Another solution is to use a different type of file system--one that protects itself from corruption or uses some sort of journaling to reduce the need for a full-blown fsck.
fsck has always been a pain. There are several solutions, though.
Much of the time used by fsck is for reading all the inodes. If you reduce the number of inodes, you speed up fsck. I did this with my MP3 partition. Unfortunately, ext2 won't let me have one inode per 1024K. Since with such large storage systems most people will be storing very large (by today's standards) files (excluding news/mail archives with one file per message), it makes sense to alter the file system to reserve fewer inodes. Using dynamic inode allocation makes a lot of sense here. You can also save some time by using larger block groups and larger block sizes, but the advantages there will be relatively insignificant.
Well, not really, but my employer, EMC, has been selling multi-terabyte storage systems for years. If you've got the money, we'll set up a 10TB system for you.
Generally, EMC storage systems are partitioned into separate volumes, which show up as separate devices when viewed by a host computer.
Still, the point is that people are dealing with storage systems larger than what we're talking about here.
Not that I believe in this technology, but one big consumer application would be digital VCRs. You could record a thousand hours of DVD-quality video with one of those. So you could record every episode of your favorite TV shows. Or get HBO for a few months and build up a library of movies.
Of course, this is still a long way from being able to record every channel all the time. With only 100 channels, you would run out of storage within a day. You could, though, pick your favorite channels, set up a profile of stuff you know you don't want to watch (e.g., golf), and have it record everything that doesn't fit the profile. You would then have a week or so after something was recorded to decide to watch or save it before it is recorded over by newer stuff.
So when will we be able to order 8-way systems with this chipset from our favorite Linux hardware vendors?
Penguin has been offering 8-way systems for a while using another chipset solution, but the price for those systems is over $100,000, while 4-way systems are more like $12,000, last I checked.
Right now, the parties are playing PR games with taxes. While they would like to pass real changes, the Republicans know that the tax cuts they're passing will never become law with a Democrat in the White House, and the Democrats know that a national Internet sales tax will never become law until they control both houses of Congress.
If Republicans really wanted to cut taxes, they would pass their massive tax cut as a bunch of separate bills, each with one cut--the lower rates, capital gains cuts, marriage penalty relief, and estate tax elimination. Then Clinton would sign one or two and veto the rest, but at least we would get some tax cuts. Of course, that's bad politics because then it's easier to portray some of those as tax cuts for the rich, playing into the class warfare Democrats always use when tax cuts are proposed.
As long as cryptography is legal, the disclosure that governments are monitoring communications will serve the public interest in encouraging additional use of cryptography.
Seriously, you shouldn't expect anything sent over an open channel to be private. Now the US government is helping people to realize that. Now if only we can get lucky on some cryptography export rule relaxation, we'll be all set.
And I just saw an article reporting that Reno, I believe in a letter to the German government, stated her desire to see all cryptographic products banned from the Internet. Despite the technical and legal impossibilities involved, that was one more indication that the JustUs Department is solidly anti-privacy.
Interestingly, this seems to be a non-partisan issue, so I can't fall back on my traditional response of complaining about the Democrats. I've yet to see either party integrate a cohesive position on privacy and technology.
[Note that SMP is very different from clustering, so forget all you've heard about Beowulf, Mosix, and such for this discussion.]
Yes, Linux supports SMP. If you have a bunch of user processes that are doing mostly user-space computation, then you should get a mostly linear speedup even with a 2.0 kernel. But if the processes depend on a bunch of kernel services, watch the kernel version carefully.
You see, with SMP the kernel has to be sure that two processors don't try to modify the same data structure at the same time so as to avoid confusion (e.g, deadlock or a crash). The 2.0 kernels supported SMP by putting one big lock on the kernel, so only one process can be doing something in the kernel at a time. One of the major features of 2.2 is that this one big lock was broken up, allowing Linux to scale far better. Unfortunately, there are still many places where the locks need to be broken up further, and work is ongoing.
I believe that it was such a locking issue that caused Linux problems in the Mindcraft benchmarks when multiple network cards were in use.
So the performance of SMP will vary depending on your application and your system. (I've heard, for example, that building the kernel on a 4-way box gave a 3.7 times speedup in one case, which is pretty good.)
Note that many of the commercial Unixes scale better than Linux (i.e., have finer grained locking). I know Solaris has a very good reputation for large numbers of processors. I would suspect that Irix must be good, based on the systems that SGI sells. Linux is getting there. Upto 4 CPUs should be fine for most tasks. I haven't heard reports of Linux with more than 4 SMP CPUs.
Thanks to sites like eBay, people are sending more mail than ever. Think of all those checks that people are mailing. Sure, eventually, this can all be done electronically, but not for a few years, at least. Also, while big businesses may get better deals for packages from UPS and the likes, most consumers are better off with US Mail. Again, eBay and Amazon Auctions result in many more small packages from the USPS.
This brings up an important point: We need a good resource to check to see which kernel versions are stable.
Imagine the following fictional site (infomation is fictional, too):
www.kernelbugs.org/2.2/
2.2.0 Has severe problem with the wigglenet driver.
2.2.6 Improved memory management for better SMP performance.
2.2.8 Severe memory system problems
2.2.9 Mysterious and rare file corruption bug, possibly actually due to faulty hardware.
2.2.10 File corruption bug still not fixed.
Most stable version: 2.2.7 High performance version: 2.2.10
The idea is to have a review of different kernel releases at a higher level than just what features have been tweaked. Something where an experienced user can figure out easily what kernel makes the most sense.
Alan may be mostly running the 2.2 tree. He certainly ran the 2.0 tree long after Linus had moved totally to 2.1. However, when it comes time to release the next official version, Alan still gets Linus's approval first.
In many cases, the AC patches include things that Linus has specifically said won't make it into his kernel, or at least not in the 2.2 kernel. I'm not sure if that is still the case, though, now that 2.3 is underway--most of those patches that Linus wasn't ready to accept into 2.2 are now in 2.3.
Anyway, we can expect to see Alan working on both 2.2 and 2.3 until 2.4 comes out.
2.2.10 was released on June 13, based on the timestamps I see at my kernel.org mirror site. However, there are some obscure file corruption bugs in 2.2.9 and 2.2.10 that they've been working on finding, so I would consider 2.2.8 to be the latest stable version. That bug is the reason you haven't seen 2.2.11 in over a month.
Personally, I live on the edge with 2.3.11-pre5. So what if FAT file system support is broken? That's what development kernels are all about.
Does a 200W power supply really consume 200W? I thought that that just meant that it was capable of providing upto 200W of power, but usually consumed less, based on demand by the motherboard, drives, and such. If so, then using APM features to reduce power consumption on desktops makes a lot of sense.
I've always thought it would be nice to have my own power meter--one that I could plug into an outlet, and then plug a power strip into it. Then I would really know how much power I used for my computer (or how much power the VCR uses when it is "off" but updating the clock and listening for remote control power on).
Intel cuts prices all the time. It used to be a quarterly event, and a year or two ago, they started doing it more frequently; the PC makers wanted smaller but more frequent cuts to help avoid huge overnight price changes.
Now this isn't to say that Intel is being more agressive just because AMD is looking like serious competition; I only want to give some perspective.
Remember when The Open Group (which I worked for at the time) decided that it needed more money to fund X11, so it changed the licensing terms for R6.4?
Well, The Open Group isn't doing X development anymore.
And RDist isn't nearly as essential or complicated as X (and there are competing products), so why do they think they can make money with that model?
So if 6.0 has a free license, and 6.1 doesn't, why would we want to use 6.1? Anyone who actually uses rdist 6.x care to comment on any feature/bug changes that real users might care about?
I can see the next version of Solaris being called "Solaris Linux" now, and including the Linux source code. Hey, it is a Linux distribution, so rdist is free!
The license doesn't address the issue of redistributing modified versions of the software. Also, it is only free under certain circumstances. These factors combine to make it a bad basis for a branch if future versions have a more restrictive license. Hence, the community would be better served by branching from a version with a free software license.
Amazon has always said that books were just the start. I like what they're doing. Who cares what shows up on the default page? All you ever do is click on the tab for the category you want and then enter something in the search box.
The inability to get the raw scrambled data off the disk was cited as the major reason you can't use a Windows player with Wine to watch DVD movies under Linux.
Now all that is needed is to export the raw stream through Wine with the right API. That part shouldn't be secret, so there's no longer any big stumbling block for DVD under Linux.
(And of course, this will allow for easier use of debuggers to reverse-engineer the code.)
Slashdot Mirror
So does anyone have a list of names of people who were involved with WEB? It would be interesting to track down where they work now.
I wonder what other scams they've tried to defraud venture capitalists?
Another solution is to use a different type of file system--one that protects itself from corruption or uses some sort of journaling to reduce the need for a full-blown fsck.
fsck has always been a pain. There are several solutions, though.
Much of the time used by fsck is for reading all the inodes. If you reduce the number of inodes, you speed up fsck. I did this with my MP3 partition. Unfortunately, ext2 won't let me have one inode per 1024K. Since with such large storage systems most people will be storing very large (by today's standards) files (excluding news/mail archives with one file per message), it makes sense to alter the file system to reserve fewer inodes. Using dynamic inode allocation makes a lot of sense here. You can also save some time by using larger block groups and larger block sizes, but the advantages there will be relatively insignificant.
Well, not really, but my employer, EMC, has been selling multi-terabyte storage systems for years. If you've got the money, we'll set up a 10TB system for you.
Generally, EMC storage systems are partitioned into separate volumes, which show up as separate devices when viewed by a host computer.
Still, the point is that people are dealing with storage systems larger than what we're talking about here.
Not that I believe in this technology, but one big consumer application would be digital VCRs. You could record a thousand hours of DVD-quality video with one of those. So you could record every episode of your favorite TV shows. Or get HBO for a few months and build up a library of movies.
Of course, this is still a long way from being able to record every channel all the time. With only 100 channels, you would run out of storage within a day. You could, though, pick your favorite channels, set up a profile of stuff you know you don't want to watch (e.g., golf), and have it record everything that doesn't fit the profile. You would then have a week or so after something was recorded to decide to watch or save it before it is recorded over by newer stuff.
So when will we be able to order 8-way systems with this chipset from our favorite Linux hardware vendors?
Penguin has been offering 8-way systems for a while using another chipset solution, but the price for those systems is over $100,000, while 4-way systems are more like $12,000, last I checked.
Right now, the parties are playing PR games with taxes. While they would like to pass real changes, the Republicans know that the tax cuts they're passing will never become law with a Democrat in the White House, and the Democrats know that a national Internet sales tax will never become law until they control both houses of Congress.
If Republicans really wanted to cut taxes, they would pass their massive tax cut as a bunch of separate bills, each with one cut--the lower rates, capital gains cuts, marriage penalty relief, and estate tax elimination. Then Clinton would sign one or two and veto the rest, but at least we would get some tax cuts. Of course, that's bad politics because then it's easier to portray some of those as tax cuts for the rich, playing into the class warfare Democrats always use when tax cuts are proposed.
As long as cryptography is legal, the disclosure that governments are monitoring communications will serve the public interest in encouraging additional use of cryptography.
Seriously, you shouldn't expect anything sent over an open channel to be private. Now the US government is helping people to realize that. Now if only we can get lucky on some cryptography export rule relaxation, we'll be all set.
And I just saw an article reporting that Reno, I believe in a letter to the German government, stated her desire to see all cryptographic products banned from the Internet. Despite the technical and legal impossibilities involved, that was one more indication that the JustUs Department is solidly anti-privacy.
Interestingly, this seems to be a non-partisan issue, so I can't fall back on my traditional response of complaining about the Democrats. I've yet to see either party integrate a cohesive position on privacy and technology.
[Note that SMP is very different from clustering, so forget all you've heard about Beowulf, Mosix, and such for this discussion.]
Yes, Linux supports SMP. If you have a bunch of user processes that are doing mostly user-space computation, then you should get a mostly linear speedup even with a 2.0 kernel. But if the processes depend on a bunch of kernel services, watch the kernel version carefully.
You see, with SMP the kernel has to be sure that two processors don't try to modify the same data structure at the same time so as to avoid confusion (e.g, deadlock or a crash). The 2.0 kernels supported SMP by putting one big lock on the kernel, so only one process can be doing something in the kernel at a time. One of the major features of 2.2 is that this one big lock was broken up, allowing Linux to scale far better. Unfortunately, there are still many places where the locks need to be broken up further, and work is ongoing.
I believe that it was such a locking issue that caused Linux problems in the Mindcraft benchmarks when multiple network cards were in use.
So the performance of SMP will vary depending on your application and your system. (I've heard, for example, that building the kernel on a 4-way box gave a 3.7 times speedup in one case, which is pretty good.)
Note that many of the commercial Unixes scale better than Linux (i.e., have finer grained locking). I know Solaris has a very good reputation for large numbers of processors. I would suspect that Irix must be good, based on the systems that SGI sells. Linux is getting there. Upto 4 CPUs should be fine for most tasks. I haven't heard reports of Linux with more than 4 SMP CPUs.
What I would like to see is a FIPSish tool for NTFS. That would be really helpful for installing Linux in corporate settings.
If they can put that into this tool, it could well become the standard partitioning tool.
Thanks to sites like eBay, people are sending more mail than ever. Think of all those checks that people are mailing. Sure, eventually, this can all be done electronically, but not for a few years, at least. Also, while big businesses may get better deals for packages from UPS and the likes, most consumers are better off with US Mail. Again, eBay and Amazon Auctions result in many more small packages from the USPS.
Thanks. I missed that.
This brings up an important point: We need a good resource to check to see which kernel versions are stable.
Imagine the following fictional site (infomation is fictional, too):
www.kernelbugs.org/2.2/
2.2.0
Has severe problem with the wigglenet driver.
2.2.6
Improved memory management for better SMP performance.
2.2.8
Severe memory system problems
2.2.9
Mysterious and rare file corruption bug, possibly actually due to faulty hardware.
2.2.10
File corruption bug still not fixed.
Most stable version: 2.2.7
High performance version: 2.2.10
The idea is to have a review of different kernel releases at a higher level than just what features have been tweaked. Something where an experienced user can figure out easily what kernel makes the most sense.
Alan may be mostly running the 2.2 tree. He certainly ran the 2.0 tree long after Linus had moved totally to 2.1. However, when it comes time to release the next official version, Alan still gets Linus's approval first.
In many cases, the AC patches include things that Linus has specifically said won't make it into his kernel, or at least not in the 2.2 kernel. I'm not sure if that is still the case, though, now that 2.3 is underway--most of those patches that Linus wasn't ready to accept into 2.2 are now in 2.3.
Anyway, we can expect to see Alan working on both 2.2 and 2.3 until 2.4 comes out.
2.2.10 was released on June 13, based on the timestamps I see at my kernel.org mirror site. However, there are some obscure file corruption bugs in 2.2.9 and 2.2.10 that they've been working on finding, so I would consider 2.2.8 to be the latest stable version. That bug is the reason you haven't seen 2.2.11 in over a month.
Personally, I live on the edge with 2.3.11-pre5. So what if FAT file system support is broken? That's what development kernels are all about.
Does a 200W power supply really consume 200W? I thought that that just meant that it was capable of providing upto 200W of power, but usually consumed less, based on demand by the motherboard, drives, and such. If so, then using APM features to reduce power consumption on desktops makes a lot of sense.
I've always thought it would be nice to have my own power meter--one that I could plug into an outlet, and then plug a power strip into it. Then I would really know how much power I used for my computer (or how much power the VCR uses when it is "off" but updating the clock and listening for remote control power on).
Intel cuts prices all the time. It used to be a quarterly event, and a year or two ago, they started doing it more frequently; the PC makers wanted smaller but more frequent cuts to help avoid huge overnight price changes.
Now this isn't to say that Intel is being more agressive just because AMD is looking like serious competition; I only want to give some perspective.
Disclaimer: I own stock in INTC, and not AMD.
Remember when The Open Group (which I worked for at the time) decided that it needed more money to fund X11, so it changed the licensing terms for R6.4?
Well, The Open Group isn't doing X development anymore.
And RDist isn't nearly as essential or complicated as X (and there are competing products), so why do they think they can make money with that model?
So if 6.0 has a free license, and 6.1 doesn't, why would we want to use 6.1? Anyone who actually uses rdist 6.x care to comment on any feature/bug changes that real users might care about?
I can see the next version of Solaris being called "Solaris Linux" now, and including the Linux source code. Hey, it is a Linux distribution, so rdist is free!
The license doesn't address the issue of redistributing modified versions of the software. Also, it is only free under certain circumstances. These factors combine to make it a bad basis for a branch if future versions have a more restrictive license. Hence, the community would be better served by branching from a version with a free software license.
Their web page says that it is freely distributable with *BSD, but the license only excepts Linux distributions from commercial redistribution.
So much for Red Hat FreeBSD. [That would be amusing.]
Or is there a different license agreement with the versions distributed with the BSDs?
So you can distribute it for free as a part of a Linux distribution, but not FreeBSD. Ugh.
Amazon has always said that books were just the start. I like what they're doing. Who cares what shows up on the default page? All you ever do is click on the tab for the category you want and then enter something in the search box.
More is better.
The inability to get the raw scrambled data off the disk was cited as the major reason you can't use a Windows player with Wine to watch DVD movies under Linux.
Now all that is needed is to export the raw stream through Wine with the right API. That part shouldn't be secret, so there's no longer any big stumbling block for DVD under Linux.
(And of course, this will allow for easier use of debuggers to reverse-engineer the code.)