Slashdot Mirror

Does this mean that the applications running on the "old" clusters, presumably using some flavor of MPI to communicate between nodes, will have to be ported somehow to become multithreaded applications ?

NCSA still has plenty of "old" style clusters around. Two of the more aging clusters, Platinum and Titan are being retired, to make room for newer systems like Cobalt. Indeed, the official notice was made just recently--they're going down tommorrow. However, as the retirement notice points out, we still have Tungsten, Copper, and Mercury (Terragrid). Indeed, Tungsten is number 5 on the Top 500, so it should provide more than enough cycles for any message-passing jobs people require.

So, anyone has any insights as to why/how this matters for the programmers ?

What it means is that programming big jobs is easier. You no longer need to learn MPI, or figure out how to structure your job so that individual nodes are relatively loosely-coupled. Also, jobs that have more tightly-coupled parallelism are now possible. The older clusters used high-speed interconnects like Myrinet or Infiniband (NCSA doesn't own any Infiniband AFAIK, but we're looking at it for the next cluster supercomputer). Although they provided really good latency and bandwidth, they aren't as high-performing as shared memory. Also, Myrinet's ability to scale to huge numbers of nodes isn't all that great--Tugsten may have 1280 compute nodes, but a job that uses all 1280 nodes isn't practical. Indeed, untill recently the Myrinet didn't work at all, even after partitioning the cluster into smaller subclusters.

This new shared-memory machine will be more powerful, more convienient, and easier to maintain than the cluster-style supercomputers. Hopefully it will allow better scheduling algorithms than on the clusters too--an appaling number of cycles get thrown away because cluster scheduling is non-preemptive.

I'd also like to point out some errors in the Computerworld article. NCSA is *currently* storing 940 TB in near-line storage (Legato DiskXtender running on an obscenely big tape library), and growing at 2TB a week. The DiskXtender is licenced for up to 2 petabytes--we're coming close to half of that now. The article therefore vastly understates our storage capacity. On the other hand, I'd like to know where we're hiding all those teraflops of compute--35 TFLOPS after getting 6 TFLOPS from Cobalt sounds more than just a little high. That number smells of the most optimistic peak performance values of all currently connected compute nodes. I.e. - how many single-precision operations could the nodes do if they didn't have to communicate, everything was in L1 cache, we managed to schedule something on all of them, and they were all actually functioning. Realistically, I'd guess that we can clear maybe a quarter of that figure, given machines being down, jobs being non-ideal, etc. etc. etc.

As a disclaimer, I do work at NCSA, but in Security Research, not High-Performance Computing.

Does this mean that the applications running on the "old" clusters, presumably using some flavor of MPI to communicate between nodes, will have to be ported somehow to become multithreaded applications ?

NCSA still has plenty of "old" style clusters around. Two of the more aging clusters, Platinum and Titan are being retired, to make room for newer systems like Cobalt. Indeed, the official notice was made just recently--they're going down tommorrow. However, as the retirement notice points out, we still have Tungsten, Copper, and Mercury (Terragrid). Indeed, Tungsten is number 5 on the Top 500, so it should provide more than enough cycles for any message-passing jobs people require.

So, anyone has any insights as to why/how this matters for the programmers ?

What it means is that programming big jobs is easier. You no longer need to learn MPI, or figure out how to structure your job so that individual nodes are relatively loosely-coupled. Also, jobs that have more tightly-coupled parallelism are now possible. The older clusters used high-speed interconnects like Myrinet or Infiniband (NCSA doesn't own any Infiniband AFAIK, but we're looking at it for the next cluster supercomputer). Although they provided really good latency and bandwidth, they aren't as high-performing as shared memory. Also, Myrinet's ability to scale to huge numbers of nodes isn't all that great--Tugsten may have 1280 compute nodes, but a job that uses all 1280 nodes isn't practical. Indeed, untill recently the Myrinet didn't work at all, even after partitioning the cluster into smaller subclusters.

This new shared-memory machine will be more powerful, more convienient, and easier to maintain than the cluster-style supercomputers. Hopefully it will allow better scheduling algorithms than on the clusters too--an appaling number of cycles get thrown away because cluster scheduling is non-preemptive.

I'd also like to point out some errors in the Computerworld article. NCSA is *currently* storing 940 TB in near-line storage (Legato DiskXtender running on an obscenely big tape library), and growing at 2TB a week. The DiskXtender is licenced for up to 2 petabytes--we're coming close to half of that now. The article therefore vastly understates our storage capacity. On the other hand, I'd like to know where we're hiding all those teraflops of compute--35 TFLOPS after getting 6 TFLOPS from Cobalt sounds more than just a little high. That number smells of the most optimistic peak performance values of all currently connected compute nodes. I.e. - how many single-precision operations could the nodes do if they didn't have to communicate, everything was in L1 cache, we managed to schedule something on all of them, and they were all actually functioning. Realistically, I'd guess that we can clear maybe a quarter of that figure, given machines being down, jobs being non-ideal, etc. etc. etc.

As a disclaimer, I do work at NCSA, but in Security Research, not High-Performance Computing.

Does this mean that the applications running on the "old" clusters, presumably using some flavor of MPI to communicate between nodes, will have to be ported somehow to become multithreaded applications ?

NCSA still has plenty of "old" style clusters around. Two of the more aging clusters, Platinum and Titan are being retired, to make room for newer systems like Cobalt. Indeed, the official notice was made just recently--they're going down tommorrow. However, as the retirement notice points out, we still have Tungsten, Copper, and Mercury (Terragrid). Indeed, Tungsten is number 5 on the Top 500, so it should provide more than enough cycles for any message-passing jobs people require.

So, anyone has any insights as to why/how this matters for the programmers ?

What it means is that programming big jobs is easier. You no longer need to learn MPI, or figure out how to structure your job so that individual nodes are relatively loosely-coupled. Also, jobs that have more tightly-coupled parallelism are now possible. The older clusters used high-speed interconnects like Myrinet or Infiniband (NCSA doesn't own any Infiniband AFAIK, but we're looking at it for the next cluster supercomputer). Although they provided really good latency and bandwidth, they aren't as high-performing as shared memory. Also, Myrinet's ability to scale to huge numbers of nodes isn't all that great--Tugsten may have 1280 compute nodes, but a job that uses all 1280 nodes isn't practical. Indeed, untill recently the Myrinet didn't work at all, even after partitioning the cluster into smaller subclusters.

This new shared-memory machine will be more powerful, more convienient, and easier to maintain than the cluster-style supercomputers. Hopefully it will allow better scheduling algorithms than on the clusters too--an appaling number of cycles get thrown away because cluster scheduling is non-preemptive.

I'd also like to point out some errors in the Computerworld article. NCSA is *currently* storing 940 TB in near-line storage (Legato DiskXtender running on an obscenely big tape library), and growing at 2TB a week. The DiskXtender is licenced for up to 2 petabytes--we're coming close to half of that now. The article therefore vastly understates our storage capacity. On the other hand, I'd like to know where we're hiding all those teraflops of compute--35 TFLOPS after getting 6 TFLOPS from Cobalt sounds more than just a little high. That number smells of the most optimistic peak performance values of all currently connected compute nodes. I.e. - how many single-precision operations could the nodes do if they didn't have to communicate, everything was in L1 cache, we managed to schedule something on all of them, and they were all actually functioning. Realistically, I'd guess that we can clear maybe a quarter of that figure, given machines being down, jobs being non-ideal, etc. etc. etc.

As a disclaimer, I do work at NCSA, but in Security Research, not High-Performance Computing.

BYU Internet Security Research Lab
Urbana-Champaigne Database and Information Systems Laboratory

> Who the wants to pay to go to NYC, and get harassed by facist policemen searching for bin Laden in every subway car?

How about an Native American reservation then? Finding one with a casino might fulfill the need for a nightlife that others have mentioned.

> If I were holding a conference like this, I'd find some depressed mid-western or just rural city that is cheap as shit and as easy as possible to get to.

I think you're on to something. People are complaining about nothing to do, but consider The Burning Man. It's in the middle of the desert, for crying out loud! In the midwest, your spouses and kids could explore nature while you're hacking away: horse rides, biking/hiking trails, camping, fishing, etc.

> A small college town might do; it would have to be close enough to an airport serviced by Southwestern for cheap flights, that you could run a shuttle van back and forth to get people to it.

Don't forget about Amtrak or Greyhound either!

I can think of several college towns that might fit the bill: Urbana-Champaign, IL; Columbia, MO; and Lawrence, KS. I know all three have live music scenes. The computer science program at the University of Illinois Urbana-Champaign is considered top-rate. They also have a spiffy new building to show off.

> Who the wants to pay to go to NYC, and get harassed by facist policemen searching for bin Laden in every subway car?

How about an Native American reservation then? Finding one with a casino might fulfill the need for a nightlife that others have mentioned.

> If I were holding a conference like this, I'd find some depressed mid-western or just rural city that is cheap as shit and as easy as possible to get to.

I think you're on to something. People are complaining about nothing to do, but consider The Burning Man. It's in the middle of the desert, for crying out loud! In the midwest, your spouses and kids could explore nature while you're hacking away: horse rides, biking/hiking trails, camping, fishing, etc.

> A small college town might do; it would have to be close enough to an airport serviced by Southwestern for cheap flights, that you could run a shuttle van back and forth to get people to it.

Don't forget about Amtrak or Greyhound either!

I can think of several college towns that might fit the bill: Urbana-Champaign, IL; Columbia, MO; and Lawrence, KS. I know all three have live music scenes. The computer science program at the University of Illinois Urbana-Champaign is considered top-rate. They also have a spiffy new building to show off.

Last year, WA-based Digipen held its first video game creation camp in California, and the University of Illinois hosted one for girls.

Quantum chemistry is my field, and we are some of the few who actually study biochemical systems with it.

Unfortunately, I can't really say much about long-range effects myself; The number of atoms which can practically be modelled at the same time is about 100. (that's one-hundred! Not much!)

I'd say possibly the most important QM effect strucurally in enzymes is pi-stacking and pi-cation interactions. It's not something which can be modelled well by a point charge.
(And what is worse, not all QM methods can model it either; DFT is infamous for not predicting pi-stacking or VdW effects)

Another significant thing which only QM can really model is transition-metal complexes. The coordination is very difficult to predict offhand. (Field splitting, spin states, spin interactions, Jahn-Teller effect, etc, etc..)

Just the other month, Science had some interesting results, including a completeley unprecedented mode of binding for nitric oxide to copper in nitrite reductase.

Our area is the study of the catalytic functions of metalloenzyme active sites. Again, this is not something which is easily predictable.

(Like, look at Cytochrome c Oxidase.. The function was determined in 1977, X-ray crystal structures have been known for over a decade. The mechanism of proton-pumping is still unknown. (And there's lots of notable people studying it.)

The tools are actually getting quite good. I just went to a seminar last week where a student here had found that her folding prediction based on these worked out very similar to the X-ray crystalography that was later done. The changes were not very significant, and very close to the tolerances that were in the crystal.

There are ways to determine the function and regulation of a protein, though. None are in silico, and few are simple, but they do exsist. They are done all the time.

I think you're wrong when you say that 'Biology does NOT follow any simple rules of logic'. It does, and we know what most of them are. The problem is that it is a n-body problem. Give an astrophysicist a 200-body gravitational problem and he'll give yo a blank stare. A cell is a billion+ body problem, and there is more than just one interacting force. We know the rules, the system is just too complex to look at effectively. But I agree that the comparison to computers is a weak one. Biological systems are infinitely more complex than computers. Just to take the brain into consideration, there are 6 orders of magnitude more connections in the human brain than in a P4*. And that's just looking at a simple structure (synapse) as a single functional unit; within each synapse are dozens, if not hundreds of different kinds of players. With 100s to 100000s of copies of each player. And beyond that, you have long term regulatory effects,

*avg of 2000 synapses at each CNS neuron, and ~10^11 neurons in the human brain. Gives 10^13 connections total.

No, it is not. We have to pick our problems carefully to avoid running into No-Free-Lunch, but there are problems that hillclimbing is better on, and there are problems that GAs are better on.

If you're interested in toy problems, consider the simple "deceptive" problems of David Goldberg. These tend to be difficult for hill climbers, as the derivatives for decomposable segments lead to local optima. With folded traps, for instance, there are many many local optima. Bad for hill climbers.

Hill climbers are a great first thing to try with real world hard problems. But if they fail, GAs are a good tool to try. The only problem is that there are a lot of variants currently, and some research is not generally applicable. The field is immature.

Simmulated annealing, on the other hand, is a much more mature tool. Also very respectable and very useful.

If I am not being trolled, all I can recommend is reading more or, better yet, hire someone that actually, genuinely knows what they're talking about. Jürgen Branke is a perfect example, he has a very impressive track record.

Misses the point. GAs do not look at search spaces and ask 'is this function too bumpy'. GAs (as well as many other evolutionary algorithms, see GECCO) are more similar to stochastic greedy strategies. I'd suggest looking at Walsh coefficients instead of derivatives as a rule of thumb.

A large part of the reason GAs aren't catching on more is that there aren't enough really good text books on them. All of the really useful information is still in the form of research papers, these things take time. Also, we're only starting (last 5 years) to know enough to be able to routinely apply them. Without that, they're too hard and too expensive to use.

As for random search, good papers look at the scaling factor for their algorithms. The goal is to build GAs that are quadratic. Any paper which suggests an exponential time GA (like random search) is wasting your time. The idea is to have a range of algorithms, greedy search for logarithmic time (or gradient descent), GAs for quadratic, and enumerative search for NP-Hard problems.

The parent is either ignorant, overly jaded, or being a jerk.

Are you claiming that ipv6 has enough addresses for every particle in the universe? It doesn't. IPv6 has 2^128, or about 10^40 addresses. Even assuming 100% utilization of the available space, you're still a factor of 10^40 or more away from covering every particle in the universe.

BBC article, NCSA Web page. The NCSA built a supercomputer out of a cluster of 70 Playstation 2 computers. They're not actually using the main CPU for numbercrunching - they're using the Emotion Engine graphics chip.

If only Apple would have taken the time to implement the awesome history viewing of Trailblazer. Even though TrailBlazer is not a full featured browser, I still use it for some "show off" tendencies on my mac.

Steve, find these kids and put the functionality into Safari!

GroupShares Inc. - A Free and Interactive Stock Trading Community

BTW, on a slightly more serious note, there is a spiffy freeware 3D Mac OS X file browser called, appropriately, 3DOSX (screenshots page).

I would advise the use of Syzygy.
We use it here at the U of I, it's GPL, and it really excellent as far as power and stability go.
More importantly, it can run on a off-the-shelf PC. It can also run on windows or linux.

I strongly advise looking into it.

At least 5 of the top 10 systems are running Linux, starting at number two with Thunder at the Lawrence Livermore National Laboratory. The others are IBM BlueGene/L clusters at places #4 and #8, Tungsten at NCSA at #5, MPP2 at Pacific Northwest National Laboratory at #9, and probably also the Dawning 4000A at the Shanghai Supercomputer Center as #10, though I'm not 100% sure about this last one.

Interestingly enough, Peter Bentley's group results on car racing would not be considered human competitive, unless the results obtained in the simulation will be tried in the real world, or if the simulator is something experts actually use to shave of seconds. In any case, it seems the Evolutionary Computation world is starting to obtain very strong results, for a part due to Moore's law. It's possible that this is caused by the fact that the field simply tries to solve things, instead of first proving that it works (AI/ML), or proving that it doesn't work (Operations Research).

Interestingly enough, Peter Bentley's group results on car racing would not be considered human competitive, unless the results obtained in the simulation will be tried in the real world, or if the simulator is something experts actually use to shave of seconds. In any case, it seems the Evolutionary Computation world is starting to obtain very strong results, for a part due to Moore's law. It's possible that this is caused by the fact that the field simply tries to solve things, instead of first proving that it works (AI/ML), or proving that it doesn't work (Operations Research).

Hmmm, it looks like that site is down - try here instead.

As others have pointed out this is a new solution to a classic computer graphics problem. The first technique I know of to automatically reduce the poly count of meshes, while preserving the overall appearance was Garland and Heckbert's QSLIM algorithm. This was first published in SIGGRAPH 97. Or actually, hmmm, no, it looks like Hoppe's work on mesh optimization came a good bit earlier (1993).

Anyway, it's a pretty old problem in graphics. The USC press release that prompted this slashdot story is simply advertising Cohen-Steiner, Alliez, and Desbrun's paper which will appear at SIGGRAPH 2004 later this summer. That's all it is. They have a new way to do automatic poly reduction. Now it could be that it's vastly superior to anything else that's been done in the area, but even if so, this isn't likely to cause any revolutions. Why? Because the existing poly reduction algorithms already work pretty well. They work well enough that they're already in production use (as others have pointed out there are plugins for most major 3D packages already, and most game engines have had "continuous level of detail" systems for a good long while). So at best this is going to make life easier for some 3D content creators who won't have to do so much hand-tweaking of LODs (levels-of-detail, aka "optimized" meshes). So don't expect to see any huge changes in the games you play or movies or whatever because of this. Mesh optimization/LOD techniques are already being used pretty much everywhere it make sense to do so.

But here's an idea for all you Karma Whores out there: go to the list of papers on the SIGGRAPH 2004 web site (or go to Tim Rowley's easier to browse version of the list), pick something that looks interesting, and send the story to slashdot! There's at least 50 more slashdot stories there just waiting to burst! Happy hunting! There's enough Karma for everyone, so don't be greedy now.

As others have pointed out this is a new solution to a classic computer graphics problem. The first technique I know of to automatically reduce the poly count of meshes, while preserving the overall appearance was Garland and Heckbert's QSLIM algorithm. This was first published in SIGGRAPH 97. Or actually, hmmm, no, it looks like Hoppe's work on mesh optimization came a good bit earlier (1993).

Anyway, it's a pretty old problem in graphics. The USC press release that prompted this slashdot story is simply advertising Cohen-Steiner, Alliez, and Desbrun's paper which will appear at SIGGRAPH 2004 later this summer. That's all it is. They have a new way to do automatic poly reduction. Now it could be that it's vastly superior to anything else that's been done in the area, but even if so, this isn't likely to cause any revolutions. Why? Because the existing poly reduction algorithms already work pretty well. They work well enough that they're already in production use (as others have pointed out there are plugins for most major 3D packages already, and most game engines have had "continuous level of detail" systems for a good long while). So at best this is going to make life easier for some 3D content creators who won't have to do so much hand-tweaking of LODs (levels-of-detail, aka "optimized" meshes). So don't expect to see any huge changes in the games you play or movies or whatever because of this. Mesh optimization/LOD techniques are already being used pretty much everywhere it make sense to do so.

But here's an idea for all you Karma Whores out there: go to the list of papers on the SIGGRAPH 2004 web site (or go to Tim Rowley's easier to browse version of the list), pick something that looks interesting, and send the story to slashdot! There's at least 50 more slashdot stories there just waiting to burst! Happy hunting! There's enough Karma for everyone, so don't be greedy now.

FlightGear polygon reduction is based on this work

Or maybe I'm in that 0.003%? My mommy always told me I was special....

This is liquid cooled.

This is liquid cooled.

This is liquid cooled.

I do the same thing. The very last line of my resume indicates that I'm a couple months away from a PhD in Physics. No sense scaring them away early.

This came after NCSA told me I was overqualified. Course, it was just a help-desk position. ;)

Its not an insignificant amount...

Actually it is more than you may think. Microsoft's $56b cash hoard is bigger than most investment funds in the world. Hence, they get the best rates, the best opportunities and the best return.

For a lot of Americans, "security" means "looks alright," and for others, security is just something to get in the way. I work front-line tech support staff for the IT department at the U of Illinois at Urbana-Champaign (the university that Telnet and Mosiac came from), and we require a secure master password, and different and secure passwords on everything else (such as email, Active Directory services, etc). I routinely have to deal with customers who are irate because they can't put in their dog's name as the one and only password for everything, and that the master password expires in a year. Our guidelines are here. Do they seem too much to ask?

To sum up, most people just don't want security, because of the "it won't happen to me" mentality. Give some of these people an RSA card, and they've just gotten something else to yell about because they can't get into things as easy as they want. (One man started complaining about "First my bank wanted me to use some PIN number on my bank card, and now this!" in all seriousness while stumbling through setting his emails)

You're free to have a low opinion of Solaris, but you really should be more informed about it first. I'm not trolling or flaming, just hear me out before you mod me down.

Lacking GNU stuff in the default install may be a killer for you, but it really depends on your perspective. That is changing, anyway - solaris 9 has assorted gnu stuff included in /usr/sfw (if you install it), and I'm sure solaris 10 will have more. I really don't find it to be that big of a deal, because at my site I end up custom-building so much stuff anyway that it hardly matters.

I have no idea what you mean about commandline editing. Solaris comes with bash, and I bet that's what you use elsewhere.

The package format is old, and it will remain that way. It works the same now as it did ten years ago, and for many people that is a Good Thing. Anyhow, you don't need to make your own SysV packages. Make use of something like Encap to manage your /usr/local tree and keep your site-local packages away from the system package manager.

Patching isn't as hard as you make it out to be either. You could just wget ftp://sunsolve.sun.com/pub/patches/9_Recommended.z ip and run the install script in there to pick up any security fixes. Personally, I use a cron job to download that once a day, and then I run Superglue, which is just a shell script, to figure out which patches I need and install them out of an nfs share. It's simple, it works. Note that I don't cron the installation of packages, just downloading them. Cronned patch installation makes me nervous.

The installer sucks. Oh btw, DO NOT use the "install" cd. Boot "disc 1" and throw away the install disc. Seriously. Also, you really should not be using the installer if you have more than one or two systems. Set up a Jumpstart and net-install your machines. It really isn't difficult. One again, simple enough and it works.

Your ultra5 sucks, sorry to say it. Those were just lame machines. There is a reason they cost like $90 now: they suck. They can be made more useful with the addition of a $40 scsi card from ebay; that onboard cmd640 ide controller is godawful. Fwiw, I had an ultra5/360 with 256mb of ram and a pair of scsi drives (no goddamn IDE) withstand a full force slashdotting with ease. If you want a cheap sun machine that sucks less, find yourself an ultra2 2x300 with creator3d. Dual cpu, lots of ram slots that take normal sun ram, scsi storage. Should run you $300 or less. Sbus cards are cheap nowadays, too.

Solaris does support 24bit color depth on that machine; you just have to set it using fbconfig. I have no clue why they don't support 16bpp, and you're right, it is retarded.

And yeah, CDE sucks. Install Gnome. It comes with solaris 9 and it will be the default desktop in solaris 10.

Finally... I just don't bother using solaris as a desktop system. Almost all the workstations I run are Linux or MacOS. But damn, Solaris makes a good server OS.

Oh, I dunno, somewhere like here?

Large collaborative multi-user video games will still be kicking around with better graphics and resolution, normal progression from the old school days of Amiga.

The biggest leap will be in a new catagory tied into videogames, but on the educational side. The industry of edutainment will explode in the coming years, as educators realize that to make people learn, you make them interested. When I left University of Illinois in Urbana in 2000, a cool guy by the name of Sridaar Iyer was already working on a vrml multi-user videogame (in the traditional sense) that teaches High school kids about chemistry. His website is WhoolaIt challenged kids in groups to compete and learn. Yes, there has been issues with VRML and whatnot, but there are other alternatives. The point is kids love it, and get value from it.

At the University of Washington, my fiance is in a Masters program that teaches about Edutainment (cognitive studies). Talented engineers and artists are making great strides in this field. I think it will be one of those fields such as video game programming. Years ago, you didnt see specific programes JUST for video games. Now there everywhere (as are network security degrees and such). You will see these programs break off into seperate educationl only programs for games tied to teaching.

Multi-user online environments will also be suppliments to Lectures for courses. I use to do multi-user ADL (advanced distant learning) courses and testbed studies. We found that not only do people like collaborating like that, but physically challened people also greatly enjoy it, as they can still participate in the community and course, without fear of being judged due to a dissability. At the time the biggest problem was sending textures over the wires, but with pre-loaded disks with the textures already on them and stored locally, the only bandwith issues are avatar movement (UDP) and the live chat (TCP). It has a lot of potential.

As for home gaming, as computing needs become cheaper, you will see more CAVE VR environments. Paul Rajlick, formerly of the NCSA, is already working ona home version of the CAVE (using true VR/shutter glasses, wand, etc). You play CAVE quake and its unlike anything console system.

Overall, the market for games/edutainment will rival Hollywood. And the best part is, the field doesnt revolve around a bunch of conceited pricks :)

According to this this while the brighter (Alpha-2) is a much warmer white class A star with a temperature of 8500 Kelvin.
Which google says it's about 123.9 gas mark.

Why do you ask?

• Assuming the algae are 4% efficient. Solar cells are roughly 5X as efficient, and therefore would need cover only 10 thousand square kilometers. At $400/M^2, covering ~10,000 square kilometers would cost 4.14 trillion dollars, compared to the stated cost in the article of 169 billion for algae farms. Algae win with a 30:1 cost advantage.
• If you are more realistic and assume that the algae are more like 1% efficient, the solar cells will need to cover 2500 square kilometers, costing an even trillion dollars: The algae maintain a 6:1 cost advantage.

Hopefully, Microsoft will be doing something more worthwhile in their Windows HPC project.

Try this.if you actually facy the comparison.

There's an interesting presentation (8 MB PDF) from George Gollin, who researched (mostly on the Internet) these diploma mills. There are a few players who operate under a lot of different names. It's 123 pages, but basically a slide show, so it goes really fast.

This is very innovative!!! :

27 Reasons for the Iraq War
and the Lying Liars Who Tell Them

Enjoy!!!

Patriotically yours,
Kilgore Trout

Text file is here

They love volunteers

Actually, that statement is far from true. Volunteers generally cost more to train and manage than the amount spent on an employee.

My wife is a reference librarian at a small, rural public library and I am currently in Library School. I am a system admin in my department and have plenty of linux experience. The director of my wife's library will not consider any changes to their environment due to the need for full-time support. I've even offered a test system of my own that they can play with.

They use consultants from a local computer store to maintain their Windows 2K based network. Some representative from the local LUG isn't going to convince most librarians to hop on a new train that hasn't, in their experience, been proven. They've got enough headaches with filtering issues, perverts, and bratty kids whose parents think "library" means "free babysitting" in Greek.

IIRC, 1993, I started with Mosaic. Was not long after that when I moved to Netscape. Have been faithful ever since, and I stay current on Mozilla.

Engineering Hall? It was relatively spiffy the last time I was in there (last week). I'm pretty sure that they did remodel the whole place a few years back (as part of the Super Engineering Quad rebuild). Also - Dean Kamen? The Segway guy? He's not on the faculty here. Are you thinking of Sam Kamin?

Yeah... except for THIS soda machine (which just so happens to be in the seibel center) The link points at the web server running inside the pop machine itself. The only photo I can find of the thing is here, with one of the guys who worked on it sitting in front of it. And a BeBox perched on top.

Yeah... except for THIS soda machine (which just so happens to be in the seibel center) The link points at the web server running inside the pop machine itself. The only photo I can find of the thing is here, with one of the guys who worked on it sitting in front of it. And a BeBox perched on top.

I'll let you know before you read the rest of my post that I'm a current student at UIUC.

I got into the PhD programs at Stanford, Berkeley, MIT, Carnegie-Mellon, and UIUC--and UIUC compares very well with the rest of these schools. The only thing UIUC lacks is the publicity to go with the quality of research that happens here. On the other hand, this is a good thing since the students here can concentrate more on research instead of just working very hard at appearing smart like some other schools promote.

At UIUC, the professors are generally fairly young, which I view as a good thing. At the 'bigger' name schools you end up with a bunch of dinosaurs who may have contributed to the field in the past but are simply living off the legacy insteading doing new research. If you actually care about this, check out the UIUC research page at: http://www.cs.uiuc.edu/research/areas.html

I have personally found the AI, Databases, and Theory groups to be very impressive and have had experience working with them.

If you want an interesting comparison, check out MIT's new building.http://web.mit.edu/buildings/statacenter/ I took a tour of it, and the impression it gave me was, "Look at us, we're MIT! This building looks so crazy, we must be geniuses to work here!"

UIUC has a much more honest and less flashy style, which I find rather refreshing.
I do agree that most of the ubiquitous computing features of the building seem a little silly, but why not make your new computer science building a functional experiment in computer science itself?

Surprisingly, this is basically the same thing you get when you walk up to an information display on the wall (except the website doesn't have the information specific to the room you're standing in front of).

Everything actually looks really awesome right now. Too bad most of it isn't staying in the building after the weekend.

Anyway, to provide you with some other cool associated things:
web cam, VRML model of building. Enjoy.

PS Though I have complaints, I'm leaving those to the other trolls.

Surprisingly, this is basically the same thing you get when you walk up to an information display on the wall (except the website doesn't have the information specific to the room you're standing in front of).

Everything actually looks really awesome right now. Too bad most of it isn't staying in the building after the weekend.

Anyway, to provide you with some other cool associated things:
web cam, VRML model of building. Enjoy.

PS Though I have complaints, I'm leaving those to the other trolls.

Actually... here in the ACM office that dream may become a reality...

Caffeine is our own little magcard swipe internet enabled (crappy) soda machine...

It is out of Dew right now, however

Heh, and I'm a student right next door...