One of the themes of this column for some years now has been the ascent of what might be called the Internet's own operating system - GNU/Linux. For GNU/Linux not only runs a decent wodge of all Internet servers - notably for Web and e-mail - but is a direct result of the distributed development that the Internet made possible. In the light of some interesting developments over the last month or so, it seems a good time to review how GNU/Linux is faring.
One straw in the wind was the release of a report by the Butler Group called Operating Systems - Winners & Losers in the Open/Proprietary OS Market. In fact, its view is unequivocal: GNU/Linux will be a "clear winner" here, along with Microsoft's .net.
It is something of a watershed that a major IT analyst is prepared to stick its neck out to this extent. Historically, analysts tend to stay close to the status quo, with only marginal fluctuations either side of the average; the Butler report is therefore a key moment, and I expect other analysts to follow in their turn.
Of course, analysis and prediction is one thing: what is really happening often quite another. But some other recent events have shown that the GNU/Linux market penetration is not just getting deeper, but broader too.
GNU/Linux is already strong in the mainstream business server market - which is why it is no surprise that the Butler Group report makes its bold prediction here, rather than elsewhere. In fact, because of the innate conservatism of analysts, Butler's mildly daring leap into the future is more a confirmation of the fact that GNU/Linux is now so well-established in the business server market that nobody even questions the fact.
The other recent events are interesting because they have concerned areas far removed from this open source heartland. The most striking example of GNU/Linux's success at the top end was the announcement that the world's most powerful computer, IBM's Blue Gene/L, would be running not some specialised, high-performance custom-built operating system, but basically the same code that started life as a hack in Linus Torvalds' bedroom in April 1991, running on a single 386 processor with 8Mbytes of memory.
Blue Gene/L, by contrast, will have 65,536 processors and 16,000Gbytes of memory. It is expected to operate at 200 teraflops per second (200,000,000,000,000 floating point operations per second), which IBM claims is larger than the combined computing power of the top 500 supercomputers in the world today.
Blue Gene/L is actually a baby version of the main Blue Gene machine, which
is designed to run at 1 petaflop per second (1,000,000,000,000,000 floating point operations per second). And if you're wondering quite what all this power is for, it's to simulate a protein chain as it folds into the three-dimensional shape that determines most of its salient properties.
Although petaflop capabilities might seem far removed from the reality of everyday business computing, it is worth remembering that when the IBM PC was introduced in 1981 its processor used 16bits internally, and ran at just 4.77MHz, compared with PCs today that use 32bits and run at 2GHz or more. The point is, research machines of today become the desktop norm of tomorrow, and GNU/Linux has already planted its flag.
Still, it is true that currently only IBM can afford to use GNU/Linux in this kind of way. But at a slightly lower performance level there is an alternative that any company can buy: a cluster of standard PCs running GNU/Linux. Such systems provide supercomputer-level power without the price tag.
One indication that these systems are now off-the-shelf products is the fact that Dell Computers - the commodity PC maker par excellence - is making GNU/Linux clusters. Dell even envisages selling these systems directly from its Web site like any desktop or portable. High-end GNU/Linux computing has clearly arrived.
This was first published in November 2002