"CD's same thing. They used to always work [..] Now with the increasing HD density we can no longer rely on HD either!"
Um, I think the problem here is that, rather than hardware getting less reliable, you've actually dimension shifted from some freaky alternative reality where storage is far more reliable.
"I rather go back the smaller 80 GIG drives then continue down this wobbly path."
What, the ones with largely the same measured failure rates as disks 10x larger and 10x smaller? Was there a time when manufacturers claimed lower AFR's and higher MTBF's, or are you just suffering from confirmation bias?
If you really want more reliable disks and are prepared to sacrifice storage capacity (and money) for it, go SCSI/SAS. Don't forget to whinge when they turn out to be less than perfect too.
No, that's the disk throwing a tantrum because it can't find the data it wants.
You can demonstrate this yourself; open up a running hard disk and remove the platter - in pretty much all cases a rather physically violent ending will occur. That's because the disk is *upset*; you took away its data!
It's hoped that, once we have disks who's lifetimes can be measured in decades instead of a handful of years, the devices will be mature enough to take such failures in their stride.
I have plenty of my own hosting (several racks, a few dozen machines), but I still run a webserver at home.
A large part of it is that it's easy, and the machine and software it lives on would exist anyway (I'm a web developer, among other things). I find it a good place to dump low priority things which I don't want to faff about dealing with remotely; like, photos. I could easily be using GB's of disk space with photos hardly anyone will bother to look at, and that can add up rapidly on a server which may only have a few 36G HD's shared between various users. In the mean time, the users most likely to be interested (me, others in the house) would actually end up with slower access, while everyone suffers from the increased time it takes to add new images.
There other stuff too; prototype applications, or things with dependencies which don't happen to match those on a handy server. I can use mod_proxy to give them nice URLs and even provide caching from a "proper" server, while still serving them from home where I can keep an eye on them and develop them safely and comfortably.
Er? Biological systems are tricker to repair and restart after catastrophic failure than a car engine, ergo consciousness has to have some additional component?
It's fairly clear that consciousness is a process, since it's part of cognition, which involves the manipulation of information.. over time. That doesn't mean it's not something you can't start and stop at will; about all you might be able to vaguely infer is that biological brains perhaps don't bother storing everything in non-volatile forms which can be restarted easily.
Personally I feel it has something to do with the continuity of brain activity. You interrupt that, and whatever that "spark" is ceases to be, and if the brain is turned back on, it would be a different "you". Smells like latent dualism to me. Either you think you're embodied in the information stored and processed by your nervous system, or you think there's some mysterious extra, which provides "youness".
And true, dualism kind of feels right; it can't just be some dumb electrochemical process going on inside our heads truely "experiencing" being us, there's got to be some extra spark which seperates us from that, because, damnit, I'm here! Experiencing stuff!
I try not to do my thinking with my feelings, though; especially not in cases they're unlikely to be optimized for. And copies of mind-states and nature of consciousness? Yeah..
A single G80 GPU contains about 690 million transistors; enough for about 3 dual core Opterons. Then you've got 768MB of highly clocked GDDR3 memory to go with it; it's not really surprising if it takes about as much power as an entire computer, especially when it's pushed that hard -- similarly high end CPU's can easily eat 125W just on their own.
Greg Egan - best ideas ever. I'm currently re-re-reading Diaspora Charles Stross - fun. I read Accelerando (free book!), then bought all his other stuff and wasn't disappointed. Richard Morgan - really likes his Lone Genetically Modified Male protagonists, but luckily he does them well enough for it not to get old. Alastair Reynolds - the Revelation Space universe is one of my favourites. Iain (M.) Banks - The Culture novels are quite interesting, and his other books aren't bad either.
Honourable mentions:
Peter Watts - all his books appear to be online. Blindsight is very, very good, but I've not read much else from him yet. Greg Bear - some of his older works are among my favourites. Queen of Angels, Slant (literally "/") and Moving Mars are one of my favourite trilogies. I'm behind on his newer stuff though, and his latest "terrorist thriller" makes me suspicious.
"Yes, PostgreSQL has MVCC, while InnoDB has to make do with row-level locking." Er, no, InnoDB uses MVCC, as can be trivially verified by playing about with it. Start a transaction in 2 clients, update a table in client 1 and commit, then read those rows in client 2; note client 2 sees the version of the table prior to client 1's modifications because you've got a versioned snapshot.
You can modify this behavior by selecting a different isolation level.
mod_fastcgi runs fine in Apache 2, threaded mode and all, and has done so for many years. mod_fcgi is an alternative with, supposedly, a smarter application manager (that's the thing which spawns additional FastCGI servers as needed), and it too should work fine in Apache 2, though we've only used mod_fastcgi in production.
This is our preferred way of running PHP, too. Keeps the webserver nice and lean and isolated and stable. On the other hand, I think it would be tidier if PHP had something like mongrel instead; a minimal webserver for serving PHP applications, so we can just set up mod_proxy(_balancer) instead of faffing about with what is, ultimately, a somewhat crufty solution to a problem HTTP already solves. Anyone fancy porting the HTTP bits of mongrel to a PHP SAPI?
"latency doesn't seem to matter at all" It might not matter much on your desktop, I can assure you it matters quite a lot on plenty of servers. People don't buy things like this for nothing. It's less important when all your performance-critical data fits in memory, but that's not always feasable.
Also, Raptors are 10kRPM disks, and still trade off density for latency at a level most high performance drives do not.
"Perhaps we'll have hard drives with four platters internally RAID1-ed" This crops up from time to time, and I'm yet to quite grasp the thinking behind it. How exactly do you manage the "Independent" part when you put it all inside a single drive? Oh, I know.. make your "drive" an enclosure for two or more smaller ones. Hmmm...
"I have always wondered why drives couldn't be configured with two independent arm assemblies" They can, it's just not worth it; it's a lot of additional expense and complexity (and thus reduced MTBF) all for a very low volume part, when most people would prefer you to just make a physically smaller, cheaper disk so they can get more of them when needed.
Read-write on all platters at once isn't really feasable because the tracks aren't going to line up reliably; leaving aside imperfect manufacturing, components aren't all going to see uniform levels of thermal expansion or vibration, and even microscopic differences in where each head settles will leave you screwed -- lining up with one track will, most likely, be mutually exclusive to lining up with a second, and get worse from there.
"I have not yet come across a single disk that could do 100MB/s" Seagate Savvio 2.5" 15kRPM SAS disks can manage roughly 110MB/s in good conditions.
Seagate's next consumer drives, 7200.11, supposedly manage similar serial transfer rates with their 250GB platters, though will of course have massively higher seek times.
That's an.. interesting line of rhetoric, what do you base it on? Do you think the poor are poor because they're lazy or something? What the is your "American dream" anyway?
Of course, thousands of people, including a significant majority of FreeBSD's developers, use AMD64 and don't see problems like this, so I don't think it's so much "AMD64 hates FreeBSD 6.2" as "I've run into an annoying bug and would like to overgeneralize in a vague attempt to solicit sympathy and interest".
Given that you're faced with having to move to Linux, I can sympathise, but playing the "FreeBSD/AMD64 is just too unstable to use!!1!!!!1!!!2" card is just counterproductive FUD, especially when you don't even bother linking to a PR or relevent thread.
"sleeping on a non-sleepable lock" is very non-specific, fwiw. Hardly better than "it doesn't work", since it doesn't say anything about where it happens. A search on just that is going to uncover a tonne of unrelated issues which just happen to result in the same panic. They're not likely to be related to AMD processor bugs either; more likely it's a driver problem.
If I was a juror I would vote "not guilty" on this evidence. I'm a big believer in "proven beyond all reasonable doubt." Quite, but I have serious problems trusting a selection of my "peers" to be quite so impartial and clear thinking. Especially when a massive proportion of them repeatedly demonstrate their poor reasoning skills and/or ethics with beliefs like the creator of the universe has a personal relationship with them, crystals have healing energy, homosexuals are evil, atheists are worse, Bush is awesome, American Idol is pretty good, and the Iraq war is about the 9/11 terrorists.
What's the criteria for deciding whether someone's mentally competent again?
3) The on-board reality distortion field generator resolves the problem. At the expense of making nearby computers lacking such devices somewhat *less* reliable, of course.
I swear, our servers have been behaving more strangely since the guy in the next rack over installed a few Mac Mini's; maybe the RDFG's emmit gamma rays from their embedded naked singularity...
Funny, I hate physical media. I hate having huge chunks of space taken up with things I barely touch, I hate having to rip and tag every CD I buy, especially when half of them have systems which try to make that difficult. I hate the wait to actually get hold of the media, too.
A collection of properly tagged.flac's and a couple of.jpg's have far more utility to me than a physical CD I'm going to have to process myself (even if that only involved me inserting it and waiting 10 minutes).
Of course, the majority of those in the music industry doesn't seem to want me to have either; they would much prefer I get some crappy encumbered lossy files or some optical media I can't rip properly. More fool them.
"That's a bit unfair to those who need religion" Nobody *needs* religion. People do need effective social support, and religion can prove to be a useful nucleating agent to help provide that; indeed this is probably a large factor in why it's so common. This doesn't mean it's remotely the best way to provide it, especially when there are so many unfortunate side-effects.
"when you trash on religion in general that's a step to far" Nope. Sorry, religion does not deserve a "get out of criticism free" card, or a "get unconditional respect free" card. You have a right to your beliefs, you *absolutely* do not have a right to not be called out on them, especially when you're hanging around primary schools indoctrinating 6 year olds with your damaging nonsense*.
* The UK has publically funded "faith schools"; the majority of the public schools in my area are R.C. (Roman Catholic), including the two primary schools I went to; churches "just happened" to be built right next to them, and during my time in them it was a regular thing to be taken to them by the school for confession and services, in addition to being taught the basics of the faith, culminating in being "confirmed". Anyone know if this is still common?
A more specific reference: "a weak magnetic field--1 microtesla, which is roughly that generated by a computer monitor--rotating anticlockwise in a complex pattern about the temporal lobes will cause four out of five people to feel a spectral presence in the room with them."
Maxtor are Seagate's cheap-and-cheerful brand (as if Seagate's own weren't cheap enough). If you really care about long term reliability with a lot of disks, you probably want to spend a bit more and get a NL/NS/ES (Near-Line Storage/Enterprise Storage) drive, which really don't cost a whole lot more. Well, maybe they will if you're buying tiny drives.
Either way, why are you going directly to Seagate for your "satisfaction"? Return them to your vendor and let them deal with the manufacturer if it's so soon after you bought them?
Also, someone probably dropped the box your drives came in at some point, if you weren't just unlucky with a bad batch. Try not buying from the cheapest and nastiest vendor possible.
And then find the chip doesn't really want to run at 2.6GHz, and that it really deserved to be in the 1.8 bin. You will, of course, not immediately put the clock up to 2.6, you'll creep the clockrate up and put each stage through hours if not days of exhaustive testing exercising all the components of the chip (and not just, say, the ALUs). Don't forget to do it on a very hot day, and without your air conditioning on full blast.
Not treating overclocking with the respect it deserves is a path to mysterious crashes you can't track down, creeping system corruption which doesn't show up obviously for months, crashing games, corrupt media encodes, archives which spew CRC errors, broken backups and recovery files. It's not just a case of "if I push it too far it'll crash obviously and very quickly". I dunno about you, but that additional couple of hundred dollars is *nothing* compared with the effort involved in overclocking a lower end part and testing it properly.
Ultimately, yes, more performance is nice, and sometimes you can get really impressive boosts, but large ones like this are the exception rather than the rule, and can be surprisingly difficult to pull off properly, especially as time goes on and the manufacturer gets more savvy with their production and binning.
Also, it's spelled "heart". And I think you mean "*en*coding".
Yes, of course it would be broken either way, that would just be the first thing I'd look at if I were seeing such behavior. Sorry if my vague attempt to be slightly helpful upset you so:P
Personally my most pressing performance issue with Opera is the rather clunky RSS/Atom support. 5 second delays prior to "oi, your feeds have been updated" is somewhat distracting. Well, that and the somewhat extended load times with 40+ tabs...
I count 10 on my MX518's; Left, Right, Middle, Scrollwheel Up, Scrollwheel Down, Back, Forward, +, - (think these are hardcoded to setting the mouse resolution though) and.. um, another one which going by the glyph is meant for something to do with window stacking.
Slashdot Mirror
"CD's same thing. They used to always work [..] Now with the increasing HD density we can no longer rely on HD either!"
Um, I think the problem here is that, rather than hardware getting less reliable, you've actually dimension shifted from some freaky alternative reality where storage is far more reliable.
"I rather go back the smaller 80 GIG drives then continue down this wobbly path."
What, the ones with largely the same measured failure rates as disks 10x larger and 10x smaller? Was there a time when manufacturers claimed lower AFR's and higher MTBF's, or are you just suffering from confirmation bias?
If you really want more reliable disks and are prepared to sacrifice storage capacity (and money) for it, go SCSI/SAS. Don't forget to whinge when they turn out to be less than perfect too.
No, that's the disk throwing a tantrum because it can't find the data it wants.
You can demonstrate this yourself; open up a running hard disk and remove the platter - in pretty much all cases a rather physically violent ending will occur. That's because the disk is *upset*; you took away its data!
It's hoped that, once we have disks who's lifetimes can be measured in decades instead of a handful of years, the devices will be mature enough to take such failures in their stride.
I have plenty of my own hosting (several racks, a few dozen machines), but I still run a webserver at home.
A large part of it is that it's easy, and the machine and software it lives on would exist anyway (I'm a web developer, among other things). I find it a good place to dump low priority things which I don't want to faff about dealing with remotely; like, photos. I could easily be using GB's of disk space with photos hardly anyone will bother to look at, and that can add up rapidly on a server which may only have a few 36G HD's shared between various users. In the mean time, the users most likely to be interested (me, others in the house) would actually end up with slower access, while everyone suffers from the increased time it takes to add new images.
There other stuff too; prototype applications, or things with dependencies which don't happen to match those on a handy server. I can use mod_proxy to give them nice URLs and even provide caching from a "proper" server, while still serving them from home where I can keep an eye on them and develop them safely and comfortably.
Er? Biological systems are tricker to repair and restart after catastrophic failure than a car engine, ergo consciousness has to have some additional component?
It's fairly clear that consciousness is a process, since it's part of cognition, which involves the manipulation of information.. over time. That doesn't mean it's not something you can't start and stop at will; about all you might be able to vaguely infer is that biological brains perhaps don't bother storing everything in non-volatile forms which can be restarted easily.
And true, dualism kind of feels right; it can't just be some dumb electrochemical process going on inside our heads truely "experiencing" being us, there's got to be some extra spark which seperates us from that, because, damnit, I'm here! Experiencing stuff!
I try not to do my thinking with my feelings, though; especially not in cases they're unlikely to be optimized for. And copies of mind-states and nature of consciousness? Yeah..
A single G80 GPU contains about 690 million transistors; enough for about 3 dual core Opterons. Then you've got 768MB of highly clocked GDDR3 memory to go with it; it's not really surprising if it takes about as much power as an entire computer, especially when it's pushed that hard -- similarly high end CPU's can easily eat 125W just on their own.
Greg Egan - best ideas ever. I'm currently re-re-reading Diaspora
Charles Stross - fun. I read Accelerando (free book!), then bought all his other stuff and wasn't disappointed.
Richard Morgan - really likes his Lone Genetically Modified Male protagonists, but luckily he does them well enough for it not to get old.
Alastair Reynolds - the Revelation Space universe is one of my favourites.
Iain (M.) Banks - The Culture novels are quite interesting, and his other books aren't bad either.
Honourable mentions:
Peter Watts - all his books appear to be online. Blindsight is very, very good, but I've not read much else from him yet.
Greg Bear - some of his older works are among my favourites. Queen of Angels, Slant (literally "/") and Moving Mars are one of my favourite trilogies. I'm behind on his newer stuff though, and his latest "terrorist thriller" makes me suspicious.
The list was empty here. What was on it?
You can modify this behavior by selecting a different isolation level.
mod_fastcgi runs fine in Apache 2, threaded mode and all, and has done so for many years. mod_fcgi is an alternative with, supposedly, a smarter application manager (that's the thing which spawns additional FastCGI servers as needed), and it too should work fine in Apache 2, though we've only used mod_fastcgi in production.
This is our preferred way of running PHP, too. Keeps the webserver nice and lean and isolated and stable. On the other hand, I think it would be tidier if PHP had something like mongrel instead; a minimal webserver for serving PHP applications, so we can just set up mod_proxy(_balancer) instead of faffing about with what is, ultimately, a somewhat crufty solution to a problem HTTP already solves. Anyone fancy porting the HTTP bits of mongrel to a PHP SAPI?
Also, Raptors are 10kRPM disks, and still trade off density for latency at a level most high performance drives do not. "Perhaps we'll have hard drives with four platters internally RAID1-ed" This crops up from time to time, and I'm yet to quite grasp the thinking behind it. How exactly do you manage the "Independent" part when you put it all inside a single drive? Oh, I know.. make your "drive" an enclosure for two or more smaller ones. Hmmm...
Read-write on all platters at once isn't really feasable because the tracks aren't going to line up reliably; leaving aside imperfect manufacturing, components aren't all going to see uniform levels of thermal expansion or vibration, and even microscopic differences in where each head settles will leave you screwed -- lining up with one track will, most likely, be mutually exclusive to lining up with a second, and get worse from there.
Of course IANAHDM.
Seagate's next consumer drives, 7200.11, supposedly manage similar serial transfer rates with their 250GB platters, though will of course have massively higher seek times.
That's an.. interesting line of rhetoric, what do you base it on? Do you think the poor are poor because they're lazy or something? What the is your "American dream" anyway?
"Core 2 kicks Barcelona's ass"
Do you know something the rest of us don't?
Of course, thousands of people, including a significant majority of FreeBSD's developers, use AMD64 and don't see problems like this, so I don't think it's so much "AMD64 hates FreeBSD 6.2" as "I've run into an annoying bug and would like to overgeneralize in a vague attempt to solicit sympathy and interest".
Given that you're faced with having to move to Linux, I can sympathise, but playing the "FreeBSD/AMD64 is just too unstable to use!!1!!!!1!!!2" card is just counterproductive FUD, especially when you don't even bother linking to a PR or relevent thread.
"sleeping on a non-sleepable lock" is very non-specific, fwiw. Hardly better than "it doesn't work", since it doesn't say anything about where it happens. A search on just that is going to uncover a tonne of unrelated issues which just happen to result in the same panic. They're not likely to be related to AMD processor bugs either; more likely it's a driver problem.
What's the criteria for deciding whether someone's mentally competent again?
3) The on-board reality distortion field generator resolves the problem. At the expense of making nearby computers lacking such devices somewhat *less* reliable, of course.
I swear, our servers have been behaving more strangely since the guy in the next rack over installed a few Mac Mini's; maybe the RDFG's emmit gamma rays from their embedded naked singularity...
Funny, I hate physical media. I hate having huge chunks of space taken up with things I barely touch, I hate having to rip and tag every CD I buy, especially when half of them have systems which try to make that difficult. I hate the wait to actually get hold of the media, too.
.flac's and a couple of .jpg's have far more utility to me than a physical CD I'm going to have to process myself (even if that only involved me inserting it and waiting 10 minutes).
A collection of properly tagged
Of course, the majority of those in the music industry doesn't seem to want me to have either; they would much prefer I get some crappy encumbered lossy files or some optical media I can't rip properly. More fool them.
* The UK has publically funded "faith schools"; the majority of the public schools in my area are R.C. (Roman Catholic), including the two primary schools I went to; churches "just happened" to be built right next to them, and during my time in them it was a regular thing to be taken to them by the school for confession and services, in addition to being taught the basics of the faith, culminating in being "confirmed". Anyone know if this is still common?
See transcranial magnetic stimulation: "magnetic field (B): often about 2 tesla on the coil surface and 0.5 T in the cortex".
A more specific reference: "a weak magnetic field--1 microtesla, which is roughly that generated by a computer monitor--rotating anticlockwise in a complex pattern about the temporal lobes will cause four out of five people to feel a spectral presence in the room with them."
Maxtor are Seagate's cheap-and-cheerful brand (as if Seagate's own weren't cheap enough). If you really care about long term reliability with a lot of disks, you probably want to spend a bit more and get a NL/NS/ES (Near-Line Storage/Enterprise Storage) drive, which really don't cost a whole lot more. Well, maybe they will if you're buying tiny drives.
Either way, why are you going directly to Seagate for your "satisfaction"? Return them to your vendor and let them deal with the manufacturer if it's so soon after you bought them?
Also, someone probably dropped the box your drives came in at some point, if you weren't just unlucky with a bad batch. Try not buying from the cheapest and nastiest vendor possible.
And then find the chip doesn't really want to run at 2.6GHz, and that it really deserved to be in the 1.8 bin. You will, of course, not immediately put the clock up to 2.6, you'll creep the clockrate up and put each stage through hours if not days of exhaustive testing exercising all the components of the chip (and not just, say, the ALUs). Don't forget to do it on a very hot day, and without your air conditioning on full blast.
Not treating overclocking with the respect it deserves is a path to mysterious crashes you can't track down, creeping system corruption which doesn't show up obviously for months, crashing games, corrupt media encodes, archives which spew CRC errors, broken backups and recovery files. It's not just a case of "if I push it too far it'll crash obviously and very quickly". I dunno about you, but that additional couple of hundred dollars is *nothing* compared with the effort involved in overclocking a lower end part and testing it properly.
Ultimately, yes, more performance is nice, and sometimes you can get really impressive boosts, but large ones like this are the exception rather than the rule, and can be surprisingly difficult to pull off properly, especially as time goes on and the manufacturer gets more savvy with their production and binning.
Also, it's spelled "heart". And I think you mean "*en*coding".
Yes, of course it would be broken either way, that would just be the first thing I'd look at if I were seeing such behavior. Sorry if my vague attempt to be slightly helpful upset you so :P
Personally my most pressing performance issue with Opera is the rather clunky RSS/Atom support. 5 second delays prior to "oi, your feeds have been updated" is somewhat distracting. Well, that and the somewhat extended load times with 40+ tabs...
I count 10 on my MX518's; Left, Right, Middle, Scrollwheel Up, Scrollwheel Down, Back, Forward, +, - (think these are hardcoded to setting the mouse resolution though) and.. um, another one which going by the glyph is meant for something to do with window stacking.