C|Net News has a provocative piece on the rising importance of power consumption as an element in computing costs, based on a recent paper by Google thinker Luis Andre Barroso. Under some not-outlandish assumptions, power costs could soon equal or far outstrip the costs of servers, an important issue for the many firms that rely on massive server farms.

This is an issue with many tentacles.

It will affect competitive balance; Sun/AMD is claiming primacy in improving performance per watt, but Intel, too, has got religion.

It will also affect the proprietary software vs. FOSS wars. Barroso (and Sun) see the solution as multithreaded processors, but, as as a couple of Microsoftees note, this will create major software issues:

But concurrency is hard. Not only are today’s languages and tools inadequate to transform applications into parallel programs, but also it is difficult to find parallelism in mainstream applications, and—worst of all—concurrency requires programmers to think in a way humans find difficult.

Nevertheless, multicore machines are the future, and we must figure out how to program them.

If the adaptive effort must be major, then how will the redevelopment of FOSS be paid for? Will the corporate sponsors of the Open Source Development Lab pony up the huge costs involved without obtaining any IP rights? One can envision the chip makers funding the effort as a way of promoting new chips, but will the other players be willing to contribute to a community effort when they could probably free ride anyway? And is a combination of hardware/software really the most efficient form of industrial organization? It was tried in the 1960s, and the world eventually moved to specialized software firms. But without property rights, the unified approach is the only one possible.

IMHO, much of the general discussion of FOSS, Microsoft, patents, and other software issues has been based on an unspoken premise that software is a mature industry, with its great leaps of innovation behind it, and that public policy should be devoted not to fostering innovation but to turning software into a cheap commodity and to preventing its purveyors from milking products for which they have already recovered the creation costs.

If this premise is wrong, if the situation is one in which massive leaps of creativity are needed, along with the funding for such leaps, then a great many currently popular policy recommendations – such as “no software patents” or “FOSS preferences” – go out the window.

But I am not a programmer, so I would like to hear more from the tech community.

Also, the concern about servers’ power consumption has a bonus for those with a slightly malicious sense of humor. It will be fun to watch the tech community, which tends to be reflexively pro anything that claims the label “environmentalist,” forced to look hard at the realities (and unrealities) of the debate over climate change. One can hope that tech hyper-rationality will change the Kyoto dialogue for the better.

posted by James DeLong @ 11:00 AM | Software

Link to this Entry | Printer-Friendly | Email a Comment | Post a Comment(0)