First of all, the soonest the preempt patch would be merged, if at all would be in 2.5, and it is still in development. It only supports x86/32 and ARM so far...
"The thing about the preemptible kernel is that it is only for uniprocessor - SMP kernels aren't preemptible. "
This has been fixed over a month ago. UP and SMP kernels are now both preempt safe. So is a highmem (more than 2GB of ram) kernel.
"So unless you want the SMP case to be capable of tying up a processor for "too long" at a time, then you need to re-do each bit of code which is capable of long latencies anyway."
There are some small parts of the kernel that have been marked "not preempt safe". Fortunately, there aren't too many of them, and they're planned to be fixed in 2.5.
"The other thing which came up is that responsiveness of the system improved quite a bit recently with VM fixes (2.4.14 was the improved version, I think)."
Actually, the new VM was put in in 2.4.10pre11.
"It was a matter of the VM queueing up too much I/O (and the drivers trying to throttle it, instead of just throttling it all in the VM - or something like that)"
Actually that was one of the outward signs... The kernel favored the page cache over most anything else, and it grew particalarly large. That was since fixed, which allowed other problems to be noticed... The main problem was the swapping code. It would choose the wrong things to swap out, take too long to do so, and basically do bad things. Also there were some races in there that would cause kswapd to use 100% of the CPU (or just one cpu if you have SMP).
"The preemptible kernel won't solve that kind of problem"
Wow! The first thing that was 100% correct! Congratulations.
Don't forget that ext2/3 has indexed directories in development.
Right now, it's only for test partitions, but it is coming along nicely.
Ext2 has had changes from 2.0, 2.2 and 2.4 kernels, each time making it more efficient. Only recently has the actual format on disk changed, but that won't go in until 2.5 from what I can see.
Ext3 has three journal modes (available with the data=mode option on mount or in/etc/fstab):
Writeback: Don't care about ordering, just write in whatever order is most convenient. Possibly fastest for some loads... (This is basically the same level of journaling guarantees that you get from ReiserFS)
Ordered(default): Write out data before filesystem accounting information. (will protect all data modifications except for writing to the middle of the file - this is rare in most situations)
journal: Write all data to journal first (Guarantees all completed writes will be recoverable, including writes to the middle of files. This mode can be fastest with random writing because they all go directly to the journal - more effective write clustering...)
That said, ext3 does it's best to protect the data. Also, no JFS will protect you from hardware failure, use at least RAID 1 for that....
Take a look in one of those drug side-affects books (you know the ones that are almost as thick as they are wide). You'll find that one of those side-affects is *Swelling*!
I think that using a hard drive for backup would be a great idea, depending on the data, and size.
I have a set of data that would only need to be backed up for a month or two, and then could be easily removed, then selectively backing up to control files to tape.
This would not be good for our.rtf and.docs etc. Which would need to go directly to tape.
Just delete the old.bz2 archives as you need space. Could have statistics telling you how long the average file lasts in the archive. Anyone seen this done before?
When was the last time you got a good idea from a slashdot post?
"At least they weren't doing the Lynch-ian gun-pointing and saying "AAaahdd-CHAK" and having large explosions suddenly happen, though. The lameness of that alone exceeds this entire series."
I don't know, I liked it that way in the movie. Mind you, I saw it when I was 12 years old at the time.
When it first started, I thought it was going to show the movie. Since I only saw a few minutes of the movie, I was pretty disappointed when I found out that they didn't yell when they shot.;)
Anyway, I think I should see the movie, and read the books.:)
"Finally, we once recieved gift boxes - all of us - with some sort of appreciation gift inside. Turned out to be a styofoam oversized puzzle piece with the words "you're part of the solution" (or something) on a laminated side. One intrepid individual boxed them all up, sent them back to corporate with a letter asking for a working mailserver instead. Oh, and the pieces didn't fit. We thought that was foreboding."
...
I just couldn't keep from laughing! I'm glad just about everyone has left for home, or I'd have a lot of people staring at me.:)
"It's the same old marketese that Micro$oft is always saying: use our stuff and all your problems will go
away. You'll be able to retire at 15 to a deserted desert isle where bodacious babes will attend to your
every need and want. It's also amusing to hear someone from Sun grumbling about Linux fragmentation
while at the same time holding up their own fragment of Unix as the solution to the fragmentation problem!"
I have seen many do this in the linux camp. I think everyone is doing it, linux, be, bsd...
Aren't Snapshots and Phase Trees conceptually the same?
In Phase Tree, you take the changes made over a small priod of time, and make a "snapshot" out of that (called a phase, iirc) and begin to write it to the disk. While this is happening you have a new phase that all of the newer modifications are written to.
Doesn't this remind you of the description of snapshots?
Slashdot Mirror
Hello,
1 00 655627817897&w=2
1 00 655058402225&w=2
1 00 655322609664&w=2
It looks like this may be another BPB - Brown Paper Bag - 2.4 kernel release.
There is already a patch available though from Al Viro.
I believe Al can say it best. Here he is replying to war on LKML:
Damn this junk removal feature. I could've quoted if it didn't have trouble with ">"!! Grr.
http://marc.theaimsgroup.com/?l=linux-kernel&m=
References:
http://marc.theaimsgroup.com/?l=linux-kernel&m=
Patch:
http://marc.theaimsgroup.com/?l=linux-kernel&m=
Oh, boy... Where to begin...
First of all, the soonest the preempt patch would be merged, if at all would be in 2.5, and it is still in development. It only supports x86/32 and ARM so far...
"The thing about the preemptible kernel is that it is only for uniprocessor - SMP kernels aren't preemptible. "
This has been fixed over a month ago. UP and SMP kernels are now both preempt safe. So is a highmem (more than 2GB of ram) kernel.
"So unless you want the SMP case to be capable of tying up a processor for "too long" at a time, then you need to re-do each bit of code which is capable of long latencies anyway."
There are some small parts of the kernel that have been marked "not preempt safe". Fortunately, there aren't too many of them, and they're planned to be fixed in 2.5.
"The other thing which came up is that responsiveness of the system improved quite a bit recently with VM fixes (2.4.14 was the improved version, I think)."
Actually, the new VM was put in in 2.4.10pre11.
"It was a matter of the VM queueing up too much I/O (and the drivers trying to throttle it, instead of just throttling it all in the VM - or something like that)"
Actually that was one of the outward signs... The kernel favored the page cache over most anything else, and it grew particalarly large. That was since fixed, which allowed other problems to be noticed... The main problem was the swapping code. It would choose the wrong things to swap out, take too long to do so, and basically do bad things. Also there were some races in there that would cause kswapd to use 100% of the CPU (or just one cpu if you have SMP).
"The preemptible kernel won't solve that kind of problem"
Wow! The first thing that was 100% correct! Congratulations.
Don't forget that ext2/3 has indexed directories in development.
Right now, it's only for test partitions, but it is coming along nicely.
Ext2 has had changes from 2.0, 2.2 and 2.4 kernels, each time making it more efficient. Only recently has the actual format on disk changed, but that won't go in until 2.5 from what I can see.
Ext3 has three journal modes (available with the data=mode option on mount or in /etc/fstab):
Writeback: Don't care about ordering, just write in whatever order is most convenient. Possibly fastest for some loads... (This is basically the same level of journaling guarantees that you get from ReiserFS)
Ordered(default): Write out data before filesystem accounting information. (will protect all data modifications except for writing to the middle of the file - this is rare in most situations)
journal: Write all data to journal first (Guarantees all completed writes will be recoverable, including writes to the middle of files. This mode can be fastest with random writing because they all go directly to the journal - more effective write clustering...)
That said, ext3 does it's best to protect the data. Also, no JFS will protect you from hardware failure, use at least RAID 1 for that....
OMG!
:)
If there was a +10 funny, this would get it.
I have a dual celeron right here. It's 366 though :(
Take a look in one of those drug side-affects books (you know the ones that are almost as thick as they are wide). You'll find that one of those side-affects is *Swelling*!
And the more _YOU_ pay the more you'll have to pay the other side if you loose.
So if Joe Shmo sues Big_Co, he'll get paid what they paid their Lawers? Pay your Lawer and keep the diff? cool!
I don't understand why you don't just standardize on Rich Text Format...
.doc to .rtf
.xls. Anyone know of one?
In my company, I'm working on converting over from
Our Letter Art dept already uses rtf with DocuRight to work with our large laser printers.
This will also ease transition to a linux based word processor.
The only problem I've seen, is that there isn't an open format to replace
What do you want to bet that someone is going to market something called "Skynet"?
The concept for this is similar to the Crypto community. No crypto is trusted until it has been published, tested, and tested for years.
Here, published means basically the same as Open Source.
Actually, I have DSL at home, work, and a friend's house. All are faster than comcast@home in SoCal.
DSL: 15-30ms to first hop.
Cable: 60-150 to first hop.
NO, more like win 3.0, or MS DOS.
Really?!
I'll have to download it when it comes out.
Can you give a MacOS Version and hardware model?
I had a hell of a time with MacOS 8.5 on an Oldworld g3.
It's running great on linux as a vpn gateway...
:)
Mike
I'm a christian, and I thought it was a little funny, maybe not 5, but funny...
Are you trying to tell me that:
SunOS 5 runs SunOS 4 kernel modules?
That NT will use win(9x,3.1x)/os2 drivers?
That BSD will run kernel modules from five years ago on older versions?
That is what the cfleming was talking about *Kernel Modules*.
I think that using a hard drive for backup would be a great idea, depending on the data, and size.
.rtf and .docs etc. Which would need to go directly to tape.
.bz2 archives as you need space. Could have statistics telling you how long the average file lasts in the archive. Anyone seen this done before?
I have a set of data that would only need to be backed up for a month or two, and then could be easily removed, then selectively backing up to control files to tape.
This would not be good for our
Just delete the old
When was the last time you got a good idea from a slashdot post?
...
;)
:)
"At least they weren't doing the Lynch-ian gun-pointing and saying "AAaahdd-CHAK" and having large explosions suddenly happen, though. The lameness of that alone exceeds this entire series."
I don't know, I liked it that way in the movie. Mind you, I saw it when I was 12 years old at the time.
When it first started, I thought it was going to show the movie. Since I only saw a few minutes of the movie, I was pretty disappointed when I found out that they didn't yell when they shot.
Anyway, I think I should see the movie, and read the books.
....
:)
"Finally, we once recieved gift boxes - all of us - with some sort of appreciation gift inside. Turned out to be a styofoam oversized puzzle piece with the words "you're part of the solution" (or something) on a laminated side. One intrepid individual boxed them all up, sent them back to corporate with a letter asking for a working mailserver instead. Oh, and the pieces didn't fit. We thought that was foreboding."
...
I just couldn't keep from laughing! I'm glad just about everyone has left for home, or I'd have a lot of people staring at me.
Someone mod this up as funny!!!
"IE on mac os is pretty fast too though so it might not be entirely a IE Windows thing. but what do i know?"
Actually, it's true. But that's because MS does the same damn thing with MacOS. They add several extentions to your system to get IE to work.
You can take the installed folder of netscape and copy it to another mac and run it on the other mac WITHOUT INSTALLING IT! Try that with IE. ha!
I have seen many do this in the linux camp. I think everyone is doing it, linux, be, bsd...
You are quite right.
Although I don't program in a compiled language yet, I've seen myself fall into that trap.
I ended up trying to solve a problem I didn't even need to have in there.
From that point on, I took time to comment what I needed to do before I wrote the first line of code.
I hope this helps someone.
I can imagine it now!
"The MS Internet will be based on NetBEUI enabling everyone to leverage the power of Windows TM."
Aren't Snapshots and Phase Trees conceptually the same?
In Phase Tree, you take the changes made over a small priod of time, and make a "snapshot" out of that (called a phase, iirc) and begin to write it to the disk. While this is happening you have a new phase that all of the newer modifications are written to.
Doesn't this remind you of the description of snapshots?