All these SBC are nice, but I would really love one with two network interface. So far, my quest has not been successful. I'm not looking to route 1Gbps, but "normar" traffic. I know of the soekris & alix, but I would prefer an ARM based model. I was hoping to find an expansion board that would do the job, but still no luck.
Does anyone know any which would do the job ?
Conveniently, most home routers are ARM (sometimes MIPS) based embedded devices with plenty of NICs (100mb in the cheap seats, GbE doable). Find one with good OpenWRT support, ideally a USB port because they never come with enough storage, and you are ready to rock.
The usual nuisance is tracking down 'Unobtanium custom connector of inscrutable density' more than any real terror of pinout documentation. It's amazing how often the vendor fails to specify "We used connector X, you want connector Y to mate with it. Try Mouser catalog number X or Digikey catalog number Y, the part itself is a HiRose FOOPART in case you buy elsewhere." and how poor the search tools among part vendors(not just electrical, any sort of part that has two complementary pieces) can be for "I have a gizmo type XYZ. FFS, man, I want to give you my money! Just tell me what gizmo type XYZ mates with!"
It might run an older NES emulator at acceptable speed, or perhaps a native remake of the old game.
It probably depends more on what games you want, and how fanatical about fidelity you are. Reportedly, 100% timing-accurate simulation of even the humble NES is surprisingly computationally intensive, and nothing less will do to hunt down every last oddball that exploited weird edge conditions in the hardware and run it glitchlessly in its original form; but 'usually good enough' emulation was happening back when we were rocking PIII 450s on the desktop and worse than that on the laptop side.
Trust me... you're paying the Intel (TM) premium. Someone will do it better, cheaper. Just give it time.
You'd better be certain that you actually need all that power before you go and pay a little less than half as much for a Beaglebone black; but ~$100 is actually pretty much the going rate for a 'dev board' style arrangement with properly punchy ARM application processor, and those tend to have extra happy-fun-dicking-around-with-the-worst-graphics-driver-on-god's-black-earth, as opposed to 'Install Debian, have Intel's nonthrilling-but-endurable and in-kernel driver just work'.
The Intel Galileo seems like a product in pathetic search for purpose (can't bitbang even as well as a 16MHz AVR, rather more expensive than an arduino, weird and limited enough that the slightly less costly BB black or rPi is a better move, etc.); but this Minnowboard revision is markedly more compelling.
If you don't actually need that much power, you can get weaker-and-still-runs-full-linux ARM boards for about half that; but if you want a devboard (as opposed to hacking up some tightly integrated AllWinner SomethingSomething from ebay that may not even have serial debug headers), with a high end ARM application processor, you are looking at about $100 and not wildly dissimilar energy consumption.
If anything, the main competition (outside of space-constrained scenarios), is probably the (surprisingly aggressively priced) full bay trail motherboards (some other vendors as well). That will be a bit bigger, and you'll need a 24-pin PSU of some kind; but no need for expansion boards just to get PCIe/miniPCIe sockets, more I/O, and enough change to buy a low end arduino to substitute for the low-speed expansion.
I don't know if ARM scared intel good and hard, or if this is some price-dumping long game; but they appear to be practically giving 'Bay Trail' dice away.
'Cheap and nasty' = 'purchased and installed by amateurs trying to save money'. Down that path lies nothing good. Extra demerits are, of course, awarded to any vendor whose shitty 'cloud monitoring' service uPnPs like a madman trying to punch through whatever feeble pretense of security your equally crap router might have provided in order to be 'user friendly' and allow you to watch your house be burglarized from your smartphone or whatnot.
Trying to determine whether a series of hashing operations resulting in a mathematically valid bitcoin is like trying to determine whether or not a file is copyright-infringing by examining it with a hex editor.
Sure, I'd cry approximately -6 tears if the person behind this were to be caught and hauled off, and if he actually managed to mine anything(which would surprise me) I'd have no problem with the notion of his being forced to disburse the minings to his victims; but attempting to determine, from the results of a calculation, whether that calculation was conducted on a CPU not owned by the person who instructed the calculation to be performed is practically a category error. It just doesn't make sense.
If you have outside knowledge(like the arrest and conviction of the cracker), you can make inferences from that(and also use that as a basis for forcing him to disgorge the ill-gotten gains); but absent such additional information, a mathematical operation is what it is, there is no 'licitness' metadata.
I don't have an exact answer, Synology ARM devices vary, though mostly Marvell based; but for reference the 'Sheevaplug', based on a slightly obsolete Marvell storage processor, is quoted to be good for about.2 megahashes/second. I don't keep up with bitcoin difficulties, just don't care that much; but with USB-stick ASIC devices claiming 1 gigahash/second and greater, you'd need to own a mind-boggling number of these things to make it worth the time.
It's doubly weird because NASes probably have some neat stuff stashed on them, and would also be natural hosts for some sort of 'super-sleazy-CDN' type project, which would be equally illicit but might actually be worth more than a lukewarm cup of instant coffee.
Hello folks, I think the 'virus free' honeymoon is over.
Maybe I'm just pessimistic; but I thought it had been a truism for some years that embedded linux, especially in the cheap seats, was a total clusterfuck: firmware never getting released at all, firmware getting released with exploits that were known before it was even built, loads of shoddy little hacks to get the product out the door, and so on.
If memory serves, most of Synology's non-intel NASes are Marvell based. Marvell's fastest device, in terms of general compute, is the MV78460. 4 cores, ARMv7, up to 1.6GHz. As documented here most Synology NASes ship with something slower than that.
For reference, a 1.6GHz 'Kirkwood' Marvell core is good for slightly under.2 meghashes/s. About half as fast as an Atom CPU, less than 1/4000th as fast as an AMD7970, and just plain embarassing compared to the ASICs that do most of the work these days. With devices that run on USB power alone pulling north of 1gighash/s, you could probably own every Synology ARM NAS in the first world and barely pay yourself for your time.
...by this?
I'm more surprised that we haven't seen reports of infected DVD and Blu-ray players whose only purpose is to seek out more powerful devices (PCs, smartphones) on peoples' networks to compromise and turn into bitcoin zombies. After all, it only takes a few people to come up with the exploits in the first place, and then 5kr1p7 k1dd13s can use the tools others have created.
The main surprise is just that it's worth the trouble. Synology's high end has a few systems built around notably undistinguished Xeons(more for ECC support than anything else, they don't use very speedy ones); but if this attack is built for ARM, you are talking the relative cheap seats. Probably kilohashes to low megahashes per second, depending on how much capacity you reserve for the intended function of the device.
Even free-as-in-stolen, you're telling me that the best use somebody can think of for a botnet of network attached storage devices is generating maybe as many hashes as one of those cheapo USB-stick ASICs, rather than, say, basking in juicy private data and massive stolen storage space?
You missed (or deliberately ignored, or agree with) the central untruth that made that concept suspect: Is dressing your soldiers up as civilians in voilation of the Geneva convention, and do such soldiers forfeit assorted protections afforded uniformed forces Yes. So far, so good.
If somebody isn't a soldier and commits a crime in civilian cloths (as civilians are wont to do), can you argue that he violated the Geneva convention? Hardly, it only applies to soldiers.
That is the big lie of 'illegal enemy combatants'. Had the State of Terrorstan actually sent disguised soliders in, the illegality of their activity under the Convention would be cut and dry. No such state exists. Instead, the US decided to apply the standards of the Geneva Convention (selectively) to certain non-state actors who, being non-state, didn't act in uniform, in order to keep them in limbo between the protections afforded civilians and the protections afforded regular soldiers.
I'm well aware that the Geneva Convention takes a dim view of spy types; but it is wholly orthogonal to the treatment of civilian criminals, no matter how noxious.
The examples I provided aren't really 'plausible deniability'; but 'distinction without difference', which seems to be exactly the same thing that the Japanese are doing here.
Anyone who falls into that belief might as well be written off and put up against the wall, second in line to the people who believe that their own possession of arbitrary power is the only way to ensure the nation's safety. They can go first.
If I remember correctly, 'the metaverse' in Snow Crash was a VR environment with design and governance very similar to the internet's. Having it administered as a single corporation's private walled garden would be even more suitable to the flavor of the book; but it probably says something that, at least at that time, Stephenson either failed to think of that or considered it unsuitably implausible.
By the same token, though, I'd have to imagine that European law enforcement types have(formally or informally) had to adapt to the fact that "Eh, just report the details to border control and call it a day" doesn't work anywhere in the Schengen Area anymore. Most of Europe also has suitably compatible cellular operations, so I'd feel about as safe across a European 'national' border as I would across a US 'state' one.
Either way, once you get 'outside the country' whether literally (US) or figuratively(the parts of Europe that actually play nicely with each other) and strip the phone-home features, you probably do have a saleable product; but definitely a shady one. Tesla has never been secretive about (indeed, they consider it part of their customer service) the fact that vehicles phone home to report issues, assist in necessary maintenance, etc. so selling a suitably de-fanged Tesla would probably be more like selling enterprise network gear that mysteriously lacks any warranty entitlements, valid serial numbers, etc. than it would be like selling an ordinary 'used car of dubious provenance'.
For the moment, I'd imagine that the better money is either something mundane but trivially flippable/partable that you can do in relative volume with limited expertise, or bespoke hits on very high value stuff. Now, give it another few years, maybe a decade, and you'll have a number of legitimate Tesla buyers who no longer have warranty coverage and are looking for a good deal on a new battery/other part. That could change the equation.
In this litigious society, who can blame them. You can damn near guarantee that they'd have hit one bad jar in a lot that large and gotten the tar sued out of them. If you want to fix this situation and make sure it never happens again, demand tort reform in this country.
What I find strange is not that the product would be destroyed; but that it would be landfilled. Peanut butter, with minor differences by formulation and whatnot, is largely peanut oil with more or less finely ground nut fragments, all in an ostensibly-safe-for-food-contact plastic shell. Sounds like the sort of thing that a coal plant wouldn't even notice if you dropped a case or two into each rail car full of fuel, and a few free BTUs.
You seem to be making the optimistic assumption that the facebook integration will go no deeper than the logo... They may or may not be motivated to do so; but if they fancy making the hardware next to useless without talking to their mothership, that's totally doable.
What will be interesting to see(and I'm honestly not sure, so this is a question to anyone who might know as much as it is speculation) will be whether the bar for what facebook wants will turn out to be lower, higher, or just different from what game developers want.
If they want to do something 'social', it wouldn't be a huge surprise to see much greater emphasis on accurately capturing facial expressions of the wearer, so that those can be painted on an avatar of some kind(and emotional responses to advertisements gauged, naturally). On the other hand, they might be much less concerned with whether everything works properly when you whip round because you think you hear a zombie/xenomorph/commie coming up behind you.
We know that, in-engine, plausible ballistics is peanuts compared to plausible dialog or animating characters that don't break absurdly from time to time. What will the outcome be here? Is covering the 'gamer' use cases a subset of whatever dark plans Facebook has? Is it a (likely to be neglected) superset? Are they two somewhat divergent areas of emphasis?
I assume that certain baseline specs are more or less universal if you want a VR system that won't have the user heaving up their guts within the hour; but that there is nontrivial room for variation/quality level beyond that, and Facebook's intentions might not match those of gamers terribly strongly.
And color me surprised that the guy who just got a $2B cash infusion thinks this is a good idea. Gaming is barely a blip on the radar? Yeah, that's the problem, you asshole. Gamers bankrolled it; developers kept the momentum going; anticipated titles created the buzz.... but all of that barely registers against Facebook's piles and piles of money, right?
There's also the fact that, pending some sort of sea change, Facebook is kind of an engorged pustule on the ass of gaming. 'Social' gaming (ironically, usually rather less actually social than 'non-social' multiplayer games of almost any genre) is a horrible place where lousiness that would make the bargain bin of old blush is combined with ruthless exploitation of human weakness that would make your local heroin dealer a trifle queasy.
That's just not very encouraging. Though, that said, given that 'facebook' and 'gaming' have never been anything other than a mess, perhaps the best we can hope for is that they'll mostly ignore it/hack together a dubiously successful Second Life/Playstation Home clone and then focus on synergizing the graphic-centric virtual friendspace or some nonsense, and provide benign neglect elsewhere.
Slashdot Mirror
If product is 'My' cloud, why is your failure causing me these issues?
All these SBC are nice, but I would really love one with two network interface. So far, my quest has not been successful. I'm not looking to route 1Gbps, but "normar" traffic. I know of the soekris & alix, but I would prefer an ARM based model. I was hoping to find an expansion board that would do the job, but still no luck.
Does anyone know any which would do the job ?
Conveniently, most home routers are ARM (sometimes MIPS) based embedded devices with plenty of NICs (100mb in the cheap seats, GbE doable). Find one with good OpenWRT support, ideally a USB port because they never come with enough storage, and you are ready to rock.
The usual nuisance is tracking down 'Unobtanium custom connector of inscrutable density' more than any real terror of pinout documentation. It's amazing how often the vendor fails to specify "We used connector X, you want connector Y to mate with it. Try Mouser catalog number X or Digikey catalog number Y, the part itself is a HiRose FOOPART in case you buy elsewhere." and how poor the search tools among part vendors(not just electrical, any sort of part that has two complementary pieces) can be for "I have a gizmo type XYZ. FFS, man, I want to give you my money! Just tell me what gizmo type XYZ mates with!"
It might run an older NES emulator at acceptable speed, or perhaps a native remake of the old game.
It probably depends more on what games you want, and how fanatical about fidelity you are. Reportedly, 100% timing-accurate simulation of even the humble NES is surprisingly computationally intensive, and nothing less will do to hunt down every last oddball that exploited weird edge conditions in the hardware and run it glitchlessly in its original form; but 'usually good enough' emulation was happening back when we were rocking PIII 450s on the desktop and worse than that on the laptop side.
Trust me ... you're paying the Intel (TM) premium. Someone will do it better, cheaper. Just give it time.
You'd better be certain that you actually need all that power before you go and pay a little less than half as much for a Beaglebone black; but ~$100 is actually pretty much the going rate for a 'dev board' style arrangement with properly punchy ARM application processor, and those tend to have extra happy-fun-dicking-around-with-the-worst-graphics-driver-on-god's-black-earth, as opposed to 'Install Debian, have Intel's nonthrilling-but-endurable and in-kernel driver just work'.
The Intel Galileo seems like a product in pathetic search for purpose (can't bitbang even as well as a 16MHz AVR, rather more expensive than an arduino, weird and limited enough that the slightly less costly BB black or rPi is a better move, etc.); but this Minnowboard revision is markedly more compelling.
If you don't actually need that much power, you can get weaker-and-still-runs-full-linux ARM boards for about half that; but if you want a devboard (as opposed to hacking up some tightly integrated AllWinner SomethingSomething from ebay that may not even have serial debug headers), with a high end ARM application processor, you are looking at about $100 and not wildly dissimilar energy consumption.
If anything, the main competition (outside of space-constrained scenarios), is probably the (surprisingly aggressively priced) full bay trail motherboards (some other vendors as well). That will be a bit bigger, and you'll need a 24-pin PSU of some kind; but no need for expansion boards just to get PCIe/miniPCIe sockets, more I/O, and enough change to buy a low end arduino to substitute for the low-speed expansion.
I don't know if ARM scared intel good and hard, or if this is some price-dumping long game; but they appear to be practically giving 'Bay Trail' dice away.
'Cheap and nasty' = 'purchased and installed by amateurs trying to save money'. Down that path lies nothing good. Extra demerits are, of course, awarded to any vendor whose shitty 'cloud monitoring' service uPnPs like a madman trying to punch through whatever feeble pretense of security your equally crap router might have provided in order to be 'user friendly' and allow you to watch your house be burglarized from your smartphone or whatnot.
Trying to determine whether a series of hashing operations resulting in a mathematically valid bitcoin is like trying to determine whether or not a file is copyright-infringing by examining it with a hex editor.
Sure, I'd cry approximately -6 tears if the person behind this were to be caught and hauled off, and if he actually managed to mine anything(which would surprise me) I'd have no problem with the notion of his being forced to disburse the minings to his victims; but attempting to determine, from the results of a calculation, whether that calculation was conducted on a CPU not owned by the person who instructed the calculation to be performed is practically a category error. It just doesn't make sense.
If you have outside knowledge(like the arrest and conviction of the cracker), you can make inferences from that(and also use that as a basis for forcing him to disgorge the ill-gotten gains); but absent such additional information, a mathematical operation is what it is, there is no 'licitness' metadata.
I don't have an exact answer, Synology ARM devices vary, though mostly Marvell based; but for reference the 'Sheevaplug', based on a slightly obsolete Marvell storage processor, is quoted to be good for about .2 megahashes/second. I don't keep up with bitcoin difficulties, just don't care that much; but with USB-stick ASIC devices claiming 1 gigahash/second and greater, you'd need to own a mind-boggling number of these things to make it worth the time.
It's doubly weird because NASes probably have some neat stuff stashed on them, and would also be natural hosts for some sort of 'super-sleazy-CDN' type project, which would be equally illicit but might actually be worth more than a lukewarm cup of instant coffee.
It's JVMs all the way down. Except for the one that's actually Dalvik and willing to go head-to-head with Oracle to prove it.
and Synology devices run ... LINUX.
Hello folks, I think the 'virus free' honeymoon is over.
Maybe I'm just pessimistic; but I thought it had been a truism for some years that embedded linux, especially in the cheap seats, was a total clusterfuck: firmware never getting released at all, firmware getting released with exploits that were known before it was even built, loads of shoddy little hacks to get the product out the door, and so on.
If memory serves, most of Synology's non-intel NASes are Marvell based. Marvell's fastest device, in terms of general compute, is the MV78460. 4 cores, ARMv7, up to 1.6GHz. As documented here most Synology NASes ship with something slower than that.
.2 meghashes/s. About half as fast as an Atom CPU, less than 1/4000th as fast as an AMD7970, and just plain embarassing compared to the ASICs that do most of the work these days. With devices that run on USB power alone pulling north of 1gighash/s, you could probably own every Synology ARM NAS in the first world and barely pay yourself for your time.
For reference, a 1.6GHz 'Kirkwood' Marvell core is good for slightly under
...by this? I'm more surprised that we haven't seen reports of infected DVD and Blu-ray players whose only purpose is to seek out more powerful devices (PCs, smartphones) on peoples' networks to compromise and turn into bitcoin zombies. After all, it only takes a few people to come up with the exploits in the first place, and then 5kr1p7 k1dd13s can use the tools others have created.
The main surprise is just that it's worth the trouble. Synology's high end has a few systems built around notably undistinguished Xeons(more for ECC support than anything else, they don't use very speedy ones); but if this attack is built for ARM, you are talking the relative cheap seats. Probably kilohashes to low megahashes per second, depending on how much capacity you reserve for the intended function of the device.
Even free-as-in-stolen, you're telling me that the best use somebody can think of for a botnet of network attached storage devices is generating maybe as many hashes as one of those cheapo USB-stick ASICs, rather than, say, basking in juicy private data and massive stolen storage space?
You missed (or deliberately ignored, or agree with) the central untruth that made that concept suspect: Is dressing your soldiers up as civilians in voilation of the Geneva convention, and do such soldiers forfeit assorted protections afforded uniformed forces Yes. So far, so good.
If somebody isn't a soldier and commits a crime in civilian cloths (as civilians are wont to do), can you argue that he violated the Geneva convention? Hardly, it only applies to soldiers.
That is the big lie of 'illegal enemy combatants'. Had the State of Terrorstan actually sent disguised soliders in, the illegality of their activity under the Convention would be cut and dry. No such state exists. Instead, the US decided to apply the standards of the Geneva Convention (selectively) to certain non-state actors who, being non-state, didn't act in uniform, in order to keep them in limbo between the protections afforded civilians and the protections afforded regular soldiers.
I'm well aware that the Geneva Convention takes a dim view of spy types; but it is wholly orthogonal to the treatment of civilian criminals, no matter how noxious.
The examples I provided aren't really 'plausible deniability'; but 'distinction without difference', which seems to be exactly the same thing that the Japanese are doing here.
Anyone who falls into that belief might as well be written off and put up against the wall, second in line to the people who believe that their own possession of arbitrary power is the only way to ensure the nation's safety. They can go first.
"Oh, hey, I think the toy computers are fighting again..."
-Servers, Compute Clusters, Storage Systems, et al.
Continue the noble science against UN oppression with the Cetacean Research Simulator!
As opposed to those other countries, that haven't invented 'using euphemisms to evade established law' yet?
I'll be right back, the illegal enemy combatants in administrative detention are causing trouble again.
If I remember correctly, 'the metaverse' in Snow Crash was a VR environment with design and governance very similar to the internet's. Having it administered as a single corporation's private walled garden would be even more suitable to the flavor of the book; but it probably says something that, at least at that time, Stephenson either failed to think of that or considered it unsuitably implausible.
Can you imagine what that would do to Second Life's core 'BDSM Furries' demographic?
By the same token, though, I'd have to imagine that European law enforcement types have(formally or informally) had to adapt to the fact that "Eh, just report the details to border control and call it a day" doesn't work anywhere in the Schengen Area anymore. Most of Europe also has suitably compatible cellular operations, so I'd feel about as safe across a European 'national' border as I would across a US 'state' one.
Either way, once you get 'outside the country' whether literally (US) or figuratively(the parts of Europe that actually play nicely with each other) and strip the phone-home features, you probably do have a saleable product; but definitely a shady one. Tesla has never been secretive about (indeed, they consider it part of their customer service) the fact that vehicles phone home to report issues, assist in necessary maintenance, etc. so selling a suitably de-fanged Tesla would probably be more like selling enterprise network gear that mysteriously lacks any warranty entitlements, valid serial numbers, etc. than it would be like selling an ordinary 'used car of dubious provenance'.
For the moment, I'd imagine that the better money is either something mundane but trivially flippable/partable that you can do in relative volume with limited expertise, or bespoke hits on very high value stuff. Now, give it another few years, maybe a decade, and you'll have a number of legitimate Tesla buyers who no longer have warranty coverage and are looking for a good deal on a new battery/other part. That could change the equation.
In this litigious society, who can blame them. You can damn near guarantee that they'd have hit one bad jar in a lot that large and gotten the tar sued out of them. If you want to fix this situation and make sure it never happens again, demand tort reform in this country.
What I find strange is not that the product would be destroyed; but that it would be landfilled. Peanut butter, with minor differences by formulation and whatnot, is largely peanut oil with more or less finely ground nut fragments, all in an ostensibly-safe-for-food-contact plastic shell. Sounds like the sort of thing that a coal plant wouldn't even notice if you dropped a case or two into each rail car full of fuel, and a few free BTUs.
You seem to be making the optimistic assumption that the facebook integration will go no deeper than the logo... They may or may not be motivated to do so; but if they fancy making the hardware next to useless without talking to their mothership, that's totally doable.
What will be interesting to see(and I'm honestly not sure, so this is a question to anyone who might know as much as it is speculation) will be whether the bar for what facebook wants will turn out to be lower, higher, or just different from what game developers want.
If they want to do something 'social', it wouldn't be a huge surprise to see much greater emphasis on accurately capturing facial expressions of the wearer, so that those can be painted on an avatar of some kind(and emotional responses to advertisements gauged, naturally). On the other hand, they might be much less concerned with whether everything works properly when you whip round because you think you hear a zombie/xenomorph/commie coming up behind you.
We know that, in-engine, plausible ballistics is peanuts compared to plausible dialog or animating characters that don't break absurdly from time to time. What will the outcome be here? Is covering the 'gamer' use cases a subset of whatever dark plans Facebook has? Is it a (likely to be neglected) superset? Are they two somewhat divergent areas of emphasis?
I assume that certain baseline specs are more or less universal if you want a VR system that won't have the user heaving up their guts within the hour; but that there is nontrivial room for variation/quality level beyond that, and Facebook's intentions might not match those of gamers terribly strongly.
And color me surprised that the guy who just got a $2B cash infusion thinks this is a good idea. Gaming is barely a blip on the radar? Yeah, that's the problem, you asshole. Gamers bankrolled it; developers kept the momentum going; anticipated titles created the buzz.... but all of that barely registers against Facebook's piles and piles of money, right?
There's also the fact that, pending some sort of sea change, Facebook is kind of an engorged pustule on the ass of gaming. 'Social' gaming (ironically, usually rather less actually social than 'non-social' multiplayer games of almost any genre) is a horrible place where lousiness that would make the bargain bin of old blush is combined with ruthless exploitation of human weakness that would make your local heroin dealer a trifle queasy.
That's just not very encouraging. Though, that said, given that 'facebook' and 'gaming' have never been anything other than a mess, perhaps the best we can hope for is that they'll mostly ignore it/hack together a dubiously successful Second Life/Playstation Home clone and then focus on synergizing the graphic-centric virtual friendspace or some nonsense, and provide benign neglect elsewhere.