yes, of course. Our cluster is currently so unstable that I just perform simulations in chunks of steps and restart at every chunk if it went down again. It could be done more efficiently maybe, but then it would cost me more time and patience to administer all that.
Can you imagine a top500 supercomputer with scheduled downtime every week on the same day for maintenance? We tried asking nicely to get rid of that, but apparently there's no way out. Doing official complaints would just make the sysadmins being pissed at you, and most likely not solve the problem. Since we still need the sysadmins for any greater problem that might occur, we just go with it now. For ab-initio runs where you have to do huge matrix operations that take several days, this one day of downtime means that the actual speed of the cluster is half of what it could be:( Probably it's faster to have another parallelization in the code.
indeed, apparently more people seem to mix these up. It's a useless list anyway, as I argued in a similar reply to this article.
A bit off-topic, if you used a T3E you must have been out of science for a while now;) do you think that in the long run, the loss of freedom in your work and working hours is worth the higher pay and maybe higher sense of usefulness when switching to the big bucks industry?
Hi. I've had access to at least two systems of the current top500 list, and let me tell you and others referring to the top500, these are supercomputers, not mainframes. I hope that a mainframe will never reach this list, because of several reasons:
the top500 machines are made for showing off computerpower. whereas mainframes have not so much to do with clock cycles, more with handling heavy loads. Probably there are a lot of mainframes in use that can be easily outperformed by my EEE, but do it reliably for years after each other, whereas my EEE would probably be molten by the continuous load.
because of this, expenses go to fast cpus and fast networks, not so much solid data storage. Of course there is some terabyte storage present, but these are for scratch data storage mostly, there are no backups (some PhD ignored the warning, everything went well of course for a long time. that time was unfortunately before writing the thesis together, so all calculations had to be redone in a few months)
same for redundancy. If part of a supercomputer goes down, anything running on it is lost. Bad news for a scientist running on it, he needs to restart his work. Impossible news for a Wall Street data processing machine.
actually I call bogus on this list. The list has become a political cause on itself for countries/organisations to put themselves in the spotlight. Many of the supercomputing tasks could be done more effectively on local smaller clusters. In practice, a supercomputer is built to run as soon as possible the top500 benchmark, after which for the actual users several years of pain start including
MPI hardware incompatibilities, sudden events of slow data access, badly configured queueing systems because the staff is already busy enough dealing with getting the machine to work at all.
see it like this: probably every machine on that list is a 1st generation product. It has too be, otherwise it is too slow. That means that the staff is faced with a shitload of first generation bugs for which there is no standard protocol to go by. By the time they ironed it out, the machine has gotten redundant, a new "state of the art" has been ordered, and the whole shitty process starts again. I hope it is clear that all these points are not good properties for a mainframe that manages stock handling
I am also wondering if the tone of the parent post is fitting, as the poster seems not to know very much what he is talking about himself. Of course a big part of banking is now also to do simulations of economics, for which they will need clusters. But that work is additional to the mainframe administrative jobs, not substituting it.
Actually, I am a CLI person now, reading mail with pine, using bitlbee to chat on msn over irc. (ok, so those are interactive, but you get the point). All open in a screen session, so I can open it in a second from everywhere where I can open an ssh connection.
Paradoxically, I do that for the same reason that I like Hypercard: I want to use the tool that is the fastest and easiest way to do something. I use latex for large text/equations documents, but powerpoint for presentations, because the outlining in latex became tedious, and powerpoint is amazing for sloppy and very fast layout work.
What desktop? My Xandros EEE is my main computer now. If I find an external TV card for it, I have no reason to boot up the old desktop anymore. It's just taking space.
In the natural sciences, one tries to develop theories that have the power to describe the underlying principles of a naturally occurring phenomenon, with the goal to use these theories as predictive tools. When a theory fails to predict certain phenomena, it needs tweaking or a paradigm shift, until both its describing and predicting value are in correspondence with the natural state of things.
Say, this is where Intelligent Design fails as a scientific theory. Even though you could use it as a tool to describe underlying principles, its predictive value is 0.
For economic theories, you have the problem that the phenomena that occur are not "natural". You can most likely influence the course of an economy by publishing new theories. Try influencing the speed of light just by convincing the light that its fundamental theory is different than before.
As someone who used HyperCard as a 10-12 year old, without using the manual ever, I realize only now that I never realized that by using HyperCard I was actually programming. The program must have been made in such a way that you could perform pretty complex operations with it, without even knowing that what you are doing is complex.
Years later I tried to do similar simple interactive animations for adobe flash. It faced me with multitudes of concepts, each with their own drop-down menus and rules, before I could even start drawing something. Maybe it was more easy as a child because I had no idea of what I was doing, but more likely HyperCard was just designed very elegantly.
Amazing! The best kind of advertising you can have. That is, as long as your content is worthwhile (probably the reason why piracy of most blockbuster movies is so strongly enforced, there is absolutely no reason to have a high-quality version of them).
News for content creators: Advertising sells.
Something related that bugs me: a music video is a bloody "promotional video" for the music it is made to. Doing your best to limit the distribution of that advertisement is just shooting yourself in the foot. Idiots.
Seriously, can someone who is in this business explain me why you would need SMP for a mobile device? If you are going to deploy small CPUs in parallel, is the performance loss in communication not so huge that it just makes more sense to switch to a more powerful CPU instead?
Yeah, I was thinking the same about mini-itx board for a long time, too. Maybe this board is a bit cheaper, but I found any mini-itx configuration I looked for until now to have a higher price/performance than a mac-mini. It's not just the board, you'll still need the power supply, memory, HD, enclosure, maybe DVD drive etc. I guess that for mom and dad a mac-mini or EEE would be a better choice, because it's about the same price, but both the hardware and the OS are standardized. They will be able to solve some software problems all by themselves without having to ask you. And you will save time fine-tuning your CentOS install.
That said, this motherboard might find its way into linux preinstalled boxes, although more likely the motherboards for mom-and-dad PCs won't have the parallel and serial port. This motherboard will then have to find its way into industrial applications then, where these legacy ports are of main importance. Pretty neat little thing, though, can't wait until I get my hands on a nice Atom-based laptop.
Sometimes, when voting, you actually have a choice. Try to buy a bike that has not been made in china/taiwan. Bike manufacturer Scott put stickers with american flags on its bikes, next to a sticker "made in taiwan". Thinking you can vote with your money is just fooling yourself.
Or... these developers just could have seen it coming. The whole idea was nice but flawed. Creating a computer that is effectively _limited_ to developing countries hardly makes any sense, because you not only exclude yourself from the big market of all other countries but also make the adoption and improvement of it more difficult because you'll end up with much less user-contributed support. Only the people who went to great hoops to get one can contribute.
So what would have worked, then? A friend of mine went to volunteer in Botswana, and specifically left his digital camera at home. To find out that most people in Botswana had newer mobile phones than himself, and some had pretty nifty digital cameras as well. So, compared to a monthly salary they have to pay a lot for this stuff, but it's also a prestige issue. Furthermore, having the OLPC distributed via the governments is one hell of a way to make sure they will end up not at the people they were supposed to. We can presume however that any product freely available on the world market will reach anyone willing to buy it, developing country or not.
I think that the development of commercial sublaptops will eventually fulfill the goal of the OPLC much better than the OPLC could have ever done. So they are a bit more expensive and miss some of the nifty battery-life/sturdiness/instant network features of the OPLC, but their performance specs are mostly better, they (well, the EEE) already have been tested and tried by millions of people, and any info you would need to know about it to change whatever setting you want to change can be found on the web. They have a bit of "chic" to them, so people will actually want to show off with them. I know I do, and I do;)
Furthermore, look what Asus managed to get done just by their commercial motives! They had Xandros create a very simple interface specially for this machine, and it can also play movies right out of the box, they managed Microsoft to get on their knees and please please install XP on it. And just look at the new EEE: the XP version will come with 12 GB of SSD, whereas the linux version comes with 20 GB, making the linux version automatically the better choice (don't know how the price difference is, though). That is the opposite of what the OLPC-project managed to reach. Now I'd be the last to say that a free market economy is the solution to everything, but in this case we are talking about the economics of bulk: getting a product out to as many people as possible. And you just cannot do that without considering economic motives.
It says that you can do illegal things as long as you pay up (in this case you pay your time to do the quiz). Though of course not per se morally supportable, but at least it resembles reality.
If you leave your door open, you are liable to getting your house robbed and getting no insurance refund for it. However, it could also just happen that you came back and everything was still there. That's how your liability depends on the actions of others.
The originals were defining movies for a generation.
Correct, but along with that these "originals" also weren't very good from a film critic perspective. They were just the right kind of movies at the right kind of time. If any if them would have been released 5-10 years later than they were, they would have gotten poor reviews and had marginal response by the public as well. I convinced my gf to watch "A new hope", she had never seen a star wars movie before, nor known about the story, etc. If you watch it without all that luggage, it really is a thin little story. Face it, people, George Lucas always did what he did, make simple entertaining movies for a large public. Except maybe thx 1138. Not per se a good movie, but at least not entertaining to anyone;)
In old world terms: Why is there no newspaper called "newspaper", or, if you want to see music videos, would you watch a TV station called MTV/Music television?
In the early days, music videos were called "promotional clips", because that is what they do, they help sell the music. Actually, if you make a video to a piece of music you are doing just that, promoting the music to others. Why the hell wouldn't they want their free advertising? It's a mad world. (Please listen to Gary Jules or Tears for Fears (depening on when you were born) while reading this post)
Actually the cans are recycled into... cans. Apparently recycled aluminum are not good for the strength needed in bikes or cars.
As for cheap chinese framemakers, I have to agree with you there. Also if the factory was working on the penny they might have skipped some crucial hardening steps and then your frame can just rip apart easily at some point. You should however be able to buy a good-quality aluminum bike from about 500 euro. Still a factor 5-10 better than in the early days. Carbon fiber is a different story, because you need actually skilled, well-paid people to do that by hand, the price will remain on the high side for a long time to come.
Just like any company, Apple is not some fairy godmother who goes around granting people's wishes. If they had interest in the low-cost laptop business, why aren't they in it by themselves? Intel did it, Via did it, Asus did it. Apparently Apple can adapt their OSX to low-cpu power systems like on the iPhone, but I am not sure if they are willing to lose their profit margins and go into that sector. A mini apple in the same price and hardware configuration like the EEE would wipe out the whole field. I know, competition is supposed to be good, but as an EEE fan I had a sigh of relief when apple instead of an "eee-killer" introduced an overly expensive Airbook. Anyone out there that thinks the apple cheap notebook is still to come?
Ok, thank you for pointing that out, I didn't know. So, to correct: I welcome the rich pensioners who bought Mercedes cars with the first airbags that didn't kill you, which Mercedes developed DESPITE the bad reputation the inferior early airbags had in the US at that time. Now that must have taken some guts.
And, thanks for pointing that other one out, I thank the people who bought the first expensive cars with anti-lock brakes (again, not used the first time by Mercedes, but still greatly improved by them and Bosch, so that not only the driver with amazing driving "skillz" like you can brake safely, but also the unexperienced driver behind him (!) so that that one wont crash into the driver who knew who to stop safely. You're not the only one on the road, you know.
I earn my living trying to add my little contribution to advance society through material science, and I, for one, welcome our selfish rich people overlords.
I welcome the rich pensioners that bought Mercedes cars with airbags in the 80's, so that development by Mercedes could be financed and now you get life-saving airbag in even the smallest cars.
I welcome the yuppies that bought the first aluminum bikes, costing probably several thousand dollars back then, but now anyone can have a bike that is light and doesn't rust.
I welcome the showoffs that wanted a mobile phone in the early 90s, so now wireless technology is cheap enough to be used in third world countries, and get people connected.
Should I go on? Advances, especially in materials, are often sustainable because of some marginal hobbies of rich people. They want the lightest and strongest, even when it is actually not needed for their cause (do fishing rods really need to be made out of carbon-fiber?). But the amount of money that they want to invest can keep small innovative companies alive. In the end, we all win.
Slashdot Mirror
Can you imagine a top500 supercomputer with scheduled downtime every week on the same day for maintenance? We tried asking nicely to get rid of that, but apparently there's no way out. Doing official complaints would just make the sysadmins being pissed at you, and most likely not solve the problem. Since we still need the sysadmins for any greater problem that might occur, we just go with it now. For ab-initio runs where you have to do huge matrix operations that take several days, this one day of downtime means that the actual speed of the cluster is half of what it could be :( Probably it's faster to have another parallelization in the code.
A bit off-topic, if you used a T3E you must have been out of science for a while now ;) do you think that in the long run, the loss of freedom in your work and working hours is worth the higher pay and maybe higher sense of usefulness when switching to the big bucks industry?
- the top500 machines are made for showing off computerpower. whereas mainframes have not so much to do with clock cycles, more with handling heavy loads. Probably there are a lot of mainframes in use that can be easily outperformed by my EEE, but do it reliably for years after each other, whereas my EEE would probably be molten by the continuous load.
- because of this, expenses go to fast cpus and fast networks, not so much solid data storage. Of course there is some terabyte storage present, but these are for scratch data storage mostly, there are no backups (some PhD ignored the warning, everything went well of course for a long time. that time was unfortunately before writing the thesis together, so all calculations had to be redone in a few months)
- same for redundancy. If part of a supercomputer goes down, anything running on it is lost. Bad news for a scientist running on it, he needs to restart his work. Impossible news for a Wall Street data processing machine.
- actually I call bogus on this list. The list has become a political cause on itself for countries/organisations to put themselves in the spotlight. Many of the supercomputing tasks could be done more effectively on local smaller clusters. In practice, a supercomputer is built to run as soon as possible the top500 benchmark, after which for the actual users several years of pain start including
MPI hardware incompatibilities, sudden events of slow data access, badly configured queueing systems because the staff is already busy enough dealing with getting the machine to work at all.
- see it like this: probably every machine on that list is a 1st generation product. It has too be, otherwise it is too slow. That means that the staff is faced with a shitload of first generation bugs for which there is no standard protocol to go by. By the time they ironed it out, the machine has gotten redundant, a new "state of the art" has been ordered, and the whole shitty process starts again. I hope it is clear that all these points are not good properties for a mainframe that manages stock handling
I am also wondering if the tone of the parent post is fitting, as the poster seems not to know very much what he is talking about himself. Of course a big part of banking is now also to do simulations of economics, for which they will need clusters. But that work is additional to the mainframe administrative jobs, not substituting it.What if you bring along, say, two girls. And a cup.
Paradoxically, I do that for the same reason that I like Hypercard: I want to use the tool that is the fastest and easiest way to do something. I use latex for large text/equations documents, but powerpoint for presentations, because the outlining in latex became tedious, and powerpoint is amazing for sloppy and very fast layout work.
What desktop? My Xandros EEE is my main computer now. If I find an external TV card for it, I have no reason to boot up the old desktop anymore. It's just taking space.
Say, this is where Intelligent Design fails as a scientific theory. Even though you could use it as a tool to describe underlying principles, its predictive value is 0.
For economic theories, you have the problem that the phenomena that occur are not "natural". You can most likely influence the course of an economy by publishing new theories. Try influencing the speed of light just by convincing the light that its fundamental theory is different than before.
Years later I tried to do similar simple interactive animations for adobe flash. It faced me with multitudes of concepts, each with their own drop-down menus and rules, before I could even start drawing something. Maybe it was more easy as a child because I had no idea of what I was doing, but more likely HyperCard was just designed very elegantly.
yeah, swell, let's have amazon's servers encrypt all traffic, that'll reduce their server load.
News for content creators: Advertising sells.
Something related that bugs me: a music video is a bloody "promotional video" for the music it is made to. Doing your best to limit the distribution of that advertisement is just shooting yourself in the foot. Idiots.
Seriously, can someone who is in this business explain me why you would need SMP for a mobile device? If you are going to deploy small CPUs in parallel, is the performance loss in communication not so huge that it just makes more sense to switch to a more powerful CPU instead?
That said, this motherboard might find its way into linux preinstalled boxes, although more likely the motherboards for mom-and-dad PCs won't have the parallel and serial port. This motherboard will then have to find its way into industrial applications then, where these legacy ports are of main importance. Pretty neat little thing, though, can't wait until I get my hands on a nice Atom-based laptop.
Sometimes, when voting, you actually have a choice. Try to buy a bike that has not been made in china/taiwan. Bike manufacturer Scott put stickers with american flags on its bikes, next to a sticker "made in taiwan". Thinking you can vote with your money is just fooling yourself.
Or... these developers just could have seen it coming. The whole idea was nice but flawed. Creating a computer that is effectively _limited_ to developing countries hardly makes any sense, because you not only exclude yourself from the big market of all other countries but also make the adoption and improvement of it more difficult because you'll end up with much less user-contributed support. Only the people who went to great hoops to get one can contribute.
So what would have worked, then? A friend of mine went to volunteer in Botswana, and specifically left his digital camera at home. To find out that most people in Botswana had newer mobile phones than himself, and some had pretty nifty digital cameras as well. So, compared to a monthly salary they have to pay a lot for this stuff, but it's also a prestige issue. Furthermore, having the OLPC distributed via the governments is one hell of a way to make sure they will end up not at the people they were supposed to. We can presume however that any product freely available on the world market will reach anyone willing to buy it, developing country or not.
I think that the development of commercial sublaptops will eventually fulfill the goal of the OPLC much better than the OPLC could have ever done. So they are a bit more expensive and miss some of the nifty battery-life/sturdiness/instant network features of the OPLC, but their performance specs are mostly better, they (well, the EEE) already have been tested and tried by millions of people, and any info you would need to know about it to change whatever setting you want to change can be found on the web. They have a bit of "chic" to them, so people will actually want to show off with them. I know I do, and I do ;)
Furthermore, look what Asus managed to get done just by their commercial motives! They had Xandros create a very simple interface specially for this machine, and it can also play movies right out of the box, they managed Microsoft to get on their knees and please please install XP on it. And just look at the new EEE: the XP version will come with 12 GB of SSD, whereas the linux version comes with 20 GB, making the linux version automatically the better choice (don't know how the price difference is, though). That is the opposite of what the OLPC-project managed to reach. Now I'd be the last to say that a free market economy is the solution to everything, but in this case we are talking about the economics of bulk: getting a product out to as many people as possible. And you just cannot do that without considering economic motives.
It says that you can do illegal things as long as you pay up (in this case you pay your time to do the quiz). Though of course not per se morally supportable, but at least it resembles reality.
If you leave your door open, you are liable to getting your house robbed and getting no insurance refund for it. However, it could also just happen that you came back and everything was still there. That's how your liability depends on the actions of others.
Correct, but along with that these "originals" also weren't very good from a film critic perspective. They were just the right kind of movies at the right kind of time. If any if them would have been released 5-10 years later than they were, they would have gotten poor reviews and had marginal response by the public as well. I convinced my gf to watch "A new hope", she had never seen a star wars movie before, nor known about the story, etc. If you watch it without all that luggage, it really is a thin little story. Face it, people, George Lucas always did what he did, make simple entertaining movies for a large public. Except maybe thx 1138. Not per se a good movie, but at least not entertaining to anyone ;)
In old world terms: Why is there no newspaper called "newspaper", or, if you want to see music videos, would you watch a TV station called MTV/Music television?
In the early days, music videos were called "promotional clips", because that is what they do, they help sell the music. Actually, if you make a video to a piece of music you are doing just that, promoting the music to others. Why the hell wouldn't they want their free advertising? It's a mad world. (Please listen to Gary Jules or Tears for Fears (depening on when you were born) while reading this post)
As for cheap chinese framemakers, I have to agree with you there. Also if the factory was working on the penny they might have skipped some crucial hardening steps and then your frame can just rip apart easily at some point. You should however be able to buy a good-quality aluminum bike from about 500 euro. Still a factor 5-10 better than in the early days. Carbon fiber is a different story, because you need actually skilled, well-paid people to do that by hand, the price will remain on the high side for a long time to come.
The Dutch police have prior art on that idea, I'm afraid, they're trying to lobby for it, so no mention of an upcoming law yet. Remains ridiculous.
Just like any company, Apple is not some fairy godmother who goes around granting people's wishes. If they had interest in the low-cost laptop business, why aren't they in it by themselves? Intel did it, Via did it, Asus did it. Apparently Apple can adapt their OSX to low-cpu power systems like on the iPhone, but I am not sure if they are willing to lose their profit margins and go into that sector. A mini apple in the same price and hardware configuration like the EEE would wipe out the whole field. I know, competition is supposed to be good, but as an EEE fan I had a sigh of relief when apple instead of an "eee-killer" introduced an overly expensive Airbook. Anyone out there that thinks the apple cheap notebook is still to come?
And, thanks for pointing that other one out, I thank the people who bought the first expensive cars with anti-lock brakes (again, not used the first time by Mercedes, but still greatly improved by them and Bosch, so that not only the driver with amazing driving "skillz" like you can brake safely, but also the unexperienced driver behind him (!) so that that one wont crash into the driver who knew who to stop safely. You're not the only one on the road, you know.
I welcome the rich pensioners that bought Mercedes cars with airbags in the 80's, so that development by Mercedes could be financed and now you get life-saving airbag in even the smallest cars.
I welcome the yuppies that bought the first aluminum bikes, costing probably several thousand dollars back then, but now anyone can have a bike that is light and doesn't rust.
I welcome the showoffs that wanted a mobile phone in the early 90s, so now wireless technology is cheap enough to be used in third world countries, and get people connected.
Should I go on? Advances, especially in materials, are often sustainable because of some marginal hobbies of rich people. They want the lightest and strongest, even when it is actually not needed for their cause (do fishing rods really need to be made out of carbon-fiber?). But the amount of money that they want to invest can keep small innovative companies alive. In the end, we all win.
The moment that flying cars become available, I will start a business selling reinforced roofs.