shouldn't you expect everything you do in public to be potentially monitored? yes, the scale of modern life has, until recently, made most activity relatively anonymous, but only because no one bothered to look. I'm not sure why we should be worried about this.
HOWEVER, we should make sure that this be done above-the-board. for instance, the activity of police in public is clearly also something that should be public, and thus legally recordable. what police do when they enter your property is up to you to record if you wish. and government records resulting from this kind of recordkeeping of the public need to be public records (accessible to anyone). (government can reasonably charge for access if some company wants to mine these records, but I should also be able to, for a nominal fee, ask whether any records exist of, say, vehicles speeding on my street.)
so there are lots of quad-copters around that have roughly similar specs. this one is a uni-copter with 8 thrust-vectoring flaps, which is, I guess somewhat novel. not sure why 8 is the right number, and seems like a fairly large number, given that each requires a servo and fairly big piece of material. but since the flaps are independent, they can provide both direction and rotational control (which is why a quad-copter needs 4 fans - and why a helicopter needs a tail fan.) the spherical cage (and uni-fan) makes it seem compact and tidy, but I'm not sure the layout is actually better than a quad-copter.
a big part of the problem is that we've permitted doctors to form a medical mafia, and what we need is something closer to "healthcare engineering". as a guild, the medical establishment is structurally prone to rackets, and the relationship between big pharma and doctors qualifies.
128 cores isn't enough to worry about - just install a distro you like and feel comfortable maintaining. although 128 cores isn't many, you should probably think about the style of install you want. lots of people seem to like diskful installs - afaikt purely because it's familiar. most significant clustering sites use diskless (NFS root) though, because it's so much easier to maintain. there's never any question of nodes getting out of sync. traffic due to NFS root is trivial. another best-practice is to configure 1-2 admin nodes (no users, provides NFS, scheduling and monitoring services), one or more dedicated login nodes, and discourage users from touching the compute nodes directly (among other things, give them non-routable addresses.) get or make a ticket system to keep track of user and system issues. monitor the heck out of your systems.
I'm an HPC center admin and system programmer, 10+ years. I think we've been in the top 50 several times.
no, we don't have any real info on how fast BD is yet. some technical papers on it indicate it's designed specifically for high clock while maintaining control on power dissipation. the shared FPU is somewhat faster than an unshared FPU would be, so this is a good choice, especially for code that's not always FPU-bound. and AMD has said that BD definitely has thermally-limited clock boosting.
tagging a picture is two separate acts: publishing the picture and publishing a statement (the tag). if I have published a legal picture of someone, I can tag it with anything non-libelous. not really anything new here, is there?
I'm sure these are very nice chips, but anyone can do similar, given funding. there are a number of cores available for licensing (like they did with MIPS), and adding vector units is the obvious way to boost your peak flops without blowing your power budget. I guess I don't really see why this merits all the coverage - for instance, what fraction of peak performance can it get on real code (say, a weather or MD simulation, not HPL)?
the quoted peak gflops/watt for this project are decent, but not much better than current commodity x86 parts, and comparable to GPUs. most architects in the field consider power-efficient computing to be a system-architecture challenge: how to move around all that data without spending all your power on fast/wide buses. a genuinely interesting new architecture would try to address this - perhaps something vaguely like IRAM. smart memory with some kind of high-order interconnect seems like the way to go, rather than putting giant vector units on a traditional design.
I would never interfere with someone else's fetish, but what is the point of getting excited about SSD speeds? how many people who buy them have any sort of thought-out reason to get them, rather than alternatives? what's your workload that works if you can write at 315 MB/s, but fails if you're limited to a measly 250?
in general, the SSD market seems driven by fetish, and that's just fine. the whole auto market is fetish-driven. and apple, too;)
what I wonder, though, is if there's someone out there designing, say, a complex website with load balancing, HA-failover, frontend/backend/storage specialization, etc who just needs a 315 MB/s storage device to win. it seems like current standard practice (inmemory nosql, fast private networks, etc) make pretty much any performance concern just a matter of turning the knob - usually just adding some nodes. but if that's the case, then there's little importance to 315 vs 250 MB/s.
The disturbing part is that Beck isn't going away - the Tea Party and related craziness has its paranoid hooks sunk deeply enough into American self-doubt that a little turn-the-other-cheek will take away their oxygen. (sorry about the metaphor pileup there.)
Debian is the socks-and-sandals distro, and this change lets them cast off that embarrassing reliance on the conformist, ideologically impure Linux kernel. WOO! great leap forward.
but aren't we all expecting a major change in patentability, especially wrt this kind of software-related topics? the current territory-denial (landmine) approach to patents is pretty doomed...
this sort of discussion, especially given the mental limitations of websites, media outlets and media consumers, always gets bogged down in the morass of non-free vs defacto vs dejure vs license pools, etc.
the fact is that when we talk about "standard" in a good sense, it is open _and_ free, or not at all. for instance TCP/IP is a real standard precisely because it is both open and free. HTTP/HTML. javascript is, but java isn't. linux is and windows isn't. you get the idea.
in a meaningful sense, GNU caused part of the problem, because it conflates RMS's ideology with this basically simple "standard=open+free" model. (and before any RMSista object: GPL is popular as a flag of convenience - operating under GPL does not mean allegiance to RMS's politics...)
seriously, why is this news? yes, naifs out there associate OS stacks with particular hardware platforms, but the people _inside_ software companies don't. msft has produced extensive app stacks for ppc before, even to some extent mips and alpha. current windows is derived through NT and NT OS/2 from a codebase that was _originally_ developed simultaneously on MIPS and x86. obviously for devices like phones and tablets, even a lot of desktops, there's no need to worry about externally-provided add-in cards and the driver complexities that introduces.
besides, who cares that much about native apps anymore? "appliancy" stuff is browser/flash/java-based. msft itself is probably the main purveyor of non-browser/flash/java stuff, and of course they can retarget their office suite, no sweat (hah).
the interesting question in this is whether there's really any reason for ARM to be more mobile-friendly - that is, what is it about the ISA or implementation that makes it _inherently_ better (if any). my suspicion is that it's mostly a matter of methodology or culture: ARM has traditionally been very parsimonious (think hybrid or high-efficiency car), and the x86 makers have traditionally been more Nascar. Intel seemed to have done Atom almost against its will or corporate culture - followons have been more power-efficient, but they hardly seem like significant, bet-the-company efforts. AMD's recent bobcat-based chips appear to be based on a modestly-tweaked version of the K8. maybe what distinguishes ARM is something simple, like a compact instruction encoding...
you seem to be conflating cloud with desktop virtualization. the network needs of cloud use depend primarily on the amount of data creation or mutation going on, not on the bandwidth of the UI. it's also not clear that apple would actually win by shifting to dumber/thinner clients, as opposed to the hardware they sell now, which are basically high-margin reformatted PCs.
what an out-of-touch twit. then again, UChi is fairly famous for "odd" faculty.
numeracy is what's needed: that people are comfortable with quantitative reasoning. the specific mathematical techniques are irrelevant, but yes, people really do need better ability to understand issues in a quantitative way. climate change, sub-prime mortgages, this week's discount on cans of soup, fluoridating water, innoculations and autism, the list goes on forever. you can't be a competent human without understanding conditional probability, for instance.
the education system does an incredibly poor job of this, producing adults who struggle and fail to find structure and place in a big, confusing world, and for lack of comfort with analytic, quantitative approaches, latch on to religious/emotional/ideological movements like the Tea Party.
My experience is that when IO affects "feel", it's mostly because of increased competition for memory. Someone else mentioned tuning the vm.dirty_* sysctls, which certainly helps, but what surprises me is that we don't use O_DIRECT and splice/vmsplice more. "cp" is still a loop of 32k read/writes, with only the obvious O_ modes on either descriptor (and needless to say, no fadvise either.)
In short, the kernel offers fine mechanisms to resolve these problems - it's user-space that isn't taking advantage of them.
Amazon is already a major e-retailer, so they have some advantage there. I think the real question is whether they'll add value to the user/customer's app-selection process. yes, Amazon already has some you-may-also-like, and user reviews, but can Google do a better job of mining such data to produce value? Amazon doesn't seem to take this as seriously as Netflicks, for instance. can Google obtain some advantage from, for instance, crash reporting? perhaps they're in a better position to profile, for instance, how an app treats your personal data...
at Frys, you tend to buy brands that you recognize. why? because the brand engenders a certain amount of trust. the difference with app stores is that we're treating the store as the brand, in part because app suppliers don't have much in the way of brand/trust identities of their own.
which is why Amazon opening an Android app store makes sense, and is not much of a problem for Google. Google may well be able to come up with some competitive advantage, but otherwise, they'll probably just get out of that business once Amazon (or others) have a decent ecosystem going. Sure, iTunes is a profit center for Apple, but that's largely a result of being a monoculture/walled-garden/etc.
South Part is all about pricking the pretentious, as outrageously as possible. Colbert does that too, but he's mainly about the politics, not the politicians. so yeah, Colbert isn't much fun if you don't actually care about politics. and why is it that you don't care about politics?
all the air in a machineroom is either hot or cold. anything else means you're mixing - that is, your containment leaks. there is basically no heat transfer through building conduction, for instance. 'cold' merely means that it's between the chiller outflow and front of servers; hot means ass-side of servers and chiller intake. the primary goal is to keep them from mixing.
a nontrivial machineroom will have multiple chillers and non-uniform heatload distribution. that doesn't change this principle, but does mean that the airflow design may have trouble getting enough air to the right spots. ideally, the chiller outflow would all go into a single large plenum (such as a deep pressurized underfloor), with hot ducting with controllable air intakes whisking the hot air back to the chillers.
look, you can already use whatever DNS server you want. if you're worried about your traffic being analyzed by someone else's DNS, just use your own (or a privacy-respecting) DNS elsewhere.
DNS is just the obvious way to ensure that clients use the best path to content.
I think you actually mean "need something better than the webcam's microphone". there's no problem capturing lectures with good webcams, since they provide plenty of sensitivity, resolution, refresh rate. you'll want to put a lapel mic on the speaker, that's all.
mbone+vic (+rat) pretty much describes AccessGrid. AG works, scales, but is not great and definitely not convenient the way a web client with java or flash would be.
Slashdot Mirror
shouldn't you expect everything you do in public to be potentially monitored? yes, the scale of modern life has, until recently, made most activity relatively anonymous, but only because no one bothered to look. I'm not sure why we should be worried about this.
HOWEVER, we should make sure that this be done above-the-board. for instance, the activity of police in public is clearly also something that should be public, and thus legally recordable. what police do when they enter your property is up to you to record if you wish. and government records resulting from this kind of recordkeeping of the public need to be public records (accessible to anyone). (government can reasonably charge for access if some company wants to mine these records, but I should also be able to, for a nominal fee, ask whether any records exist of, say, vehicles speeding on my street.)
so there are lots of quad-copters around that have roughly similar specs. this one is a uni-copter with 8 thrust-vectoring flaps, which is, I guess somewhat novel. not sure why 8 is the right number, and seems like a fairly large number, given that each requires a servo and fairly big piece of material. but since the flaps are independent, they can provide both direction and rotational control (which is why a quad-copter needs 4 fans - and why a helicopter needs a tail fan.) the spherical cage (and uni-fan) makes it seem compact and tidy, but I'm not sure the layout is actually better than a quad-copter.
a big part of the problem is that we've permitted doctors to form a medical mafia, and what we need is something closer to "healthcare engineering". as a guild, the medical establishment is structurally prone to rackets, and the relationship between big pharma and doctors qualifies.
but seriously, shouldn't the question be whether EM or audio has a more usable SNR in the random retail environment?
128 cores isn't enough to worry about - just install a distro you like and feel comfortable maintaining. although 128 cores isn't many, you should probably think about the style of install you want. lots of people seem to like diskful installs - afaikt purely because it's familiar. most significant clustering sites use diskless (NFS root) though, because it's so much easier to maintain. there's never any question of nodes getting out of sync. traffic due to NFS root is trivial. another best-practice is to configure 1-2 admin nodes (no users, provides NFS, scheduling and monitoring services), one or more dedicated login nodes, and discourage users from touching the compute nodes directly (among other things, give them non-routable addresses.) get or make a ticket system to keep track of user and system issues. monitor the heck out of your systems.
I'm an HPC center admin and system programmer, 10+ years. I think we've been in the top 50 several times.
no, we don't have any real info on how fast BD is yet. some technical papers on it indicate it's designed specifically for high clock while maintaining control on power dissipation. the shared FPU is somewhat faster than an unshared FPU would be, so this is a good choice, especially for code that's not always FPU-bound. and AMD has said that BD definitely has thermally-limited clock boosting.
tagging a picture is two separate acts: publishing the picture and publishing a statement (the tag). if I have published a legal picture of someone, I can tag it with anything non-libelous. not really anything new here, is there?
I'm sure these are very nice chips, but anyone can do similar, given funding. there are a number of cores available for licensing (like they did with MIPS), and adding vector units is the obvious way to boost your peak flops without blowing your power budget. I guess I don't really see why this merits all the coverage - for instance, what fraction of peak performance can it get on real code (say, a weather or MD simulation, not HPL)?
the quoted peak gflops/watt for this project are decent, but not much better than current commodity x86 parts, and comparable to GPUs. most architects in the field consider power-efficient computing to be a system-architecture challenge: how to move around all that data without spending all your power on fast/wide buses. a genuinely interesting new architecture would try to address this - perhaps something vaguely like IRAM. smart memory with some kind of high-order interconnect seems like the way to go, rather than putting giant vector units on a traditional design.
I would never interfere with someone else's fetish, but what is the point of getting excited about SSD speeds? how many people who buy them have any sort of thought-out reason to get them, rather than alternatives? what's your workload that works if you can write at 315 MB/s, but fails if you're limited to a measly 250?
in general, the SSD market seems driven by fetish, and that's just fine. the whole auto market is fetish-driven. and apple, too ;)
what I wonder, though, is if there's someone out there designing, say, a complex website with load balancing, HA-failover, frontend/backend/storage specialization, etc who just needs a 315 MB/s storage device to win. it seems like current standard practice (inmemory nosql, fast private networks, etc) make pretty much any performance concern just a matter of turning the knob - usually just adding some nodes. but if that's the case, then there's little importance to 315 vs 250 MB/s.
The disturbing part is that Beck isn't going away - the Tea Party and related craziness has its paranoid hooks sunk deeply enough into American self-doubt that a little turn-the-other-cheek will take away their oxygen. (sorry about the metaphor pileup there.)
Debian is the socks-and-sandals distro, and this change lets them cast off that embarrassing reliance on the conformist, ideologically impure Linux kernel. WOO! great leap forward.
but aren't we all expecting a major change in patentability, especially wrt this kind of software-related topics? the current territory-denial (landmine) approach to patents is pretty doomed...
this sort of discussion, especially given the mental limitations of websites, media outlets and media consumers, always gets bogged down in the morass of non-free vs defacto vs dejure vs license pools, etc.
the fact is that when we talk about "standard" in a good sense, it is open _and_ free, or not at all. for instance TCP/IP is a real standard precisely because it is both open and free. HTTP/HTML. javascript is, but java isn't. linux is and windows isn't. you get the idea.
in a meaningful sense, GNU caused part of the problem, because it conflates RMS's ideology with this basically simple "standard=open+free" model. (and before any RMSista object: GPL is popular as a flag of convenience - operating under GPL does not mean allegiance to RMS's politics...)
seriously, why is this news? yes, naifs out there associate OS stacks with particular hardware platforms, but the people _inside_ software companies don't. msft has produced extensive app stacks for ppc before, even to some extent mips and alpha. current windows is derived through NT and NT OS/2 from a codebase that was _originally_ developed simultaneously on MIPS and x86. obviously for devices like phones and tablets, even a lot of desktops, there's no need to worry about externally-provided add-in cards and the driver complexities that introduces.
besides, who cares that much about native apps anymore? "appliancy" stuff is browser/flash/java-based. msft itself is probably the main purveyor of non-browser/flash/java stuff, and of course they can retarget their office suite, no sweat (hah).
the interesting question in this is whether there's really any reason for ARM to be more mobile-friendly - that is, what is it about the ISA or implementation that makes it _inherently_ better (if any). my suspicion is that it's mostly a matter of methodology or culture: ARM has traditionally been very parsimonious (think hybrid or high-efficiency car), and the x86 makers have traditionally been more Nascar. Intel seemed to have done Atom almost against its will or corporate culture - followons have been more power-efficient, but they hardly seem like significant, bet-the-company efforts. AMD's recent bobcat-based chips appear to be based on a modestly-tweaked version of the K8. maybe what distinguishes ARM is something simple, like a compact instruction encoding...
you seem to be conflating cloud with desktop virtualization. the network needs of cloud use depend primarily on the amount of data creation or mutation going on, not on the bandwidth of the UI. it's also not clear that apple would actually win by shifting to dumber/thinner clients, as opposed to the hardware they sell now, which are basically high-margin reformatted PCs.
what an out-of-touch twit. then again, UChi is fairly famous for "odd" faculty.
numeracy is what's needed: that people are comfortable with quantitative reasoning. the specific mathematical techniques are irrelevant, but yes, people really do need better ability to understand issues in a quantitative way. climate change, sub-prime mortgages, this week's discount on cans of soup, fluoridating water, innoculations and autism, the list goes on forever. you can't be a competent human without understanding conditional probability, for instance.
the education system does an incredibly poor job of this, producing adults who struggle and fail to find structure and place in a big, confusing world, and for lack of comfort with analytic, quantitative approaches, latch on to religious/emotional/ideological movements like the Tea Party.
My experience is that when IO affects "feel", it's mostly because of increased competition for memory. Someone else mentioned tuning the vm.dirty_* sysctls, which certainly helps, but what surprises me is that we don't use O_DIRECT and splice/vmsplice more. "cp" is still a loop of 32k read/writes, with only the obvious O_ modes on either descriptor (and needless to say, no fadvise either.)
In short, the kernel offers fine mechanisms to resolve these problems - it's user-space that isn't taking advantage of them.
Amazon is already a major e-retailer, so they have some advantage there. I think the real question is whether they'll add value to the user/customer's app-selection process. yes, Amazon already has some you-may-also-like, and user reviews, but can Google do a better job of mining such data to produce value? Amazon doesn't seem to take this as seriously as Netflicks, for instance. can Google obtain some advantage from, for instance, crash reporting? perhaps they're in a better position to profile, for instance, how an app treats your personal data...
or android tablet, ebook reader...
at Frys, you tend to buy brands that you recognize. why? because the brand engenders a certain amount of trust. the difference with app stores is that we're treating the store as the brand, in part because app suppliers don't have much in the way of brand/trust identities of their own.
which is why Amazon opening an Android app store makes sense, and is not much of a problem for Google. Google may well be able to come up with some competitive advantage, but otherwise, they'll probably just get out of that business once Amazon (or others) have a decent ecosystem going. Sure, iTunes is a profit center for Apple, but that's largely a result of being a monoculture/walled-garden/etc.
South Part is all about pricking the pretentious, as outrageously as possible. Colbert does that too, but he's mainly about the politics, not the politicians. so yeah, Colbert isn't much fun if you don't actually care about politics. and why is it that you don't care about politics?
all the air in a machineroom is either hot or cold. anything else means you're mixing - that is, your containment leaks. there is basically no heat transfer through building conduction, for instance. 'cold' merely means that it's between the chiller outflow and front of servers; hot means ass-side of servers and chiller intake. the primary goal is to keep them from mixing.
a nontrivial machineroom will have multiple chillers and non-uniform heatload distribution. that doesn't change this principle, but does mean that the airflow design may have trouble getting enough air to the right spots. ideally, the chiller outflow would all go into a single large plenum (such as a deep pressurized underfloor), with hot ducting with controllable air intakes whisking the hot air back to the chillers.
look, you can already use whatever DNS server you want. if you're worried about your traffic being analyzed by someone else's DNS, just use your own (or a privacy-respecting) DNS elsewhere.
DNS is just the obvious way to ensure that clients use the best path to content.
I think you actually mean "need something better than the webcam's microphone". there's no problem capturing lectures with good webcams, since they provide plenty of sensitivity, resolution, refresh rate. you'll want to put a lapel mic on the speaker, that's all.
mbone+vic (+rat) pretty much describes AccessGrid. AG works, scales, but is not great and definitely not convenient the way a web client with java or flash would be.