The problem is that the waste heat from server is pretty low grade; google runs their data centers hotter than most, and they report a waste heat temp of about 50 C. I would guess that the water they use to cool the air thus gets heated to at most 45 C or so. So it's difficult to use efficiently or economically. At least over here, district heating systems have an input temperature around 100 C (in some cases slightly more, the pressure in the system prevents boiling).

I don't see how this would be any different if the server would be immersion cooled with mineral oil rather than air; in both cases the waste heat needs to be exchanged to water, and even with immersion cooling you couldn't run the system that much hotter without affecting the reliability of the servers.

That might be a relevant argument if immersion cooling (or generally, liquid cooling of the servers themselves) would somehow be new, innovative, or non-obvious. It's none of those. Secondly, I didn't mean to imply that google would turn around on a dime, but rather that at least some of the newer data centers would use something better if available. The Hamina data center seen in those picture, for instance, was opened in 2012 and seems to use the same air-cooled hot-aisle containment design. I haven't seen "Princess Bride" (assuming it's a movie or play), so I won't comment on that.

Google runs their datacenters at quite high temperatures, the cold side is around 25 C, hot side 50 C. I suppose it would be a pretty unpleasant working environment if the main space of the server rooms would be at 50 C rather than 25 C.

Current generation google datacenters already have a PUE around 1.1, so whatever they do by tweaking the cooling they cannot reduce the total energy consumption by more than 10 %. Of course, at their scale 10% is still a lot of energy, but the question is how much they could actually reduce that by going to immersion cooling. So far the anecdotal answer seems to be "not enough", since otherwise they would surely already have done it.

Every teacher individually?

AFAIU, yes. (That being said, while I have teached at the university level in Finland, I have no experience of the Finnish primary and high school system from the faculty viewpoint, so you might want to double-check with someone else). Also, consider that there are something like 5 million Finnish speakers, so it's not a particularly large market, so teachers are not exactly going to be overwhelmed by the number of available textbooks. E.g. in physics I think there are about 3-4 book series covering the high school curriculum. I suppose it's a bit different in the US, where one presumably cannot assume a teacher has time to evaluate all the available textbooks. Then again, at least from over here it seems that textbook selection in the US is extremely politicized (can a biology textbook cover evolution? WTF!?) which probably isn't conductive to a good outcome either.

Textbooks must teach to the content of the abitur and the standards being established by the Bologna Process. So, I guess the curricula are well defined. But I'm still surprised that this decision would be left to every teacher individually.

Yes, the Ministry of Education defines (broadly) the curriculum, so it's not like teachers are allowed to teach whatever they fancy. But generally, the large degree of autonomy given to teachers is often seen as one of the reasons why Finland does so well in these PISA tests. Teachers over here are pretty well educated, and it's a well regarded profession. Of course, there are other reasons as well, e.g. Finland is culturally pretty homogeneous and there are quite small socioeconomic differences compared to many other countries. Anyway, it's not like teachers are alone in choosing textbooks, of course they talk with colleagues etc., and professional societies do from time to time publish reviews of the available textbooks, which I assume teachers read carefully.

As an aside, the Bologna process AFAIK covers only higher education (at the polytechnic/university level, bachelor/master/Phd), not high school. Of course, it indirectly covers lower education as well in the sense that it effectively requires that students entering higher education have certain skills.

I think this Finnish group needs someone who is an insider on textbook selection committees to advise them. The last thing these committees want is to embarrass themselves by being seen to recommend a work that was produced in three days. They would lose their credibility, regardless of the quality of the work.

IIRC there are no textbook selection committees in Finland. Teachers are free to choose whichever book they want; or indeed to not choose any book at all and teach the class based on their own material.

The thing has a wingspan on 68m, more than an A340. Yet it weighs 1600 kg, about the same as a car. Carbon fiber and epoxy is a pretty impressive combination..

Some additional points:

- FWIW, Linux finally got rid of the BKL in the 3.0 release or thereabouts.

- Many (most?) 10Gb NIC's are multiqueue, meaning that the interrupt and packet processing load can be spread over multiple cores.

- Linux and presumably other OS'es have mechanisms to switch to polling mode when processing large numbers of incoming network packets.

That being said, your basic points about interrupt latency being an issue still stand, of course.

Google's webm project is also under a (3-clause) BSD license with an additional patent license on top (which ceases in case of a patent suite, similar to GPLv3, or the Apache license v2).

They are going to have to get their electricity from somewhere & generating capacity don't grow on trees.

Unless they burn, um, err, apples? Yes, APPLES!

1. 1. Solve world energy crisis
2. 2. Get Nobel peace price
3. 3. Profit!!!

Man, I'm awesome!

To recap, KWin currently supports:

• No compositing
• Compositing using the 2D XRender interface

• Compositing using OpenGL 1.x

• Compositing using OpenGL 2.x
• Compositing using OpenGL ES 2 (code mostly shared with the OpenGL 2.x codepath)

So what is suggested here is to delete support for compositing using OpenGL 1.x.

Personally, I can hardly blame the developer for wanting to prune that list a bit.

And, if you don't want to see this feature deleted, now is your opportunity to step up to the plate and contribute!

Haven't you understood; Watching gcc output scroll by for hours on end will make you l33t! That's why Gentoo and FreeBSD users are so hardcore.

What you're looking for is the Green500 list [green500.org]

Indeed, but the site was down when I wrote my previous reply so I had to resort to the top500 list and calculating flops/watt for the few top entries manually. :)

In any case, as one can see from the list, the best GPU machine manages to beat the K machines by a factor of 1.66, a far cry from the factor of 3-6 you originally claimed. And most GPU machines fall behind the K.

I think the sparc viiifx is quite impressive, it gets very good flops/watt without being a particularly exotic design. Basically it's just a standard OoO CPU with a couple extra FP units and lots of registers clocking at a little lower frequency than usual. No long vectors with scatter/gather memory ops, no GPU's, no low power very slow embedded CPU's like the Blue Genes etc.

I have no knowledge of the design tradeoffs of the individual systems, but I'd say that it's fairly impressive that both the top500 and the Green500 have so many GPUs in the top 10, given that they're both CPU-dominated lists.

Large GPGPU clusters are still a relatively new phenomenon, give it a few years and I suspect you'll see a lot more of them.

Oh? So how come the VIIIFX based "K computer" then, apart from being the current #1 in performance, also beats the GPGPU clusters (with the latest Nvidia Fermi cards) in flops/watt on the latest top500 list: http://top500.org/list/2011/06/100 ? And heck, that's on linpack, which should be the pretty much optimal workload for a GPU.

How much does computer time on these things cost? How is the cost calculated? Is time divided up something like how it's done on a large telescope, where the controlling organization get proposals from scientists, then divvies up the computer's available time according to what's been accepted?

On the supercomputer centers I'm familiar with, scientists write proposals which are evaluated by some kind of scientific steering committee which meets regularly (say, once per month), and gives out a certain amount of cpu-hours depending on the application.

Do they multi-task (run more than one scientists' program at one time)?

Yes. Typically the users write batch scripts requesting the amount of resources their job needs. E.g. "512 cores with at least 2 GB RAM/core, max runtime 3 days", and then they submit the batch job to a queue. At some point when there are enough free resources in the system, the batch scheduler launches the job. When the job finishes (or during its runtime) the usage is then subtracted from the quota they were awarded in the application process.

Does the computer run at top power (10pf) at all times, or does the resource usage go up and down?

Usually all functioning nodes are running and available for use, yes. Typically load is around 80-90% of maximum, due to scheduling inefficiencies etc. (e.g. a large parallel job needs to wait until there are enough idle cores before it can start, and so forth).

And lastly, how hard is it to write programs to run on these things? Do the scientists do it themselves, and if so, do the people who run the supercomputer audit the code before it runs?

Pretty tricky. Usually they use the MPI library. The programs are either written by the scientists themselves, or by other scientists working in the same field. The supercomputing center typically doesn't audit code, but may require the user to submit scalability benchmarks before allowing the user to submit large jobs. For some popular applications the supercomputing center may maintain a version themselves (so each user doesn't need to recompile it) and provide some more or less rudimentary support.