I think he fails to build up satisfying endings because he's basically a cold-blooded creature. His emotional range is rather narrow, and a good ending needs to be backed by a strong emotional apex. Pretty much like sex.:-)
In the light of the sex comparison, Stephenson seems kind of frigid to me.
What's to become of the Epiphyte corporation and its data crypt plan?
They filed for bankruptcy while the founders ran away to spend idle lifetimes sipping margueritas in the Bahamas.:-)
Sorry, couldn't resist, but it's obvious that Cryptonomicon was written during the dot-com bubble. We all know what happened to that bubble, therefore all follow-ups must depict a technology-pessimistic view.
Well, unless said follow-ups are set in a distant future.:-)
TCP is layer 4, therefore it is a container for VoIP (H.323, SIP, IAX, Skinny, etc.) and P2P (Napster, etc.). VoIP and P2P are in the layers above 4. Please go back and re-read the manuals.
Live audio has quite a bit in common with transferring large files (movies, ISO images) at the transport level, it is more or less a continuous stream of data. But there are dissimilarities as well, i agree with that.
I believe the main reason for choosing SIP has to do with firewall circumventing, not with similarities with audio. If VoIP explodes, SIP will be as ubiquitous as HTTP.
Now, if you're using some kind of external SIP proxy, then "automagically" all your clients behind NAT will work. That's all that's required for SIP to work through NAT.
Since it's good for VoIP, it's good to carry persistent streams of data, which is typical for P2P.
Also, it's more likely that SIP would be allowed by default through various corporate firewalls, while typical P2P protocols will be blocked. Think of it as a firewall circumventing feature. I was actually impressed, the idea is clever.
Also keep in mind that the Efficeon machine is only a cluster, hence it works well for (like you pointed out) heavy-duty algorithms that are easy to compute on clusters. Some algorithms cannot be computed efficiently on clusters; for those, you rather need a single-image supercomputer, such as the SGI Altix. Unfortunately, many of the examples you provided fall into this category.:-)
The SGI supercomputers are single-system-image machines. This Transmeta-based thing is essentially just a cluster.
Clusters are easy to design and build. Single-image machines are hard. Unfortunately, there are entire classes of mathematical algorithms that cannot be solved efficiently except on single-image SMP machines.
Why do you think the market of huge single-system-image thousands-of-CPUs supercomputers has, essentially, only one vendor? (Silicon Graphics, a.k.a. SGI) We're not counting here superclusters and things like that, where IBM, HP and so on are happy to sell you a bunchload of systems. We're only discussing pure single-image SMP.
It's not either black or white, there are shades of gray in between.
I said "cannot be translated into 100% parallel algorithms", with an emphasis on "100%". Of course, completely serial algos will not benefit at all from multi-CPU processing, no matter what architecture.
What i was refering to were the algos that can only be partially parallelized, that require a lot of communication between the node. But you said that yourself.;-)
How did this comment got moderated "informative"? There's definitely something wrong with the moderators today.
SGI Altix uses the Intel compilers. They're pretty damn good on IA64. They're available today.
Also, the massively parallel software is already up'n'running. NASA has been using for decades SGI supercomputers - traditionally it's been the MIPS/Irix architecture. A while ago, when SGI told NASA that they were going to migrate to Intel/Linux, NASA simply recompiled their software to Linux, which is not too difficult, since Irix is pretty much standard Unix (i did some porting from Linux to Irix and often the software simply compiles with no change). Also, Altix systems are essentially the same hardware architecture as MIPS-based SGI Origin with the exception of the CPU (and a different OS on top), so the differences are really not that big; it's just the transition from Irix to Linux.
In Sun's head, Dr. Jekyll and Mr. Hyde are still fighting for supremacy. They've got this split personality syndrome and are still scratching their collective head figuring out which way to go. By the time they make up their mind, i'm afraid the sun will be setting. (pun intended)
The "firmware" (the equivalent of BIOS) they have on the Altix is pretty damn smart, it's like an OS of it's own. It can do diagnostics, and inventory and a truckful of other things. Powering up a huge complex beast such as an Altix is no easy task. You need lots of "intelligence" at the hardware level to do that.
It makes you believe this supercomputer is made out of commodity components. That's blatantly false.
The SGI systems are highly proprietary equipments that provide very large bandwidth between the nodes, extremely low latency and tight integration. They're not regular Beowulf clusters. They really are single systems with hundreds or thousands of CPUs, all of them running the same single instance of the OS (as opposed to typical clusters which run one OS instance per node). Because of the tight integration, the software does not have to obey the same constraints as when running on commodity clusters. Especially the requirement for total parallelization does not stand anymore. Therefore, problems which cannot be translated into 100% parallel algorithms, and therefore do not run efficiently on commodity clusters, are easily tackled on SGI supercomputers. That's why they can charge a high price on their systems - because they can solve problems that are not accessible to "normal" computers.
That being said, the system at NASA is indeed a cluster, but it's a "small" cluster (a handful of nodes), each node being a supercomputer with hundreds of CPUs. It's a hybrid that provides the best of both worlds.
Can't find that link anymore, sorry, but i saw a blog page or a newspiece or something where someone from SGI explained how they benchmarked various CPU architectures and Itanium is the one that performs best in a thousand-CPU environment. I'm sorry for being too vague, but that's all i remember.
The reason SGI uses Itanium is that it performs better in a supercomputer environment than Opteron. I'm not sure which CPU might be faster in a "normal" computer (1...4 CPUs), it may or may not be Opteron, but in a single-OS-image supercomputer Intel performs better.
Yeah, in my home computer i also got an AMD, i think there's more bang for the buck with AMD in that area.
What? You call Audigy a professional card? (rolling on the floor laughing my guts out)
Have you tried to do some serious (i try to avoid the overused word "professional") with an Audigy? Then compare it with a (well, there's no better substitute) truly professional card? Did you notice that Audigy actually has a hard, fixed sampling rate of 48kHz and works around that with tricks in the DAC/ADCs? Is that what you call professional?
In any case, Audigy support is fine in Linux. One of the systems that i own has an Audigy2 Platinum and it works fine in conjunction with an external synthesizer (Alesis), several sequencers (Muse, Rosegarden, Seq24), digital multitrack recorders (Ardour), softsynths (ZynAddSubFX, amSynth, Hydrogen), etc.
FYI, professional cards supported by Linux are RME and MAudio. They work very well with ALSA. All features that they have on Windows are present in Linux as well. There is indeed a lack of support for Motu.
Linux still has trouble supporting [...] more-than-2 channel support for sound
Are you kidding?
Hardware: Linux has perfect support for cards such as RME and MAudio which offer multichannel support, 8 channels per card or more. There are people using multiple cards in one computer, having 16 or more channels active at the same time.
Drivers: ALSA works fine with many channels at the same time.
Software: Ardour can use how many sound tracks in parallel? A couple hundred?:-) (not that the CPU will keep up with it anyway)
You're probably thinking about Audiophile 2496, or maybe even the Delta 1010LT. Still, the "big" Delta 1010 (not the LT) is perhaps better, since it has balanced analog. But it's more expensive (about $600).
Send them an email, ask them to support Linux, etc. They don't have to write the driver themselves. Often, a company will sit on their card's specs and not release them to the open source community. This will prevent a free driver to be written.
Slashdot Mirror
I think he fails to build up satisfying endings because he's basically a cold-blooded creature. His emotional range is rather narrow, and a good ending needs to be backed by a strong emotional apex. Pretty much like sex. :-)
In the light of the sex comparison, Stephenson seems kind of frigid to me.
What's to become of the Epiphyte corporation and its data crypt plan?
:-)
:-)
They filed for bankruptcy while the founders ran away to spend idle lifetimes sipping margueritas in the Bahamas.
Sorry, couldn't resist, but it's obvious that Cryptonomicon was written during the dot-com bubble. We all know what happened to that bubble, therefore all follow-ups must depict a technology-pessimistic view.
Well, unless said follow-ups are set in a distant future.
TCP is layer 4, therefore it is a container for VoIP (H.323, SIP, IAX, Skinny, etc.) and P2P (Napster, etc.).
VoIP and P2P are in the layers above 4.
Please go back and re-read the manuals.
Live audio has quite a bit in common with transferring large files (movies, ISO images) at the transport level, it is more or less a continuous stream of data. But there are dissimilarities as well, i agree with that.
I believe the main reason for choosing SIP has to do with firewall circumventing, not with similarities with audio. If VoIP explodes, SIP will be as ubiquitous as HTTP.
True.
Now, if you're using some kind of external SIP proxy, then "automagically" all your clients behind NAT will work. That's all that's required for SIP to work through NAT.
Since it's good for VoIP, it's good to carry persistent streams of data, which is typical for P2P.
Also, it's more likely that SIP would be allowed by default through various corporate firewalls, while typical P2P protocols will be blocked. Think of it as a firewall circumventing feature.
I was actually impressed, the idea is clever.
Also keep in mind that the Efficeon machine is only a cluster, hence it works well for (like you pointed out) heavy-duty algorithms that are easy to compute on clusters. :-)
Some algorithms cannot be computed efficiently on clusters; for those, you rather need a single-image supercomputer, such as the SGI Altix. Unfortunately, many of the examples you provided fall into this category.
The SGI supercomputers are single-system-image machines. This Transmeta-based thing is essentially just a cluster.
Clusters are easy to design and build. Single-image machines are hard. Unfortunately, there are entire classes of mathematical algorithms that cannot be solved efficiently except on single-image SMP machines.
...high-CPU-count SMP is hard.
Why do you think the market of huge single-system-image thousands-of-CPUs supercomputers has, essentially, only one vendor? (Silicon Graphics, a.k.a. SGI) We're not counting here superclusters and things like that, where IBM, HP and so on are happy to sell you a bunchload of systems. We're only discussing pure single-image SMP.
It's not either black or white, there are shades of gray in between.
;-)
I said "cannot be translated into 100% parallel algorithms", with an emphasis on "100%". Of course, completely serial algos will not benefit at all from multi-CPU processing, no matter what architecture.
What i was refering to were the algos that can only be partially parallelized, that require a lot of communication between the node.
But you said that yourself.
How did this comment got moderated "informative"? There's definitely something wrong with the moderators today.
SGI Altix uses the Intel compilers. They're pretty damn good on IA64. They're available today.
Also, the massively parallel software is already up'n'running. NASA has been using for decades SGI supercomputers - traditionally it's been the MIPS/Irix architecture. A while ago, when SGI told NASA that they were going to migrate to Intel/Linux, NASA simply recompiled their software to Linux, which is not too difficult, since Irix is pretty much standard Unix (i did some porting from Linux to Irix and often the software simply compiles with no change).
Also, Altix systems are essentially the same hardware architecture as MIPS-based SGI Origin with the exception of the CPU (and a different OS on top), so the differences are really not that big; it's just the transition from Irix to Linux.
In Sun's head, Dr. Jekyll and Mr. Hyde are still fighting for supremacy. They've got this split personality syndrome and are still scratching their collective head figuring out which way to go. By the time they make up their mind, i'm afraid the sun will be setting. (pun intended)
The "firmware" (the equivalent of BIOS) they have on the Altix is pretty damn smart, it's like an OS of it's own. It can do diagnostics, and inventory and a truckful of other things.
Powering up a huge complex beast such as an Altix is no easy task. You need lots of "intelligence" at the hardware level to do that.
Because it's a 20-nodes cluster, each node being a supercomputer with 512 CPUs.
u ly/supercomputing_ctr.html
The article was written, unfortunately, by a rather clueless journalist. Here's a link to the proper information:
http://www.sgi.com/newsroom/press_releases/2004/j
It makes you believe this supercomputer is made out of commodity components.
That's blatantly false.
The SGI systems are highly proprietary equipments that provide very large bandwidth between the nodes, extremely low latency and tight integration. They're not regular Beowulf clusters. They really are single systems with hundreds or thousands of CPUs, all of them running the same single instance of the OS (as opposed to typical clusters which run one OS instance per node).
Because of the tight integration, the software does not have to obey the same constraints as when running on commodity clusters. Especially the requirement for total parallelization does not stand anymore.
Therefore, problems which cannot be translated into 100% parallel algorithms, and therefore do not run efficiently on commodity clusters, are easily tackled on SGI supercomputers.
That's why they can charge a high price on their systems - because they can solve problems that are not accessible to "normal" computers.
That being said, the system at NASA is indeed a cluster, but it's a "small" cluster (a handful of nodes), each node being a supercomputer with hundreds of CPUs. It's a hybrid that provides the best of both worlds.
Can't find that link anymore, sorry, but i saw a blog page or a newspiece or something where someone from SGI explained how they benchmarked various CPU architectures and Itanium is the one that performs best in a thousand-CPU environment.
I'm sorry for being too vague, but that's all i remember.
The reason SGI uses Itanium is that it performs better in a supercomputer environment than Opteron. I'm not sure which CPU might be faster in a "normal" computer (1...4 CPUs), it may or may not be Opteron, but in a single-OS-image supercomputer Intel performs better.
Yeah, in my home computer i also got an AMD, i think there's more bang for the buck with AMD in that area.
It's all explained in the web pages that this announcement links to.
Where's the FLAC support? Hello? Apple? Anyone?...
What? You call Audigy a professional card? (rolling on the floor laughing my guts out)
Have you tried to do some serious (i try to avoid the overused word "professional") with an Audigy? Then compare it with a (well, there's no better substitute) truly professional card?
Did you notice that Audigy actually has a hard, fixed sampling rate of 48kHz and works around that with tricks in the DAC/ADCs? Is that what you call professional?
In any case, Audigy support is fine in Linux. One of the systems that i own has an Audigy2 Platinum and it works fine in conjunction with an external synthesizer (Alesis), several sequencers (Muse, Rosegarden, Seq24), digital multitrack recorders (Ardour), softsynths (ZynAddSubFX, amSynth, Hydrogen), etc.
FYI, professional cards supported by Linux are RME and MAudio. They work very well with ALSA. All features that they have on Windows are present in Linux as well.
There is indeed a lack of support for Motu.
Linux still has trouble supporting [...] more-than-2 channel support for sound
:-) (not that the CPU will keep up with it anyway)
Are you kidding?
Hardware: Linux has perfect support for cards such as RME and MAudio which offer multichannel support, 8 channels per card or more. There are people using multiple cards in one computer, having 16 or more channels active at the same time.
Drivers: ALSA works fine with many channels at the same time.
Software: Ardour can use how many sound tracks in parallel? A couple hundred?
Muse is comparable with Rosegarden, not with Ardour.
Muse and Rosegarden are sequencers (MIDI and stuff), while Ardour is a DAW (digital multitrack sound recorder).
MAudio
You're probably thinking about Audiophile 2496, or maybe even the Delta 1010LT.
Still, the "big" Delta 1010 (not the LT) is perhaps better, since it has balanced analog. But it's more expensive (about $600).
If the customers don't ask them to support Linux, how would they become aware of the need?
lack of familiar professional mixing/software
Have you looked at JAMin?
Send them an email, ask them to support Linux, etc.
They don't have to write the driver themselves. Often, a company will sit on their card's specs and not release them to the open source community. This will prevent a free driver to be written.