The argument against PCIe is that if anything on that particular server fries, all of your data is now trapped inside a dead server; there's no way to fail over quickly.
If you care about high availability you probably want to look into a SAN flash appliance, i.e. Pure Storage. (shameless plug)
This is generally true; shared _disk_ will suck once you have multiple VMs accessing it because they'll each try to optimize for seeks internally, but the combined I/O stream will be relatively seek-heavy and wreck performance. Where I work we call this the "I/O blender."
Solid state storage can solve this, but then you have to balance cost, high-availability (if your data is on a direct-attached PCIe card, what do you do when that motherboard dies?) and performance.
This is the problem that a new generation of storage companies are trying to solve. Shameless plug: Pure Storage. We make a solid state storage appliance that uses compression and deduplication to bring down the price of flash while still beating the pants off performance disk arrays.
While they ditched the really massive heavyweight branch prediction, it's not accurate to say it has none. Simple branch prediction is really cheap, and pretty much required for pipelined CPUs to have good performance.
Plus, they have something that appears to make up for what they lost-- the hardware scout. It executes ahead in the instruction stream, preloading necessary data and instructions into the cache. There's lots of articles accessible via google if you want to learn more: Sun's Marc Tremblay has been giving a lot of talks about how it works.
It's not the same deal as sound, because they're claiming benefits from that a process that, done right, should only affect frequencies far above the human range of hearing.
And if it's done wrong, that's like claiming that bleach tastes good on mashed potatoes because your bleach happens to be polluted with salt.
If you upsample a track from, say, 44.1 to 88.2 you can do it perfectly. If you go from 44.1 to 48, or something that's not a multiple of your original sample rate, you have to interpolate a lot of the samples. And that's not a perfect process, because by definition you're guessing at what some of the samples should be.
Yeah, so maybe it's not aliasing, technically. Maybe I should have said "artifacts". If you use a crappy interpolation algorithm (which I believe mass market sound chips have been known to do) you get damage across a wide range of frequencies, much worse than if you had just left the signal alone.
It's possible to do very high quality upsampling; the algorithms are just computationally expensive. So I don't know how good theirs is, and I waffled by saying "at best... at worst"
That line about MP3s sounding better at 96kHz is a bunch of marketing BS.
There are reasons for 24bit/96kHz, but upsampling just to play it out of a speaker isn't one of them. That's kind of like printing out something at 2400dpi only to scan it back in again at 300. At best, you're going to wind up with exactly the same thing, while at worst you're going to have a bunch of aliasing artifacts from the upsampling.
Upsampling for playback is worthless even if your source material is perfect CD audio. Taking something even worse than that (MP3) and upsampling it is just turd polishing.
Want better sound? Buy better speakers. And a sound card that has high-quality analog components. The digitial part is not the weak part of computer sound playback. Hard to market that, though: "Now with 10db more S/N! And better capacitors!"
24bit/96kHz is good for doing high quality recordings, then manipulating the sound and mixing it. Once that's done there's no point in distributing it in anything better than 16/44.1, if all that's ever done with it after that is playback. If you want your listeners to be able to do their own remixes, that may be another story, but then you have to distribute separate mixer tracks anyway...
You're assuming that a hard drive will always return the data stored on the platter, right or wrong. In truth, hard drives use enough CRC bits internally that the data will be return correctly, or not at all.
That means with two mirrored drives, you know precisely which one went bad, because it's the one that returns an error when you try to read certain sectors.
Badnarik believes that the federal income tax has no legal authority and that people are justified in refusing to file a tax return until such time as the IRS provides them with an explanation of its authority to collect the tax. He hadn't filed income tax returns for several years. He moved from California to Texas because of Texas' more liberal gun laws, but he refused to obtain a Texas driver's license because the state requires drivers to provide their fingerprints and Social Security numbers. He has been ticketed several times for driving without a license; sometimes he has gotten off for various technical legal reasons, but on three occasions he has been convicted and paid a fine. He also refused to use postal ZIP codes, seeing them as "federal territories."
...He proposed that convicted felons serve the first month of their sentence in bed so that their muscles would atrophy and they'd be less trouble for prison guards and to blow up the U.N. building on the eighth day of his administration, after giving the building's occupants a chance to evacuate.
Why is everyone obsessing about the timestamping problem? Didn't anyone read past page 6? The entire encryption system is broken, meaning there's a man-in-the-middle attack to sniff or alter any game traffic on the wire. The paper seems to vaguely suggest that even ICC credit card payments get sent using this completely bogus cipher.
That's a fair criticism. Power lines, which various groups claim cause cancer through RF, do correlate with poor people. I figure that one claim of RF-cancer causation is just as good as another, but I guess I'll wait for a more expert debunking.
Repeat after me: correlation is not causation.
Yes, people near power transmission towers and antennas get cancer more frequently.
But poor people tend to live in the houses next to unsightly power lines or antennas. And poor people have higher cancer risk, because they tend to be exposed to more pollution and hazardous substances, live under higher stress, and are less likely to get proper health care.
Besides, you get more radation from your cellphone.
Funny how the courtroom sketch of him doesn't look like that at all. Not really worth picking on Fox News, though. CNN and lots of other news sites are using the same photo with the evil beard.
GSSAPI patches for openssh are here. The openssh maintainers aren't real impressed with the GSSAPI spec, and aren't terribly enthusiastic about supporting it, so there's some resistance to merging it.
Metcalfe has a habit of saying stupid things, I wonder why people keep listening to him. One great invention thirty years ago, paired with a huge ego, does not an oracle make.
Ah, I understand. You believe that "based in part" means "we looked at it and then produced our own version".
I think "based in part" means "we copied the code and then added a bunch of our own stuff". I still believe that my reading is correct, because if they had _not_ copied linux code nobody would have been able to find linux code in Castle's code.
I've written PCI config code before, and it's not so simple that there's only one (or ten) ways to do it. There's no way that Castle wound up with identical code (which was the original complaint) unless they deliberately copied some source from the Linux kernel.
I've read the press release, and I don't see what you think is incorrect about my post. Can you be more specific?
They say "the hardware abstraction layer (roughly analogous to a PC's BIOS) has it's PCI allocation and bridge setup based in part on the following functions from the Linux kernel sources"...
Even though they don't say "GPL" in there, there isn't any other legitimate way for them to "base in part" their code on linux code other than under the GPL conditions.
They're saying that their kernel does not include GPLed code, but another program of theirs (called the HAL), a separate piece of software, DOES include GPLed code and source will be available for this program.
I'm not sure I believe them; hiding the function names after a complaint sure does seem like they know they're doing something wrong. But that's what they're saying.
Remember, too, that just offering source isn't the only requirement of the GPL. You're also required to notify users that the code is GPLed and tell them source is available. If they haven't done this part then they're also in violation.
My boss' boss (who is quite sharp technically as well as an attorney) thinks that the GPL is stupid because it doesn't read like it was written by a lawyer. He doesn't object to the principles and methods involved-- he's just disgusted by the unlawyerly writing. He says it was written by an amateur, not a lawyer, giving the impression that everyone using it is an amateur, and not serious about their work. What would you say to that?
Slashdot Mirror
The argument against PCIe is that if anything on that particular server fries, all of your data is now trapped inside a dead server; there's no way to fail over quickly. If you care about high availability you probably want to look into a SAN flash appliance, i.e. Pure Storage. (shameless plug)
This is generally true; shared _disk_ will suck once you have multiple VMs accessing it because they'll each try to optimize for seeks internally, but the combined I/O stream will be relatively seek-heavy and wreck performance. Where I work we call this the "I/O blender."
Solid state storage can solve this, but then you have to balance cost, high-availability (if your data is on a direct-attached PCIe card, what do you do when that motherboard dies?) and performance.
This is the problem that a new generation of storage companies are trying to solve. Shameless plug: Pure Storage. We make a solid state storage appliance that uses compression and deduplication to bring down the price of flash while still beating the pants off performance disk arrays.
While they ditched the really massive heavyweight branch prediction, it's not accurate to say it has none. Simple branch prediction is really cheap, and pretty much required for pipelined CPUs to have good performance.
Plus, they have something that appears to make up for what they lost-- the hardware scout. It executes ahead in the instruction stream, preloading necessary data and instructions into the cache. There's lots of articles accessible via google if you want to learn more: Sun's Marc Tremblay has been giving a lot of talks about how it works.
It's not the same deal as sound, because they're claiming benefits from that a process that, done right, should only affect frequencies far above the human range of hearing.
And if it's done wrong, that's like claiming that bleach tastes good on mashed potatoes because your bleach happens to be polluted with salt.
If you upsample a track from, say, 44.1 to 88.2 you can do it perfectly. If you go from 44.1 to 48, or something that's not a multiple of your original sample rate, you have to interpolate a lot of the samples. And that's not a perfect process, because by definition you're guessing at what some of the samples should be.
Yeah, so maybe it's not aliasing, technically. Maybe I should have said "artifacts". If you use a crappy interpolation algorithm (which I believe mass market sound chips have been known to do) you get damage across a wide range of frequencies, much worse than if you had just left the signal alone.
It's possible to do very high quality upsampling; the algorithms are just computationally expensive. So I don't know how good theirs is, and I waffled by saying "at best... at worst"
That line about MP3s sounding better at 96kHz is a bunch of marketing BS.
There are reasons for 24bit/96kHz, but upsampling just to play it out of a speaker isn't one of them. That's kind of like printing out something at 2400dpi only to scan it back in again at 300. At best, you're going to wind up with exactly the same thing, while at worst you're going to have a bunch of aliasing artifacts from the upsampling.
Upsampling for playback is worthless even if your source material is perfect CD audio. Taking something even worse than that (MP3) and upsampling it is just turd polishing.
Want better sound? Buy better speakers. And a sound card that has high-quality analog components. The digitial part is not the weak part of computer sound playback. Hard to market that, though: "Now with 10db more S/N! And better capacitors!"
24bit/96kHz is good for doing high quality recordings, then manipulating the sound and mixing it. Once that's done there's no point in distributing it in anything better than 16/44.1, if all that's ever done with it after that is playback. If you want your listeners to be able to do their own remixes, that may be another story, but then you have to distribute separate mixer tracks anyway...
You're assuming that a hard drive will always return the data stored on the platter, right or wrong. In truth, hard drives use enough CRC bits internally that the data will be return correctly, or not at all. That means with two mirrored drives, you know precisely which one went bad, because it's the one that returns an error when you try to read certain sectors.
giant advertisement:
I'm thinking that maybe "the guy that almost crashed a bunch of planes" is not the name they were looking for.(I'm not making this up- that's really the ad I'm seeing.)
Why is everyone obsessing about the timestamping problem? Didn't anyone read past page 6? The entire encryption system is broken, meaning there's a man-in-the-middle attack to sniff or alter any game traffic on the wire. The paper seems to vaguely suggest that even ICC credit card payments get sent using this completely bogus cipher.
That's a fair criticism. Power lines, which various groups claim cause cancer through RF, do correlate with poor people. I figure that one claim of RF-cancer causation is just as good as another, but I guess I'll wait for a more expert debunking.
But do poor Koreans smoke more and eat worse? :)
I would agree with you, except for the fact that this has already been researched to death in the past 20 years.
Repeat after me: correlation is not causation. Yes, people near power transmission towers and antennas get cancer more frequently. But poor people tend to live in the houses next to unsightly power lines or antennas. And poor people have higher cancer risk, because they tend to be exposed to more pollution and hazardous substances, live under higher stress, and are less likely to get proper health care. Besides, you get more radation from your cellphone.
A ship poorly built is a raft
an airplane poorly built is a cart
a rocket poorly built is a bomb.
You can download the tracks you buy as MP3 (fixed or variable rate), WAV, OGG, or FLAC. No DRM. No country restrictions.
According to Redhat Bugzilla bug 104917, Red Hat has never shipped openssh 3.7, so they're not vulnerable to this. No workaround or fix is needed.
Funny how the courtroom sketch of him doesn't look like that at all. Not really worth picking on Fox News, though. CNN and lots of other news sites are using the same photo with the evil beard.
GSSAPI patches for openssh are here. The openssh maintainers aren't real impressed with the GSSAPI spec, and aren't terribly enthusiastic about supporting it, so there's some resistance to merging it.
Metcalfe has a habit of saying stupid things, I wonder why people keep listening to him. One great invention thirty years ago, paired with a huge ego, does not an oracle make.
I think "based in part" means "we copied the code and then added a bunch of our own stuff". I still believe that my reading is correct, because if they had _not_ copied linux code nobody would have been able to find linux code in Castle's code.
I've written PCI config code before, and it's not so simple that there's only one (or ten) ways to do it. There's no way that Castle wound up with identical code (which was the original complaint) unless they deliberately copied some source from the Linux kernel.
Finally, proof that all Ask Slashdot questions could be more quickly answered by simply checking with Google :)
They say "the hardware abstraction layer (roughly analogous to a PC's BIOS) has it's PCI allocation and bridge setup based in part on the following functions from the Linux kernel sources"...
Even though they don't say "GPL" in there, there isn't any other legitimate way for them to "base in part" their code on linux code other than under the GPL conditions.
They're saying that their kernel does not include GPLed code, but another program of theirs (called the HAL), a separate piece of software, DOES include GPLed code and source will be available for this program. I'm not sure I believe them; hiding the function names after a complaint sure does seem like they know they're doing something wrong. But that's what they're saying. Remember, too, that just offering source isn't the only requirement of the GPL. You're also required to notify users that the code is GPLed and tell them source is available. If they haven't done this part then they're also in violation.
My boss' boss (who is quite sharp technically as well as an attorney) thinks that the GPL is stupid because it doesn't read like it was written by a lawyer. He doesn't object to the principles and methods involved-- he's just disgusted by the unlawyerly writing. He says it was written by an amateur, not a lawyer, giving the impression that everyone using it is an amateur, and not serious about their work. What would you say to that?