  As I had once noted in my doctoral thesis:

Physical mechanisms proposed for reversible computing such as ballistic computers, externally clocked Brownian machines, and fully Brownian machines, however require that the system have no faults (contrary to the model in Chapter 5), have infinite storage capability (contrary to the model in Chapter 4), and operate arbitrarily slowly (contrary to the goals in Chapter 3).

Do you think the series of articles in the New York Times will also bring up the thermodynamics of computing?




  1. …Links…

    […]I am no longer certain the place you’re getting your information, but great topic.[…]…

  2. On the Google homepage today, there’s a link to “The Story of Send” which includes stops to smart heat and an efficient server floor. It is kind of interesting that they’re highlighting “the innovative ways [their] data centers … conserve energy” when the expectation is that the link will only show you how an email travels through the network.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: