You ought to go talk to amazon, especially if you have proof (even if it is affadavits from friends who saw it). I'm sure they'd be happy to hear from you.
I worked on a government funded heap of crap.. er I mean "cutting edge system" for a while.
It was very very bad, luckily I was just a drone, but it was obvious how lousy it was (apart from to the people at the top).
It wasn't a total waste, I made a reasonable wage at the time, and spent a lot of money on food and drink, and having fun, after all that's what the world is about.
Whenever fraud etc is discovered in a foreign government (or the EU and so on), British Politicians ridicule the foreign government saying how corrupt it is, and how much superior the British Government is. But everyone on the ground sees through their false words. Was there any meaningful quotation process for this work? Unlikely, even if it went to quotation, the requirements were probably highly stacked in favour of the chosen system.
Public job adverts are seen as a way of avoiding a legal challenge over employing the cousin/friend who they really want to get the job.
This just continues and continues unchallenged. The media are just as rotten.
Someone needs to give Britain a good sorting out about all this, and it would be good for the country, because it is, with inefficiency piled on inefficiency, performance is lousy. Without unprincipled UK arms sales, UK GDP would be non-existent.
I seriously doubt that Cray can put faster circuits down on silicon that Intel. part of the nature of the silicon foundry is that stuff doesn't start getting good and fast unless you make A LOT of it. It also gets cheap at this point.
I see no architectural difference between a "cluster" and a "supercomputer". The links between different CPUs are just conventionally made using different technology.
There's a lot of rubbishing of PCI (hey it's 10 years old now, and there are MUCH faster new versions happening), and what is the point of saying unquantified/unsubstantiated crap like "CRAYS HAVE VERY FAST SHARED MEMORY BUS".
Yeah - HOW FAST THEN? I'd be surprised if they are 128 bit running at 2 GHz.
Shared memory can mean one of a number of things, also: You can have one CPU sharing say a 4 meg block with each of 25 other CPUs. The first CPU acts as the hub for communication between the other CPUs. You could have 27 CPUS in a 3 x 3 x 3 cube, each CPU sharing memory with up to 6 neighbours. You could have 5 processors in a line with each one sharing memory with (up to 2) neighbours.
Or you could have a bunch of core memory that 4 processors share (they might have their own memory too).
The same thing goes for a cluster - you could have PCs with up to 6 network cards (or with unidirectional custom ethernet protocol, even 12 network cards linking to neighbours in a 27 CPU cube, and so on.
The topology will affect how the program is written for maximum speed, but also which tasks the computer is suited for. I think you could make very very fast links between ordinary PCs with say full duplex gigabit running a custom protocol (TCP has latency by the way, UDP has none since it doesn't wait to assemble packets in buffers in the kernel).
It's hard to imagine a task that is so i/o bound (in my mind this is the opposite of embarrasingly parallel problems) as to require more than 100 megabytes/second between each node, when each CPU node has a memory bandwidth of 12 gigabytes per second (based on 32 bit core of Pentium 4 at 3 GHz, assuming roughly 1 transfer per clock cycle, which in itself is unlikely).
In other words, a cluster using off the shelf gigabit ethernet hardware could transfer 1% as much data as the CPU could do with RAM. Note if the CPU is in a 27 CPU cube the combined 6 gigabit ether cards would be transferring 6% as much as the CPU could. I guess it is possible to get motherboards with larger numbers of PCI slots, say 12 in which case you could run two streams of gigabit ethernet between each CPU giving you 12% as much data being transferred over ethernet as the CPU can transfer in and out of memory (not including cache flushing from CPU to RAM).
Once again, what problems require such a huge amount of communication with other nodes that say 12% as much bandwidth between nodes versus CPU-memory is not sufficient? Say 12% isn't high enough: what CPUs, data bus widths, and shared memory speeds are used then?
Arguments people have made so far are so light on detail, and using terms like "much faster" instead of giving a figure, it sounds like FUD.
Remember parallel links between devices on chips can exhibit data skew, lowering data rate compared with a fast serial link. In fact there is talk (and I personally suggested a long time on a newsgroup) using light to get signals from one chip to another. (probably mainly serial, but not necessarily exclusively).
At least that's what I thought. Asymmetric keys are far weaker that symmetric keys. Typically people use asymmetric keys because they think distribution of quantities of OTP symmetric keys is hard. This is no harder than your bank sending out a smart card instead of your credit card. Indeed by putting a CPU on the smartcard, security of the platform would not compromise the table of keys either. I think far too much trust is put in Organisational root certifiers, and instead all keys should be instantly discardable, and trust established via a connection to a TrustProvider. (both parties connect to the trust provider, and exchange a secret through the TrustProvider, this can be a Bank, Your organisation server, or on a peer to peer basis. If it seems like anything has been compromised, you can quickly dump all keys, re-issue new ones, and trust is re-established. If you rely on less hard-to-break keys, and use them widely across numerous clients, it is very hard to issue a new org root certificate. If your org root certifier for example, in a Lotus Notes system (many countries, many servers) is compromised, then it is extremely hard to re-establish trust.
Re: ease of cracking asymmetric keys. Because so few values in the solution space are possible solutions, they are relatively easy to crack as you only have to brute force the possible solutions, not all possible values. Indeed, custom chips are made by the US government in the basement of the giant NSA building that crack 5000 per second. Imagine a 19 inch rack full of 200 of these, 6 foot tall times a few. You could crack millions per second. However, a terrorist using a dice and a piece of paper and pencil could defeat this asymmetric-key-breaking technology (OTP symmetric key). [so who are we kidding].
1. I've never heard ANYONE claim that Asymmetric Key is harder to break than Symmetric key of same length. Can you provide a reference (obviously all the books I've read on the subject are wrong, as is my view of the axiomatic). 2. In star trek maybe you can store bits and them take no power, however, in reality a bit in a ram chip takes a constant stream of energy to keep stored. Same thing goes for stuff stored on magnetic form (eg hard disks) since the domain degrades and needs to be re-written, plus fairly low MTBF on drives means you really want a RAID5 which takes energy to run.
To pre-answer any questions about how does partyA trust partyB, the answer is both trust each other by secret exchange through the Key Maintainer. In this case, the Organisation that issues CompactFlash cards full of OTPs (ie Checkbooks in man-in-street-speak) to partyA and partyB. Both contact Key Maintainer to exchange secrets. Secrets are distributed, then maybe smaller OTP used direct partyA to partyB, or maybe a traffic through KeyMaintainer.
With a perfect PRBS generator. However, there are far fewer solutions for Asymmetric keys. You also state 'Joules' to store a bit. Over what period are you considering this? Since a few transistors on a chip will consume a given amount of energy per second, you must integrate wrt time in order to get a total amount of 'work'. Therefore you must have a time period in mind (for the storage or each key combination), which you should state in your brute force calculation.
As OneTimePad with a perfect PRBS (PseudoRandomBinarySequence) generator is mathematically unbreakable (with minimised redundancy and no checksum etc), why not use this instead of DES and all the weaker Asymmetric technologies? The reasons quoted (for not using OTP) are always key distribution problems, and yet we routinely get sent periodically Credit Cards and Checkbooks. (this could just as easily be a 'book' of OTPs on a CompactFlash card. Personally in addition to OTP, I'd spray randomly and using random inversions, the data into a larger packet (to counter weaknesses in the PRBS) and also perhaps send dummy packets to. The only way for verification (like checksums) to work, is to send data at least twice before it is trusted, using different OTPs. It seems to me the whole trust based on asymmetric signed certificate is flawed since too much trust is put in a common root certificate which must be shared across millions of clients for a long time. And if it's cracked, how do you deploy new certs? Nightmare. (I've been in this situation with Lotus Notes common root certificate - it really isn't fun). My idea puts no trust in any certificate beyond the current message. Pads are deleted either end (the server has copies of all pads for all users). If the end user thinks his compact flash is compromised, all pads are deleted either end, fresh ones made (shipped by UPS or whatever in a CompactFlash) and trust is instantly re-established. If the server is cracked, this is no worse than the situation for Asymmetric Keys.
The way to protect your Linux system: Log in as a user with ability to write only to user directory. Use a Journaling type filing system that writes prior versions of modified files to a backup area or device. (Can Xfs do this?). Alternatively, allow user to only write but not delete these data files, and use some kind of "take tidy" style batch script so you can delete files by notifying a process that has ability to delete user files indirectly (perhaps with PIN access). Anyway, as I have never lost any user data in the years I have been using Linux (unlike in Windows), to date it just isn;t a problem, so less FUD please. Jeff Davies
Slashdot Mirror
and infinity even harder
You ought to go talk to amazon, especially if you have proof (even if it is affadavits from friends who saw it).
I'm sure they'd be happy to hear from you.
I worked on a government funded heap of crap .. er I mean "cutting edge system" for a while.
It was very very bad, luckily I was just a drone, but it was obvious how lousy it was (apart from to the people at the top).
It wasn't a total waste, I made a reasonable wage at the time, and spent a lot of money on food and drink, and having fun, after all that's what the world is about.
Being a buddhist helps by the way.
to me the existence of infinite primes seems obvious.
Some people find inifinity troublesome.
Whenever fraud etc is discovered in a foreign government (or the EU and so on), British Politicians ridicule the foreign government saying how corrupt it is, and how much superior the British Government is.
But everyone on the ground sees through their false words. Was there any meaningful quotation process for this work? Unlikely, even if it went to quotation, the requirements were probably highly stacked in favour of the chosen system.
Public job adverts are seen as a way of avoiding a legal challenge over employing the cousin/friend who they really want to get the job.
This just continues and continues unchallenged. The media are just as rotten.
Someone needs to give Britain a good sorting out about all this, and it would be good for the country, because it is, with inefficiency piled on inefficiency, performance is lousy. Without unprincipled UK arms sales, UK GDP would be non-existent.
check out http://sourceforge.net/projects/securekey
Its a one time pad with randomised dictionary entries. I really really can't see how this can be broken, no matter how many ASICs you use.
I seriously doubt that Cray can put faster circuits down on silicon that Intel. part of the nature of the silicon foundry is that stuff doesn't start getting good and fast unless you make A LOT of it. It also gets cheap at this point.
I see no architectural difference between a "cluster" and a "supercomputer". The links between different CPUs are just conventionally made using different technology.
There's a lot of rubbishing of PCI (hey it's 10 years old now, and there are MUCH faster new versions happening), and what is the point of saying unquantified/unsubstantiated crap like "CRAYS HAVE VERY FAST SHARED MEMORY BUS".
Yeah - HOW FAST THEN? I'd be surprised if they are 128 bit running at 2 GHz.
Shared memory can mean one of a number of things, also:
You can have one CPU sharing say a 4 meg block with each of 25 other CPUs. The first CPU acts as the hub for communication between the other CPUs.
You could have 27 CPUS in a 3 x 3 x 3 cube, each CPU sharing memory with up to 6 neighbours.
You could have 5 processors in a line with each one sharing memory with (up to 2) neighbours.
Or you could have a bunch of core memory that 4 processors share (they might have their own memory too).
The same thing goes for a cluster - you could have PCs with up to 6 network cards (or with unidirectional custom ethernet protocol, even 12 network cards linking to neighbours in a 27 CPU cube, and so on.
The topology will affect how the program is written for maximum speed, but also which tasks the computer is suited for. I think you could make very very fast links between ordinary PCs with say full duplex gigabit running a custom protocol (TCP has latency by the way, UDP has none since it doesn't wait to assemble packets in buffers in the kernel).
It's hard to imagine a task that is so i/o bound (in my mind this is the opposite of embarrasingly parallel problems) as to require more than 100 megabytes/second between each node, when each CPU node has a memory bandwidth of 12 gigabytes per second (based on 32 bit core of Pentium 4 at 3 GHz, assuming roughly 1 transfer per clock cycle, which in itself is unlikely).
In other words, a cluster using off the shelf gigabit ethernet hardware could transfer 1% as much data as the CPU could do with RAM.
Note if the CPU is in a 27 CPU cube the combined 6 gigabit ether cards would be transferring 6% as much as the CPU could. I guess it is possible to get motherboards with larger numbers of PCI slots, say 12 in which case you could run two streams of gigabit ethernet between each CPU giving you 12% as much data being transferred over ethernet as the CPU can transfer in and out of memory (not including cache flushing from CPU to RAM).
Once again, what problems require such a huge amount of communication with other nodes that say 12% as much bandwidth between nodes versus CPU-memory is not sufficient?
Say 12% isn't high enough: what CPUs, data bus widths, and shared memory speeds are used then?
Arguments people have made so far are so light on detail, and using terms like "much faster" instead of giving a figure, it sounds like FUD.
Remember parallel links between devices on chips can exhibit data skew, lowering data rate compared with a fast serial link. In fact there is talk (and I personally suggested a long time on a newsgroup) using light to get signals from one chip to another. (probably mainly serial, but not necessarily exclusively).
At least that's what I thought.
Asymmetric keys are far weaker that symmetric keys. Typically people use asymmetric keys because they think distribution of quantities of OTP symmetric keys is hard. This is no harder than your bank sending out a smart card instead of your credit card. Indeed by putting a CPU on the smartcard, security of the platform would not compromise the table of keys either.
I think far too much trust is put in Organisational root certifiers, and instead all keys should be instantly discardable, and trust established via a connection to a TrustProvider. (both parties connect to the trust provider, and exchange a secret through the TrustProvider, this can be a Bank, Your organisation server, or on a peer to peer basis. If it seems like anything has been compromised, you can quickly dump all keys, re-issue new ones, and trust is re-established. If you rely on less hard-to-break keys, and use them widely across numerous clients, it is very hard to issue a new org root certificate. If your org root certifier for example, in a Lotus Notes system (many countries, many servers) is compromised, then it is extremely hard to re-establish trust.
Re: ease of cracking asymmetric keys. Because so few values in the solution space are possible solutions, they are relatively easy to crack as you only have to brute force the possible solutions, not all possible values. Indeed, custom chips are made by the US government in the basement of the giant NSA building that crack 5000 per second. Imagine a 19 inch rack full of 200 of these, 6 foot tall times a few. You could crack millions per second. However, a terrorist using a dice and a piece of paper and pencil could defeat this asymmetric-key-breaking technology (OTP symmetric key). [so who are we kidding].
1. I've never heard ANYONE claim that Asymmetric Key is harder to break than Symmetric key of same length. Can you provide a reference (obviously all the books I've read on the subject are wrong, as is my view of the axiomatic).
2. In star trek maybe you can store bits and them take no power, however, in reality a bit in a ram chip takes a constant stream of energy to keep stored. Same thing goes for stuff stored on magnetic form (eg hard disks) since the domain degrades and needs to be re-written, plus fairly low MTBF on drives means you really want a RAID5 which takes energy to run.
To pre-answer any questions about how does partyA trust partyB, the answer is both trust each other by secret exchange through the Key Maintainer. In this case, the Organisation that issues CompactFlash cards full of OTPs (ie Checkbooks in man-in-street-speak) to partyA and partyB.
Both contact Key Maintainer to exchange secrets. Secrets are distributed, then maybe smaller OTP used direct partyA to partyB, or maybe a traffic through KeyMaintainer.
With a perfect PRBS generator. However, there are far fewer solutions for Asymmetric keys. You also state 'Joules' to store a bit. Over what period are you considering this? Since a few transistors on a chip will consume a given amount of energy per second, you must integrate wrt time in order to get a total amount of 'work'. Therefore you must have a time period in mind (for the storage or each key combination), which you should state in your brute force calculation.
As OneTimePad with a perfect PRBS (PseudoRandomBinarySequence) generator is mathematically unbreakable (with minimised redundancy and no checksum etc), why not use this instead of DES and all the weaker Asymmetric technologies?
The reasons quoted (for not using OTP) are always key distribution problems, and yet we routinely get sent periodically Credit Cards and Checkbooks. (this could just as easily be a 'book' of OTPs on a CompactFlash card.
Personally in addition to OTP, I'd spray randomly and using random inversions, the data into a larger packet (to counter weaknesses in the PRBS) and also perhaps send dummy packets to. The only way for verification (like checksums) to work, is to send data at least twice before it is trusted, using different OTPs.
It seems to me the whole trust based on asymmetric signed certificate is flawed since too much trust is put in a common root certificate which must be shared across millions of clients for a long time. And if it's cracked, how do you deploy new certs? Nightmare. (I've been in this situation with Lotus Notes common root certificate - it really isn't fun). My idea puts no trust in any certificate beyond the current message. Pads are deleted either end (the server has copies of all pads for all users). If the end user thinks his compact flash is compromised, all pads are deleted either end, fresh ones made (shipped by UPS or whatever in a CompactFlash) and trust is instantly re-established.
If the server is cracked, this is no worse than the situation for Asymmetric Keys.
The way to protect your Linux system: Log in as a user with ability to write only to user directory. Use a Journaling type filing system that writes prior versions of modified files to a backup area or device. (Can Xfs do this?). Alternatively, allow user to only write but not delete these data files, and use some kind of "take tidy" style batch script so you can delete files by notifying a process that has ability to delete user files indirectly (perhaps with PIN access). Anyway, as I have never lost any user data in the years I have been using Linux (unlike in Windows), to date it just isn;t a problem, so less FUD please. Jeff Davies