The difference between those memory technologies and this one seems to be that it could be integrated directly into the processor. "Cells" could retain memory without being powered. But that's just speculation by me.
Well, the problem with RAM is that it doesn't really hold state. Of course, you can keep it powered using battery, but would you give up so much reliability for a speed advantage? And RAM takes power all the time. This is fine if your server gets accessed all the time, but otherwise it is a waist of energy.
Also, even though prices are now fairly low, RAM is as expensive now as a flash SSD. That sounds reasonable, but remember the energy advantage of flash. Also think of the fact that flash memory as used in SSD drives has a very high price point compared with USB flash drives and Compact Flash. Sure, this is normally cheaper flash, but the difference cannot be *that* huge.
But most importantly: check the speed in which flash is getting bigger and cheaper. RAM is nowhere near this curve. PRAM also holds many promises, even though speculating on memory technology is a very risky thing to do. RAM won't be able to keep up. There are many more devices that need loads of flash compared to those that need RAM.
Finally, Intel and AMD are making chips for RAM drives. They are called chip sets (or, in case of AMD, processors) and are normally a feature of a motherboard you buy. This may sound like trolling, but most OS use the memory that is left after the processes have grabbed as much as they need as disk cache. This is actually quite near the idea of a RAM drive backed by a hard disk. This RAM drive has got a *very* speedy connection to the processor as well, beats the hell out of (S)-ATA.
As I understood, this depends on the amount of information stored per cell. Those cheaper memory modules are almost as fast as those that store a single bit per cell, but the disadvantage is more wear and tear (100k and less). Obviously, with wear leveling this is a bit of a moot point, but wear leveling slows down the process quite a bit.
Obviously if you have a web-server with many reads and few writes all this matters very little. If you have a web server with many reads and quite a few drive: use SSD *and* a drive. I mean, it's not an exclusive OR isn't it? That's only the case when the data that is written many times is also the stuff that is read many times.
From the view of the bands, it is the question how many payed downloads are needed to make up for lost CD sales. I don't know what percentage of a CD sale actually goes to the band, but I presume it's not much. Downloads and servers are getting cheaper by the minute...
Sometimes it is fun to dig into the stereotype. I can remember a birth-day party of a fellow student which was at the same time that his sister finished her school. It was pretty much fun finding out that they hadn't understood a single word we said during the entire time of the party. We also did a bit of laser gaming, very nerdy sport that can be.
In my opinion, Unit testing an interface is much less productive than Unit testing a new data structure or inner component. When I use Unit testing (not too much) it's always for core components within a system or for data structures.
Of course, I don't do Unit testing for test applications either. So all in all it does have to do with the system you are developing the application for, as well as the role of the component that you are creating within the application.
What a very distasteful article which seems only about bragging rights about his own very experimental system. Only the last paragraph seems to be getting to the point. The system this guy is devising does make an interesting read, but it unleashes many questions about the feasibility of such a system. It's certainly not something that AMD or Intel can currently adopt as a strategy.
Actually, it seems a bit about bragging rights about -your- very experimental system. In the kill file with you.
I haven't been known to be on the top of the CPU performance graph for years now, and this benchmark proves one thing to me: I'll rather go for a fast hard disk and fast internet connection before thinking about upgrading my CPU. Sure, there are things that are CPU bound. But only if I really see a need for performance for a very specific task, then I'll look for a quad core or high frequency processor. And if it is for a single task, I'll look around and ask a few friends who might have one already.
Maybe for batch graphics processing or PAR2, but even then the HDD might max out instead of the CPU.
You could send a hash over your data + your name to a publicly available time stamp server *before* you disclose it anonymously. For a list of available servers see this link:
http://security.polito.it/ts/
Oh, yeah, this is being researched. In the Netherlands, they are thinking about putting an island in the North sea and put a windmill park around it (usefull for damage repair as well). Now use the energy to pump out water from a big hole in the middle of the island, and use generators when you let it flow back in.
Much easier and cheaper and less ugly than building a big tower. Sometimes you wonder why you don't think of things like this yourself.
My company manages all that without being large (~300 persons). Are we sure we aren't talking about every company where the CEO doesn't know all the people working for the company? I can talk to the CEO on first name bases, most of us can and may do that, but he wouldn't necessarily know what's going on and how frustrating the working space can get.
Didn't the CEO of a well known power broker do the same? I can remember how well that company fared.
Yeah.
Besides it being counterproductive, I would also call it unsocial to the extreme. It can also lay down a very unhealthy strain on people and the company as a whole. What about people who have a major illness in the family?
And then there is the part on deciding who that 10 pro cent actually is. Gods, I can already see the PHB laying off Dilbert because an idea he (the PHB) had turned out sour.
Of course, the other extreme is that you have to lay of 10% of more or less randomly, as was the case in the NL some time ago. Then you can really get in trouble as a company.
Yes, but factories are coming more expensive by the minute. This is a different ballgame. Of course, the number of CPU's sold is probably also still rising.
The weird thing is the way Intel has kept its marketing share so high. The quality of the AMD server solutions seems very high in performance per watt. Still Intel is leading by a big margin, and seems to use its better production facilities again to kill AMD off.
It's Intels turn to be cheap as well. Since they've got their 45 nm online, they can sell chips very cheap. Check the comments on the actual costs of the upcoming Atom chip for instance.
Probably somebody was using your credentials against them in some sort of brute force password attack. They use the same reasoning that I've experienced with one company, they claimed I had a virus because the mails seemed to come from my mail account. Of course, they just looked at the from: address, which had been 'spoofed', if you would even call it that (it's unprotected really). It was a well known virus at that time, so somebody will probably fill it in right below this message:)
These are for internet transactions (which can also be initiated from most post order companies over there, by a popular scheme called iDEAL). So you will have to enter the number manually. The device is not connected to the PC at all, which is a major selling point (no hardware or OS requirements, no chance of loosing your PIN).
I do think these things were introduced by the Dutch Rabobank. I also know that these devices are not made by RSA but by Vasco. Easy to check, their internet address is on the back.
Basically these things work by using a generated number by a clock inside the device hashed together with some authentication info (the number of your account). This number will stay valid for a limited number of time and cannot be generated without the device. More information? Just look up the patent numbers on the back of the device: US Patents 4.599.489 and 4.609.777 .
You certainly have to fill in the total amount of the "commit" (which can consist of multiple transactions. If you go even higher you will have to fill in yet another number. Previously, at least at the Rabobank, they forgot to tell you that the number you had to fill in was the total transaction amount, so there was less safety (if only one person picks up that the amount is incorrect you can still easily identify a bad transaction leading to the attacker.
It's pretty secure, but its not foolproof if you are not sure that the site you are visiting isn't that of the bank. E.g. an attacker could just tell you on a spoofed site that you would now have to enter the amount in cents instead of euros, and 99% of the people would fill in that amount without even thinking about it. Or the spoofer could just wait until you enter a sufficiently high amount, and send it to its account on the Bahama's.
"Other parts of the world are chipping into 'software' already, and there's no reason to think that we have some kind of automatic, natural, competitive advantage in any of those."
Yes, you have. The English language is one for instance. And then there is the economy of scale. Any really interesting conference on Java and Security is in the old US of A, and it is a pretty big drawback if you're thousands of miles from the city it is held in.
Even books on all the subjects are a lot more expensive than in the US, and don't forget the higher living conditions (it's easier to do your job properly if you don't have to put in so much energy in the rest of your life).
Of course, some of these advantages are declining a bit, but I don't think they will disappear completely. So now it is up to you to compete. Somehow I am pretty sure that if I'd apply for a job over there, it would be easy as pie.
Ah, yes, I forgot about the "not replying" to each other. This was because they didn't want to make it obvious the information came from the same source. Seems that this guy forgot about this important point.
Anyway, I was not trying to say that this was the exact same thing, just that it reminded me of something. Clearly the guy doesn't have the romantic background of doing it because of trying to avoid age discrimination.
On Slashdot, I should be worried about age discrimination:)
Slashdot Mirror
It's the fifth... element... /me passes out.
The difference between those memory technologies and this one seems to be that it could be integrated directly into the processor. "Cells" could retain memory without being powered. But that's just speculation by me.
Well, the problem with RAM is that it doesn't really hold state. Of course, you can keep it powered using battery, but would you give up so much reliability for a speed advantage? And RAM takes power all the time. This is fine if your server gets accessed all the time, but otherwise it is a waist of energy.
Also, even though prices are now fairly low, RAM is as expensive now as a flash SSD. That sounds reasonable, but remember the energy advantage of flash. Also think of the fact that flash memory as used in SSD drives has a very high price point compared with USB flash drives and Compact Flash. Sure, this is normally cheaper flash, but the difference cannot be *that* huge.
But most importantly: check the speed in which flash is getting bigger and cheaper. RAM is nowhere near this curve. PRAM also holds many promises, even though speculating on memory technology is a very risky thing to do. RAM won't be able to keep up. There are many more devices that need loads of flash compared to those that need RAM.
Finally, Intel and AMD are making chips for RAM drives. They are called chip sets (or, in case of AMD, processors) and are normally a feature of a motherboard you buy. This may sound like trolling, but most OS use the memory that is left after the processes have grabbed as much as they need as disk cache. This is actually quite near the idea of a RAM drive backed by a hard disk. This RAM drive has got a *very* speedy connection to the processor as well, beats the hell out of (S)-ATA.
As I understood, this depends on the amount of information stored per cell. Those cheaper memory modules are almost as fast as those that store a single bit per cell, but the disadvantage is more wear and tear (100k and less). Obviously, with wear leveling this is a bit of a moot point, but wear leveling slows down the process quite a bit.
Obviously if you have a web-server with many reads and few writes all this matters very little. If you have a web server with many reads and quite a few drive: use SSD *and* a drive. I mean, it's not an exclusive OR isn't it? That's only the case when the data that is written many times is also the stuff that is read many times.
Don't worry, I've already clicked the thumbs down button, it seems somebody was trolling there :)
daishi
Some one who rapes little boys on the weekends
Source:
http://www.urbandictionary.com/define.php?term=daishi
From the view of the bands, it is the question how many payed downloads are needed to make up for lost CD sales. I don't know what percentage of a CD sale actually goes to the band, but I presume it's not much. Downloads and servers are getting cheaper by the minute...
Sometimes it is fun to dig into the stereotype. I can remember a birth-day party of a fellow student which was at the same time that his sister finished her school. It was pretty much fun finding out that they hadn't understood a single word we said during the entire time of the party. We also did a bit of laser gaming, very nerdy sport that can be.
In my opinion, Unit testing an interface is much less productive than Unit testing a new data structure or inner component. When I use Unit testing (not too much) it's always for core components within a system or for data structures.
Of course, I don't do Unit testing for test applications either. So all in all it does have to do with the system you are developing the application for, as well as the role of the component that you are creating within the application.
What a very distasteful article which seems only about bragging rights about his own very experimental system. Only the last paragraph seems to be getting to the point. The system this guy is devising does make an interesting read, but it unleashes many questions about the feasibility of such a system. It's certainly not something that AMD or Intel can currently adopt as a strategy. Actually, it seems a bit about bragging rights about -your- very experimental system. In the kill file with you.
Bwah, simply disable disasters and be done with it.
Well, now they do...
Yeah, so did I, the first couple of times.
Uh, can't you just put in a serial ata and power extension cable without the ice box thingy around it?
I haven't been known to be on the top of the CPU performance graph for years now, and this benchmark proves one thing to me: I'll rather go for a fast hard disk and fast internet connection before thinking about upgrading my CPU. Sure, there are things that are CPU bound. But only if I really see a need for performance for a very specific task, then I'll look for a quad core or high frequency processor. And if it is for a single task, I'll look around and ask a few friends who might have one already. Maybe for batch graphics processing or PAR2, but even then the HDD might max out instead of the CPU.
You could send a hash over your data + your name to a publicly available time stamp server *before* you disclose it anonymously. For a list of available servers see this link: http://security.polito.it/ts/
Oh, yeah, this is being researched. In the Netherlands, they are thinking about putting an island in the North sea and put a windmill park around it (usefull for damage repair as well). Now use the energy to pump out water from a big hole in the middle of the island, and use generators when you let it flow back in. Much easier and cheaper and less ugly than building a big tower. Sometimes you wonder why you don't think of things like this yourself.
My company manages all that without being large (~300 persons). Are we sure we aren't talking about every company where the CEO doesn't know all the people working for the company? I can talk to the CEO on first name bases, most of us can and may do that, but he wouldn't necessarily know what's going on and how frustrating the working space can get.
Didn't the CEO of a well known power broker do the same? I can remember how well that company fared.
Yeah.
Besides it being counterproductive, I would also call it unsocial to the extreme. It can also lay down a very unhealthy strain on people and the company as a whole. What about people who have a major illness in the family?
And then there is the part on deciding who that 10 pro cent actually is. Gods, I can already see the PHB laying off Dilbert because an idea he (the PHB) had turned out sour.
Of course, the other extreme is that you have to lay of 10% of more or less randomly, as was the case in the NL some time ago. Then you can really get in trouble as a company.
Yes, but factories are coming more expensive by the minute. This is a different ballgame. Of course, the number of CPU's sold is probably also still rising.
The weird thing is the way Intel has kept its marketing share so high. The quality of the AMD server solutions seems very high in performance per watt. Still Intel is leading by a big margin, and seems to use its better production facilities again to kill AMD off.
It's Intels turn to be cheap as well. Since they've got their 45 nm online, they can sell chips very cheap. Check the comments on the actual costs of the upcoming Atom chip for instance.
Probably somebody was using your credentials against them in some sort of brute force password attack. They use the same reasoning that I've experienced with one company, they claimed I had a virus because the mails seemed to come from my mail account. Of course, they just looked at the from: address, which had been 'spoofed', if you would even call it that (it's unprotected really). It was a well known virus at that time, so somebody will probably fill it in right below this message :)
These are for internet transactions (which can also be initiated from most post order companies over there, by a popular scheme called iDEAL). So you will have to enter the number manually. The device is not connected to the PC at all, which is a major selling point (no hardware or OS requirements, no chance of loosing your PIN).
I do think these things were introduced by the Dutch Rabobank. I also know that these devices are not made by RSA but by Vasco. Easy to check, their internet address is on the back.
Basically these things work by using a generated number by a clock inside the device hashed together with some authentication info (the number of your account). This number will stay valid for a limited number of time and cannot be generated without the device. More information? Just look up the patent numbers on the back of the device: US Patents 4.599.489 and 4.609.777 .
You certainly have to fill in the total amount of the "commit" (which can consist of multiple transactions. If you go even higher you will have to fill in yet another number. Previously, at least at the Rabobank, they forgot to tell you that the number you had to fill in was the total transaction amount, so there was less safety (if only one person picks up that the amount is incorrect you can still easily identify a bad transaction leading to the attacker.
It's pretty secure, but its not foolproof if you are not sure that the site you are visiting isn't that of the bank. E.g. an attacker could just tell you on a spoofed site that you would now have to enter the amount in cents instead of euros, and 99% of the people would fill in that amount without even thinking about it. Or the spoofer could just wait until you enter a sufficiently high amount, and send it to its account on the Bahama's.
"Other parts of the world are chipping into 'software' already, and there's no reason to think that we have some kind of automatic, natural, competitive advantage in any of those."
Yes, you have. The English language is one for instance. And then there is the economy of scale. Any really interesting conference on Java and Security is in the old US of A, and it is a pretty big drawback if you're thousands of miles from the city it is held in.
Even books on all the subjects are a lot more expensive than in the US, and don't forget the higher living conditions (it's easier to do your job properly if you don't have to put in so much energy in the rest of your life).
Of course, some of these advantages are declining a bit, but I don't think they will disappear completely. So now it is up to you to compete. Somehow I am pretty sure that if I'd apply for a job over there, it would be easy as pie.
Ah, yes, I forgot about the "not replying" to each other. This was because they didn't want to make it obvious the information came from the same source. Seems that this guy forgot about this important point.
:)
Anyway, I was not trying to say that this was the exact same thing, just that it reminded me of something. Clearly the guy doesn't have the romantic background of doing it because of trying to avoid age discrimination.
On Slashdot, I should be worried about age discrimination