I've fiddled with raid a bit but haven't yet had enough failures under my belt to know how raid systems behave in real-world drive failures. Can somebody comment on their experiences for the following hardware configs:
What happens when, say, an IDE drive fails using software raid on linux? Does the machine stay alive, or do you need to remove the drive and reboot to get back to working?
How about failures with full hardware raid, like an LSI megaraid card. I've unplugged a drive with one of these and it beeps a lot, and seems to recover when they get plugged back in, but what about a true drive failure?
I'm also really curious about these new SATA raid controllers that libata calls "fake" raid controllers. Can they gracefully handle a disk crapping out?
TIA to anyone who's seen enough failures to comment on this.
It's pretty obvious that the next wave of Moore's law seems to be moving computing towards parallelism.
This is pushing software developers to make their applications multi-threaded in order to exploit the performance gains of parallel processors. The interesting thing about this is that writing concurrent multi-threaded applications is extremely diffucult. I expect there to be an increase in demand on skilled programmers in the near future to overcome this diffuculty.
Look at it this way: the increase in CPU speed has spawned the rise of programming languages with builtin memory management, thus making programming easier to do. By using higher level languages like Visual Basic, Python, and even Java, programmers generally no longer have to worry about memory leaks. This has made software easier to develop, and has made the programming profession possible for more people.
AFAIK, there exists no anlog for making multi-threaded applications easier to write. They're damn hard, and tracking down race-conditions where one thread's actions screw up another thread mid-step is a royal pain.
It seems to me that this is going to impose a pretty big limitation on the capabilities of entry-level developers. Until somebody develops some sort of fire-and-forget race condition prevention tool, it's going to pay to have skills as a multi-threaded app developer.
I guess all of those oldschool Solaris guys who have been bragging about having true threading with Java and C/C++ since 1995 are going to finally get their day.
I heard this argument about ebay too, back when it's P/E was somewhere around 2000 (yes that's right, 2000). Take a look at ebay's stock graph and note that they never really went down. When everyone else was busting, ebay mostly stayed the same. And had you bought ebay at the height of the dotcom boom, and held it til today, you'd be quite happy.
Now, I'm not saying that google isn't overvalued right now, but if they can sustain their income growth, they still might just be a good buy. Considering I spend more time looking at google ads than just about any other media channel, it's a good bet that they can continue to grow.
FYI, the t10000 (Playstation 2 Development Tool) runs a version of redhat. Something ancient like RH4.2 if I'm not mistaken. You don't ssh into it or anything though, but it's web admin allows you to upload rpms to upgrade various subsystems on it.
Here I am reading an article about my former CEO when I stumble upon this line:
At the show, two of Robertson's engineers at MP3.com will introduce SwitchVox, which will combine PBX features with VOIP...
Oh crap that's me! Yeah, we have a fancy-pants gui front-end to asterisk. At the risk of further slashdotting ourselves, here's the site: http://www.switchvox.com.
Temperature sensors are good and all, but what would really rule would be a networked hydrometer or refractometer inside the fermentation tank giving you gravity readings. For non-homebrewers, the hydrometer reading shows the amount of dissolved sugars in your beer. This value decreases as the beer ferments (yeast eats sugar and turns it into alcohol), thus showing you when the beer is done fermenting. Normally it's a royal pain to measure because you have to extract small amounts of beer from the tank without contaminating the contents with airborne bacteria. However, with a hydrometer floating in the tank the whole time with some kind of sensors attached to it, you would know the instant your beer is ready to drink. Not only that, but correlating the slope of the hydrometer graph (fermentation velocity) with fermentation temperature would be a homebrewers wet dream. This is because certain yeast have ideal temperatures to ferment at. Too hot and the fermentation goes too quick, generating weird tastes and esters. Too slow and the yeast falls asleep. Armed with an rrdtool graph of temp and gravity though and your beer would always be juuuusst right.
My condo in downtown San Diego only has one parking spot which is permenantly claimed by my wife's car (it's nicer than my car). If I didn't have to commute via car every day I'd gladly subscribe to the flexcar that sits a block away. Heck, the subscription would pay just for the parking spot that car occupies. I can see this being a real benefit in other places where parking is brutal like beach areas or other urban centers.
From TFA: One approach is simply to make an individual system the unit of recovery; if anything fails, either restart the whole thing or fail-over to another system providing redundancy. Unfortunately, with the increasing physical resources available to each system, this approach is inherently wasteful: Why restart a whole system if you can disable a particular processor core, restart an individual application, or refrain from using a bit of your spacious memory or a particular I/O path until a repair is truly needed?
Because using stuff like stonith or heartbeat works for many more types of failures. Bad network cable? Yup. Power supply? yup. Server Catch on Fire? Yup.
I'm not saying it wouldn't be nice to have the OS route around bad memory blocks or bad processor cores due to some fancy-pants algorithm (without having to rewrite my app). But you're still going to need a redundant server for when somebody trips over the power cord.
Here's something I'd like to see in a big iron review: * Are the prices openly availiable * If not, can I get them via email, phone, fax? * How many phone calls to a sales guy does it take to get a price list. * You mean he wants to fly out to discuss pricing? * How much cheaper is my buddy at SavvyCorp able to buy it for since he knows the right guy to haggle with.
Side scrollers aren't completely dead. R-Type Final is an amazingly good (and faithful to the original) side-scroller for the Playstation 2. It combines the 3D engine with the side-scrolling genre to produce some amazing effects.
Also note in the gamespot review for Metroid Prime 2 that there are certain side-scrolling elements in it as well. The example they gave is a boss that you fight in ball-mode from a fixed, 2D perspective.
That strikes me as a cool hack. I don't see why games should pin themselves to fully 3D or fully 2D but freely mix it up both presentations in one game. I seem to remember an oldschool version of Sonic the Hedgehog that did a good job of this.
Note that these tests only uses the computing power of the main processor, while the GPU is sitting idly by doing nothing. With a little effort, and perhaps the use of some tools that harness the computing power of the gpu, these clusters would get a lot faster. It may not help in tasks like the distributed kernel compile, but things like parallel raytracing which can use the massively parallel floating point capabilities of the xbox graphics card could really benefit here.
In the future, the playstation 3 will really provide an opportunity for some enterprising cluster builders for couple of reasons. First, the initial release of most console hardware is where the manufacturer sells them for the biggest loss. Sony actually makes money on PS2s now even if you don't buy any games, but when they release the PS3, they'll be selling at a loss and your performance-to-cost ratio is going to be huge. Secondly, if the architecture decisions behind the PS3 make it anything like the PS2, it will be much easier to harness the vector engines for general purpose calculations (compared to other graphics cards). Most of the horsepower in the PS2 (and potentially in the PS3) is in it's parallel vector engines. While the general purpose processor is reasonably fast (300 mhz mips), the vector units can dispatch a ton of parallel floating point operations which enable it to run games that would crush a 300 mhz pentium with a comparable circa-2000 graphics card.
Yes and the FBI was trying to track down Mohammed Atta before the hijacking, but he disappeared somewhere in the US before they could find him. If his drivers license identified itself to some random government rfid readers (perhaps placed in every post office), then he would have been much easier to track down before he killed 2000 people.
I'm not saying I agree with this move, but this story is severely lacking the devil's advocate, so there it is.
I can't remember the last time I found a book that was more helpful than a handful of RFCs, man pages, mailing lists, and the source code to whatever I was trying to learn.
When I was first learning, I used to devour O'Reilly books like nobody's business. Lately it's just easier to use the resources at hand instead of struggling through a book that's too introductory.
Of course, there still are books that I dust off when I need them: Perl Cookbook, C++ ARM, Stephens' Network Programming. But it would seem a bit silly for slashdot to review these.
Actually, Jabber has found a very good niche doing behind the scenes work in lots of commercial software. For example, we were using it in my last job writing console video games. We're also looking to use it in a current voip product at my new company. The thing is, most of this work uses LGPL/BSD licensed libraries like iksemel so you'd never know that the underlying protocol driving that video game chat lobby is jabber unless you ran tcpdump on it.
It's probably cheaper for computer manufacturers to make (only) the latest and greatest and sell it to everybody than to try to specialize and sell one guy a 486 with DOS, somebody else a 4ghz p4, third guy gets a vt100 terminal, etc...
That's why new vt100 terminals retail for $250 while a new dell retails for $300. I'm sure the EE's on slashdot can testify about slapping a overpowered PIC microcontroller into a design instead of a cusom circuit because it simplified the design, and only bumped the product cost up from 30 cents to 40 cents.
It just makes sense from a manufacturing standpoint to mass produce one general-purpose product then try to shave a few pennies off making custom solutions for all kinds of tasks.
Hiring all of these expensive engineers must be costing them a fortune. They could seriously reduce their bottom line with a programming team or two in Bangalore.
since you mentioned the wrt54g you might be also interested in the Linksys NSLU2. It's got a single ethernet port, dual usb ports and can run linux. Attach a usb harddrive to it and you can use it for your file/mail server. The open source firmwares aren't as polished as, say sveasoft but it seems to me that you're the type that might enjoy getting it working.
Does anyone have evidence where a system was cracked due to the lack of entropy from things like interrupt timing?
I would think that there exists a limited number of people in the world who could exploit a diffie-helman exchange between systems using the usual sources of randomness on an x86 machine.
Just to add to this, I'd like to point out a reason why this is a good idea.
In taking a class, the instructor tells you directly how to do something. You may or may not retain the information long enough to reuse it the next time you have to, say, install qmail.
However, doing it yourself at home will teach you that all-imporant skill of how to google for linux howto information on the web.
I've done a couple of qmail installs in my lifetime, but any knowledge I've gained has long been forgotten. Except for the fact that I know that qmailrocks.org is the place I go to re-learn what to do.
When getting into philosophical discussions on spectrum economics, some people like to point at 2.4ghz as evidence that the FCC should step off and let anarchy rule.
However keep in mind that just because 2.4ghz is unlicensed, it is not unrestricted. If I went on the balcony of my downtown condo and put a 500dBm wifi AP with a 10dbi omni antenna, it would wreak havoc (and get me in trouble with the fcc).
For another example, imagine unlicensed wireless internet over AM radio spectrum. Yeah you could surf the web from 10 miles away from your house, but your signal would be destroyed from interference from everybody else's.
Now, I'm all for opening up as much spectrum as makes sense provided that the wavelenth is short enough to not blow through buildings etc, and provided the FCC restricts transmission strengths enough to not create anarchy.
The real bummer about this is that Nokia has far and away the most open platform for development of any phone manufacturer. They provide a huge array of sdk's and example code for both symbian and j2me developers.
Contrast this with an LG phone running brew on verizon and you have to pay all kinds of money and jump through all kinds of hoops just to write an app that verizon decides it doesn't want to distribute anyway.
My (very small) company is developing a cellphone app, and the costly barrier for starting Brew/Verizon devleopment is preventing us from using that platform. You pay through the nose for the development suite, then it's 300 bucks to register as a qcomm developer, then you have to jump through all of these verisign hoops to get a DRM key to sign your apps with, then you have to mail in your phone to be flashed into development mode, then you have to deal with verizon for distribution.
Meanwhile we're downloading compilers, tools, and example apps off the net for the nokia symbian platform that just work on an unmodded handset we bought at the store.
Global sales of recorded music - dominated by our country - quadrupled from 1980 to 1999. Then, almost on a dime, that trend line reversed, with sales figures falling by about a third to the mid point of last year. Before the launch of lawsuits by the industry last fall against those induced to steal music online, we were spiraling down with no sense of a floor.
So it seems that the financials of Ford Motor Company are similar. Their net income was in the red until last fall, and their stock price turned on a dime around 1999 and went spiraling down with no sense of a floor.
One can only conclude that pirates were illegaly copying cars produced by Ford Motor Company until the recent lawsuits of the RIAA stemmed the hemorrhaging. Either that or there was a global recession that affected all businesses.
Yes and if you relied on your linux machine to run a major database, we would accuse you of gross incompetence too.
There's ways of making systems redundant, the least of which are read-only replication or a solid backup (and recovery) process. If you don't have them, you're incompetent. And if that incompetence interferes with the freedoms of a democratic nation, so much the worse.
Slashdot Mirror
- What happens when, say, an IDE drive fails using software raid on linux? Does the machine stay alive, or do you need to remove the drive and reboot to get back to working?
- How about failures with full hardware raid, like an LSI megaraid card. I've unplugged a drive with one of these and it beeps a lot, and seems to recover when they get plugged back in, but what about a true drive failure?
- I'm also really curious about these new SATA raid controllers that libata calls "fake" raid controllers. Can they gracefully handle a disk crapping out?
TIA to anyone who's seen enough failures to comment on this.It's pretty obvious that the next wave of Moore's law seems to be moving computing towards parallelism.
This is pushing software developers to make their applications multi-threaded in order to exploit the performance gains of parallel processors.
The interesting thing about this is that writing concurrent multi-threaded applications is extremely diffucult. I expect there to be an increase in demand on skilled programmers in the near future to overcome this diffuculty.
Look at it this way: the increase in CPU speed has spawned the rise of programming languages with builtin memory management, thus making programming easier to do. By using higher level languages like Visual Basic, Python, and even Java, programmers generally no longer have to worry about memory leaks. This has made software easier to develop, and has made the programming profession possible for more people.
AFAIK, there exists no anlog for making multi-threaded applications easier to write. They're damn hard, and tracking down race-conditions where one thread's actions screw up another thread mid-step is a royal pain.
It seems to me that this is going to impose a pretty big limitation on the capabilities of entry-level developers. Until somebody develops some sort of fire-and-forget race condition prevention tool, it's going to pay to have skills as a multi-threaded app developer.
I guess all of those oldschool Solaris guys who have been bragging about having true threading with Java and C/C++ since 1995 are going to finally get their day.
For more reading, here's a good article about this stuff:
The Free Lunch Is Over: A Fundamental Turn Toward Concurrency in Software
I heard this argument about ebay too, back when it's P/E was somewhere around 2000 (yes that's right, 2000). Take a look at ebay's stock graph and note that they never really went down. When everyone else was busting, ebay mostly stayed the same. And had you bought ebay at the height of the dotcom boom, and held it til today, you'd be quite happy.
Now, I'm not saying that google isn't overvalued right now, but if they can sustain their income growth, they still might just be a good buy. Considering I spend more time looking at google ads than just about any other media channel, it's a good bet that they can continue to grow.
FYI, the t10000 (Playstation 2 Development Tool) runs a version of redhat. Something ancient like RH4.2 if I'm not mistaken. You don't ssh into it or anything though, but it's web admin allows you to upload rpms to upgrade various subsystems on it.
Here I am reading an article about my former CEO when I stumble upon this line:
At the show, two of Robertson's engineers at MP3.com will introduce SwitchVox, which will
combine PBX features with VOIP...
Oh crap that's me! Yeah, we have a fancy-pants gui front-end to asterisk. At the risk of further slashdotting ourselves, here's the site: http://www.switchvox.com.
Now to go find some bandwidth.
Temperature sensors are good and all, but what would really rule would be a networked hydrometer or refractometer inside the fermentation tank giving you gravity readings. For non-homebrewers, the hydrometer reading shows the amount of dissolved sugars in your beer. This value decreases as the beer ferments (yeast eats sugar and turns it into alcohol), thus showing you when the beer is done fermenting. Normally it's a royal pain to measure because you have to extract small amounts of beer from the tank without contaminating the contents with airborne bacteria. However, with a hydrometer floating in the tank the whole time with some kind of sensors attached to it, you would know the instant your beer is ready to drink. Not only that, but correlating the slope of the hydrometer graph (fermentation velocity) with fermentation temperature would be a homebrewers wet dream. This is because certain yeast have ideal temperatures to ferment at. Too hot and the fermentation goes too quick, generating weird tastes and esters. Too slow and the yeast falls asleep. Armed with an rrdtool graph of temp and gravity though and your beer would always be juuuusst right.
My condo in downtown San Diego only has one parking spot which is permenantly claimed by my wife's car (it's nicer than my car). If I didn't have to commute via car every day I'd gladly subscribe to the flexcar that sits a block away. Heck, the subscription would pay just for the parking spot that car occupies. I can see this being a real benefit in other places where parking is brutal like beach areas or other urban centers.
From TFA: One approach is simply to make an individual system the unit of recovery; if anything fails, either restart the whole thing or fail-over to another system providing redundancy. Unfortunately, with the increasing physical resources available to each system, this approach is inherently wasteful: Why restart a whole system if you can disable a particular processor core, restart an individual application, or refrain from using a bit of your spacious memory or a particular I/O path until a repair is truly needed?
Because using stuff like stonith or heartbeat works for many more types of failures. Bad network cable? Yup. Power supply? yup. Server Catch on Fire? Yup.
I'm not saying it wouldn't be nice to have the OS route around bad memory blocks or bad processor cores due to some fancy-pants algorithm (without having to rewrite my app). But you're still going to need a redundant server for when somebody trips over the power cord.
Here's something I'd like to see in a big iron review:
* Are the prices openly availiable
* If not, can I get them via email, phone, fax?
* How many phone calls to a sales guy does it take to get a price list.
* You mean he wants to fly out to discuss pricing?
* How much cheaper is my buddy at SavvyCorp able to buy it for since he knows the right guy to haggle with.
Side scrollers aren't completely dead.
R-Type Final is an amazingly good (and faithful to the original) side-scroller for the Playstation 2. It combines the 3D engine with the side-scrolling genre to produce some amazing effects.
Also note in the gamespot review for Metroid Prime 2 that there are certain side-scrolling elements in it as well. The example they gave is a boss that you fight in ball-mode from a fixed, 2D perspective.
That strikes me as a cool hack. I don't see why games should pin themselves to fully 3D or fully 2D but freely mix it up both presentations in one game. I seem to remember an oldschool version of Sonic the Hedgehog that did a good job of this.
Note that these tests only uses the computing power of the main processor, while the GPU is sitting idly by doing nothing. With a little effort, and perhaps the use of some tools that harness the computing power of the gpu, these clusters would get a lot faster. It may not help in tasks like the distributed kernel compile, but things like parallel raytracing which can use the massively parallel floating point capabilities of the xbox graphics card could really benefit here.
In the future, the playstation 3 will really provide an opportunity for some enterprising cluster builders for couple of reasons. First, the initial release of most console hardware is where the manufacturer sells them for the biggest loss. Sony actually makes money on PS2s now even if you don't buy any games, but when they release the PS3, they'll be selling at a loss and your performance-to-cost ratio is going to be huge. Secondly, if the architecture decisions behind the PS3 make it anything like the PS2, it will be much easier to harness the vector engines for general purpose calculations (compared to other graphics cards). Most of the horsepower in the PS2 (and potentially in the PS3) is in it's parallel vector engines. While the general purpose processor is reasonably fast (300 mhz mips), the vector units can dispatch a ton of parallel floating point operations which enable it to run games that would crush a 300 mhz pentium with a comparable circa-2000 graphics card.
Yes and the FBI was trying to track down Mohammed Atta before the hijacking, but he disappeared somewhere in the US before they could find him. If his drivers license identified itself to some random government rfid readers (perhaps placed in every post office), then he would have been much easier to track down before he killed 2000 people.
I'm not saying I agree with this move, but this story is severely lacking the devil's advocate, so there it is.
I can't remember the last time I found a book that was more helpful than a handful of RFCs, man pages, mailing lists, and the source code to whatever I was trying to learn.
When I was first learning, I used to devour O'Reilly books like nobody's business. Lately it's just easier to use the resources at hand instead of struggling through a book that's too introductory.
Of course, there still are books that I dust off when I need them: Perl Cookbook, C++ ARM, Stephens' Network Programming. But it would seem a bit silly for slashdot to review these.
Actually, Jabber has found a very good niche doing behind the scenes work in lots of commercial software. For example, we were using it in my last job writing console video games. We're also looking to use it in a current voip product at my new company. The thing is, most of this work uses LGPL/BSD licensed libraries like iksemel so you'd never know that the underlying protocol driving that video game chat lobby is jabber unless you ran tcpdump on it.
It's probably cheaper for computer manufacturers to make (only) the latest and greatest and sell it to everybody than to try to specialize and sell one guy a 486 with DOS, somebody else a 4ghz p4, third guy gets a vt100 terminal, etc...
That's why new vt100 terminals retail for $250 while a new dell retails for $300. I'm sure the EE's on slashdot can testify about slapping a overpowered PIC microcontroller into a design instead of a cusom circuit because it simplified the design, and only bumped the product cost up from 30 cents to 40 cents.
It just makes sense from a manufacturing standpoint to mass produce one general-purpose product then try to shave a few pennies off making custom solutions for all kinds of tasks.
Hiring all of these expensive engineers must be costing them a fortune. They could seriously reduce their bottom line with a programming team or two in Bangalore.
since you mentioned the wrt54g you might be also interested in the Linksys NSLU2. It's got a single ethernet port, dual usb ports and can run linux. Attach a usb harddrive to it and you can use it for your file/mail server. The open source firmwares aren't as polished as, say sveasoft but it seems to me that you're the type that might enjoy getting it working.
Does anyone have evidence where a system was cracked due to the lack of entropy from things like interrupt timing?
I would think that there exists a limited number of people in the world who could exploit a diffie-helman exchange between systems using the usual sources of randomness on an x86 machine.
Just to add to this, I'd like to point out a reason why this is a good idea.
In taking a class, the instructor tells you directly how to do something. You may or may not retain the information long enough to reuse it the next time you have to, say, install qmail.
However, doing it yourself at home will teach you that all-imporant skill of how to google for linux howto information on the web.
I've done a couple of qmail installs in my lifetime, but any knowledge I've gained has long been forgotten. Except for the fact that I know that qmailrocks.org is the place I go to re-learn what to do.
You can use apt for RedHat, Fedora, and Mandrake distros too. If that's your only reason for using debian, then you might consider a compromise.
See atrpms for more info.
When getting into philosophical discussions on spectrum economics, some people like to point at 2.4ghz as evidence that the FCC should step off and let anarchy rule.
However keep in mind that just because 2.4ghz is unlicensed, it is not unrestricted. If I went on the balcony of my downtown condo and put a 500dBm wifi AP with a 10dbi omni antenna, it would wreak havoc (and get me in trouble with the fcc).
For another example, imagine unlicensed wireless internet over AM radio spectrum. Yeah you could surf the web from 10 miles away from your house, but your signal would be destroyed from interference from everybody else's.
Now, I'm all for opening up as much spectrum as makes sense provided that the wavelenth is short enough to not blow through buildings etc, and provided the FCC restricts transmission strengths enough to not create anarchy.
The real bummer about this is that Nokia has far and away the most open platform for development of any phone manufacturer. They provide a huge array of sdk's and example code for both symbian and j2me developers.
Contrast this with an LG phone running brew on verizon and you have to pay all kinds of money and jump through all kinds of hoops just to write an app that verizon decides it doesn't want to distribute anyway.
My (very small) company is developing a cellphone app, and the costly barrier for starting Brew/Verizon devleopment is preventing us from using that platform. You pay through the nose for the development suite, then it's 300 bucks to register as a qcomm developer, then you have to jump through all of these verisign hoops to get a DRM key to sign your apps with, then you have to mail in your phone to be flashed into development mode, then you have to deal with verizon for distribution.
Meanwhile we're downloading compilers, tools, and example apps off the net for the nokia symbian platform that just work on an unmodded handset we bought at the store.
Global sales of recorded music - dominated by our country - quadrupled from 1980 to 1999. Then, almost on a dime, that trend line reversed, with sales figures falling by about a third to the mid point of last year. Before the launch of lawsuits by the industry last fall against those induced to steal music online, we were spiraling down with no sense of a floor.
So it seems that the financials of Ford Motor Company are similar. Their net income was in the red until last fall, and their stock price turned on a dime around 1999 and went spiraling down with no sense of a floor.
One can only conclude that pirates were illegaly copying cars produced by Ford Motor Company until the recent lawsuits of the RIAA stemmed the hemorrhaging. Either that or there was a global recession that affected all businesses.
Interesting that in this article, Nigeria has that highest cost/income ratio in the world.
I bet I can guess the reason
Yes and if you relied on your linux machine to run a major database, we would accuse you of gross incompetence too.
There's ways of making systems redundant, the least of which are read-only replication or a solid backup (and recovery) process. If you don't have them, you're incompetent. And if that incompetence interferes with the freedoms of a democratic nation, so much the worse.