Swapping out data before you need the free RAM would be a great idea if you kept two copies. One copy on disk and one copy in RAM. In fact it would be fine if the system swapped out 90% or more of the process memory this way. There will now be three different cases to think about.
The process needs to read the page - no problem, one copy is in RAM just read it, and keep both copies.
The process needs to write the page - no problem, you can modify the copy in RAM and discard the copy on disk. Notice that discarding the copy on disk doesn't require any disk access, as the list of swap allocations will typically be in RAM (it is much smaller than the swapspace).
You actually need memory - no problem, discard a not recently used RAM page, you still have a copy on disk.
The only problem is, that you need to make the page readonly, so you can trap the write and discard the on disk copy. In other words don't do this for pages that are frequently changed. But usually you don't have many pages that are frequently changed, and you certainly don't want to swap out those you have. And should you occationally happen to swap out one, it is not really a major problem. It will cost you a pagefault, but no disk I/O. And a pagefault is compared to a disk I/O. A system that behaves like I have described here would use a lot more space than Linux typically does, but still it should be faster. I wonder why this isn't done more often, it is not like the idea hasn't been known for years.
Another problem that many have noticed, and that isn't easy to deal with, is heavy diskaccess causing the cache to grow and stuff getting swapped out. Yes even some Linux versions suffer from this problem. A Red Hat 9 system I had running for months was really slow in the morning, because all the programs had been swapped out while cron jobs where running during the night. But you never know when it is a good idea to swap the stuff out and when it is not. When the disk access is going on, the process page might not have been used for hours. But still you might want it to be kept in RAM. File pages that have been accessed just once shouldn't be kept in cache for long time. But of course you shouldn't remove them unless the memory was needed for something else. Removing the pages too early is also bad, because you wouldn't notice, that this was really a page that was going to be accessed frequently. Some people are fanatic, and don't want process pages to ever get swapped out to make room for cache. That isn't a good idea either. You can really have process pages that may not be needed even once, do you want such a page to be kept in ram for months just in case? And notice how disabling swap is not going to solve the problem. You still have to think about memory mapped files, that in many ways must be treated like anonymous mappings.
So Ballmer thinks Linux lacks inovation in areas like compatibility? It is an interesting point, since Microsoft have been very inovative when it comes to incompatibility. When has inovation ever been good to compatibility?
Driver CDs, sealed in an envelope with a sticker that read "By breaking this seal, you agree to be bound by the enclosed license agreement."
Quite similar to the one you'd find on some boxes with OS/2. Except the one on the OS/2 box said, you should read the license inside the box, before opening the box.
if you're anything like me your machine is formatted at least a
month.
In that case I'm not anything like you. I installed Red Hat Linux 9 in april 2003 and kept using it until I upgraded to Fedora Core 1 in april 2004. You asked for the first ten programs I installed, well I haven't installed ten programs yet, because Fedora Core 1 has almost everything I need. What I have installed is:
netscape-communicator
mplayer
ogle
xine
xmms-mp3
And most of them I downloaded from freshrpms.net. And the reason I can tell you exactly what I have installed is, that I can just grep ^rpm/root/.bash_history.
A coworker pointed out that task switching can't be that
speedy.
Of course for this to be useful to general purpose machines, switching must be possible. But if you only use this feature for programs that can benefit significantly, and do lazy switching of the state, it might not be that bad after all. Do it kind of like Linux handles FPU instructions.
Taking more time to encrypt/decrypt isn't a problem (does anyone here notice the differance between 2.5ms and 5ms?) but reducing the crack time by the same proportions means that codes that were built to last years might only last months, or even mere weeks, which is a real problem.
If we assume this can speed up AES by a factor of two (not unrealistic) or even a factor of ten (I dobut it), it still doesn't mean that what took years to crack before can now be cracked in weeks. If it took two years before, and you can do it ten times faster, it still takes more than two months. But there is another and much more important point. If thise can speed up encryption and decryption, people can switch to a stronger cipher. Say people switch to this new chip to get a factor of two speedup, but at the same time they switch from AES-128 to AES-256 and lose 25% of the performance. Overall this means they will have a speedup of 50% and at the same time a stronger cipher that will take millions of times more work to break. So even if the crackers got a thousand times faster processor, they would still need a thousand times more time to break it.
Oops, I read it it as my damn windows, desktop, bookmarks, window sizes but I see now, that it really was my damn windows desktop, bookmarks, window sizes. Punctation matters. Now if they had used a better name for their damn product, I wouldn't have made that mistake.
i use the same computer cause i don't feel like setting up my damn windows desktop, bookmarks, window sizes.
Mount your home directory with NFS, and it will be the same no matter which computer you log onto. And when the NFS server needs to be replaced, just copy the whole directory to a new server. User settings and machine settings where meant to be kept seperate. Typically home directories live longer than the machines. My home directory on the department network is now on the fifth NFS server since I got my account (unless I forgot one: scandium, amigo, harald, atlantis, gorm). On my home network,/home is now on the third NFS server since I started using Linux (eddie, marvin, hactar). If you cannot replace a computer without affecting users, or replace a user without affecting the computers, something is wrong with your setup.
instead of backup, hold data in multiple places at once
Even better, instead of backup just crawl the pages again in the event of a lost disk. Of course some data needs to be in multiple places for performance reasons, but not all data are accessed frequently. How often do you think they will need the page with the lowest rank? (OK, I know there will probably be a lot with exactly the same rank, but you get the idea).
load software via NFS at node bootup
There are better protocols for this than NFS. But when you build a cluster this size, you surely want boxes, that can netboot of of the box. Actually that means you will need to use DHCP and TFTP. Security of the DHCP and TFTP servers is going to be very critical.
use nodes just to store data; keep software in RAM for speed
I wouldn't worry about the speed. Linux is going to do fine. But since they probably netboot and download kernel and a ramdisk from a server, it is of course going to be kept in ram. Now I wonder, does it all run of an initial ramdisk?
return to an IBM-dominated world is _not_ what you want.
Right now IBM is pushing Linux forward. Of course they do that because they want to make money on their hardware. But if they suceed (and I think they will, it can suddenly go very fast), how could they possibly dominate the world? If another company could make some good hardware, they could run Linux as well. And with open standards, and two hardware platforms running the same open source software, it will really be hard to monopolize the market.
What if it's 10 years later and 3ware IDE cards are ancient history?
That is my primary reason for using software raid. The other reasons being that a raid card is much more expensive than an ordinary IDE controller, and I have read more than once, that it is really still software raid. My setup with three 120GB disks and identical partitioning of all three disks goes like this. One 31MB/boot partition, one 31MB FAT partition (just in case), one 627MB partition for/, one 2GB partition for swap, and the rest for one large filesystem that I can bindmount on/usr,/home, and other places I need it. The/boot filesystem is not raid, but I have a copy of it on each disk. The root filesystem is raid-1 (on three disks, that should be very safe), and the rest is raid-5, which give me one 289GB filesystem.
its more accurate to say that BGP is "the" exterior routing protocol
Yes, that was also my understanding of how it works. Doesn't that mean, that every time you communicate with any computer outside your own AS, you will indirectly be relying on BGP?
Well - if BGP gets seriously hit by this, this might wreak *quite* massive havoc on the net...
How bad could it go? Would it stop all communication outside your own AS? Would the routers be able to restore connections when the attack stops? Or would manual action be required to make it work again?
The question is, whether any similar protocols (e.g. OSPF) are also vulnerable...
But if I understood it correctly, you can only OSPF inside your own AS. Doesn't that mean, that OSPF could never be considered a replacement for BGP?
His Monitor starts going nuts a few seconds before the cell phone
rings.
The exact same thing happened with my old cell phone and my old monitor, but only if the cell phone was lying on the table where it would be about 20 cm from the monitor. Obviously this means, the phone must know the call is comming, then why doesn't it start ringing until a few seconds later?
...that doesn't work with Linux, can I get a story on slashdot too? I mean seriously, he tried nine different versions of Linux with the same sound card, and it never occured to him, that it could be the sound card that was responsible. In my case it is AFAIK an original Sound Blaster card, which is in the computer. And when playing sound it starts fine, but suddenly the computer locks up completely. But I don't whine, I start debuging the problem myself. This particular scenario is hard to debug, so I haven't fixed it yet. Longer time ago, when I had problems with an ALS120 sound card, it didn't take me much time to modify the kernel so it would work. Remember, the Linux developers don't have access to every piece of hardware in existence. But all hardware developers have access to Linux, they just don't care about it. So if you are the only person who have that hardware and care about the problem, you have to help making it work.
some software that "zooms in" on a part of the screen
How much zoom do you need. I know some people with vision problems have found the XFree86 key combinations for changing resolution a big help. Depending on the distribution you might need to change XF86Config to add some lower resolutions. Then use Ctrl+Alt+Plus (on the numerical keypad) to switch between resolutions. Other than that I think the best solution is to use larger fonts. Really larger fonts are much more readable than just zooming on the fonts rendered for the small size.
I agree, people rarely write open source software unless they need it themself. And it is true that when you have written a program you don't need documentation to use that program. However the author of the program is the one, who knows best how it works. Of course that doesn't always mean the author of the program is also the best person to write the documentation, but at least whoever wants to write documentation should communicate with the author of the program.
I have written software with way too litle documentation, but I don't want to spend time writing documentation if nobody would ever use that documentation. I know at least one person is going to use the software, so writing the code isn't a waste of time. Should anybody actually ask me how to use a program I have written, I would help them as much as I can. If that kept happening, I would keep answering. At first I would surely be happy about people finding my program useful, later I might feel a bit bad about the time I spend answering peoples questions. But at that time I would know it was because I hadn't written documentation. So rather than blaming people about the questions, I would blame myself about the documentation I hadn't written. And then I might get started writing some documentation.
Probably not all people writing software are like me. But if you find an open source program without documentation, and you need a bit of help. Try asking the author, he might actually be a nice person, who are willing to help you.
Slashdot Mirror
- The process needs to read the page - no problem, one copy is in RAM just read it, and keep both copies.
- The process needs to write the page - no problem, you can modify the copy in RAM and discard the copy on disk. Notice that discarding the copy on disk doesn't require any disk access, as the list of swap allocations will typically be in RAM (it is much smaller than the swapspace).
- You actually need memory - no problem, discard a not recently used RAM page, you still have a copy on disk.
The only problem is, that you need to make the page readonly, so you can trap the write and discard the on disk copy. In other words don't do this for pages that are frequently changed. But usually you don't have many pages that are frequently changed, and you certainly don't want to swap out those you have. And should you occationally happen to swap out one, it is not really a major problem. It will cost you a pagefault, but no disk I/O. And a pagefault is compared to a disk I/O. A system that behaves like I have described here would use a lot more space than Linux typically does, but still it should be faster. I wonder why this isn't done more often, it is not like the idea hasn't been known for years.Another problem that many have noticed, and that isn't easy to deal with, is heavy diskaccess causing the cache to grow and stuff getting swapped out. Yes even some Linux versions suffer from this problem. A Red Hat 9 system I had running for months was really slow in the morning, because all the programs had been swapped out while cron jobs where running during the night. But you never know when it is a good idea to swap the stuff out and when it is not. When the disk access is going on, the process page might not have been used for hours. But still you might want it to be kept in RAM. File pages that have been accessed just once shouldn't be kept in cache for long time. But of course you shouldn't remove them unless the memory was needed for something else. Removing the pages too early is also bad, because you wouldn't notice, that this was really a page that was going to be accessed frequently. Some people are fanatic, and don't want process pages to ever get swapped out to make room for cache. That isn't a good idea either. You can really have process pages that may not be needed even once, do you want such a page to be kept in ram for months just in case? And notice how disabling swap is not going to solve the problem. You still have to think about memory mapped files, that in many ways must be treated like anonymous mappings.
So Ballmer thinks Linux lacks inovation in areas like compatibility? It is an interesting point, since Microsoft have been very inovative when it comes to incompatibility. When has inovation ever been good to compatibility?
Driver CDs, sealed in an envelope with a sticker that read "By breaking this seal, you agree to be bound by the enclosed license agreement."
Quite similar to the one you'd find on some boxes with OS/2. Except the one on the OS/2 box said, you should read the license inside the box, before opening the box.
In that case I'm not anything like you. I installed Red Hat Linux 9 in april 2003 and kept using it until I upgraded to Fedora Core 1 in april 2004. You asked for the first ten programs I installed, well I haven't installed ten programs yet, because Fedora Core 1 has almost everything I need. What I have installed is:
- netscape-communicator
- mplayer
- ogle
- xine
- xmms-mp3
And most of them I downloaded from freshrpms.net. And the reason I can tell you exactly what I have installed is, that I can just grep ^rpmA coworker pointed out that task switching can't be that speedy.
Of course for this to be useful to general purpose machines, switching must be possible. But if you only use this feature for programs that can benefit significantly, and do lazy switching of the state, it might not be that bad after all. Do it kind of like Linux handles FPU instructions.
Taking more time to encrypt/decrypt isn't a problem (does anyone here notice the differance between 2.5ms and 5ms?) but reducing the crack time by the same proportions means that codes that were built to last years might only last months, or even mere weeks, which is a real problem.
If we assume this can speed up AES by a factor of two (not unrealistic) or even a factor of ten (I dobut it), it still doesn't mean that what took years to crack before can now be cracked in weeks. If it took two years before, and you can do it ten times faster, it still takes more than two months. But there is another and much more important point. If thise can speed up encryption and decryption, people can switch to a stronger cipher. Say people switch to this new chip to get a factor of two speedup, but at the same time they switch from AES-128 to AES-256 and lose 25% of the performance. Overall this means they will have a speedup of 50% and at the same time a stronger cipher that will take millions of times more work to break. So even if the crackers got a thousand times faster processor, they would still need a thousand times more time to break it.
he said "my damn windows desktop".
Oops, I read it it as my damn windows, desktop, bookmarks, window sizes but I see now, that it really was my damn windows desktop, bookmarks, window sizes. Punctation matters. Now if they had used a better name for their damn product, I wouldn't have made that mistake.
i use the same computer cause i don't feel like setting up my damn windows desktop, bookmarks, window sizes.
/home is now on the third NFS server since I started using Linux (eddie, marvin, hactar). If you cannot replace a computer without affecting users, or replace a user without affecting the computers, something is wrong with your setup.
Mount your home directory with NFS, and it will be the same no matter which computer you log onto. And when the NFS server needs to be replaced, just copy the whole directory to a new server. User settings and machine settings where meant to be kept seperate. Typically home directories live longer than the machines. My home directory on the department network is now on the fifth NFS server since I got my account (unless I forgot one: scandium, amigo, harald, atlantis, gorm). On my home network,
instead of backup, hold data in multiple places at once
Even better, instead of backup just crawl the pages again in the event of a lost disk. Of course some data needs to be in multiple places for performance reasons, but not all data are accessed frequently. How often do you think they will need the page with the lowest rank? (OK, I know there will probably be a lot with exactly the same rank, but you get the idea).
load software via NFS at node bootup
There are better protocols for this than NFS. But when you build a cluster this size, you surely want boxes, that can netboot of of the box. Actually that means you will need to use DHCP and TFTP. Security of the DHCP and TFTP servers is going to be very critical.
use nodes just to store data; keep software in RAM for speed
I wouldn't worry about the speed. Linux is going to do fine. But since they probably netboot and download kernel and a ramdisk from a server, it is of course going to be kept in ram. Now I wonder, does it all run of an initial ramdisk?
return to an IBM-dominated world is _not_ what you want.
Right now IBM is pushing Linux forward. Of course they do that because they want to make money on their hardware. But if they suceed (and I think they will, it can suddenly go very fast), how could they possibly dominate the world? If another company could make some good hardware, they could run Linux as well. And with open standards, and two hardware platforms running the same open source software, it will really be hard to monopolize the market.
I've never heard of any of these "well known" groups.
Come on, who have not heard about Echelon? But I must admit I'm a bit surprised to learn, they also do software piracy.
What if it's 10 years later and 3ware IDE cards are ancient history?
/boot partition, one 31MB FAT partition (just in case), one 627MB partition for /, one 2GB partition for swap, and the rest for one large filesystem that I can bindmount on /usr, /home, and other places I need it. The /boot filesystem is not raid, but I have a copy of it on each disk. The root filesystem is raid-1 (on three disks, that should be very safe), and the rest is raid-5, which give me one 289GB filesystem.
That is my primary reason for using software raid. The other reasons being that a raid card is much more expensive than an ordinary IDE controller, and I have read more than once, that it is really still software raid. My setup with three 120GB disks and identical partitioning of all three disks goes like this. One 31MB
I'm sure he's been proven wrong a few times.
Once (according to the FAQ).
its more accurate to say that BGP is "the" exterior routing protocol
Yes, that was also my understanding of how it works. Doesn't that mean, that every time you communicate with any computer outside your own AS, you will indirectly be relying on BGP?
Well - if BGP gets seriously hit by this, this might wreak *quite* massive havoc on the net...
How bad could it go? Would it stop all communication outside your own AS? Would the routers be able to restore connections when the attack stops? Or would manual action be required to make it work again?
The question is, whether any similar protocols (e.g. OSPF) are also vulnerable...
But if I understood it correctly, you can only OSPF inside your own AS. Doesn't that mean, that OSPF could never be considered a replacement for BGP?
Just unplug your PC from the internet
How would that keep you safe from DoS attacks?
His Monitor starts going nuts a few seconds before the cell phone rings.
The exact same thing happened with my old cell phone and my old monitor, but only if the cell phone was lying on the table where it would be about 20 cm from the monitor. Obviously this means, the phone must know the call is comming, then why doesn't it start ringing until a few seconds later?
My experience is, that 15 random lowercase letters will work on most systems.
Well, if she was nice, it might take less than that before I told her my root password is "FSPWBK#hiL!JxKvS" and my IP address is 172.16.42.42.
...that doesn't work with Linux, can I get a story on slashdot too? I mean seriously, he tried nine different versions of Linux with the same sound card, and it never occured to him, that it could be the sound card that was responsible. In my case it is AFAIK an original Sound Blaster card, which is in the computer. And when playing sound it starts fine, but suddenly the computer locks up completely. But I don't whine, I start debuging the problem myself. This particular scenario is hard to debug, so I haven't fixed it yet. Longer time ago, when I had problems with an ALS120 sound card, it didn't take me much time to modify the kernel so it would work. Remember, the Linux developers don't have access to every piece of hardware in existence. But all hardware developers have access to Linux, they just don't care about it. So if you are the only person who have that hardware and care about the problem, you have to help making it work.
some software that "zooms in" on a part of the screen
How much zoom do you need. I know some people with vision problems have found the XFree86 key combinations for changing resolution a big help. Depending on the distribution you might need to change XF86Config to add some lower resolutions. Then use Ctrl+Alt+Plus (on the numerical keypad) to switch between resolutions. Other than that I think the best solution is to use larger fonts. Really larger fonts are much more readable than just zooming on the fonts rendered for the small size.
I agree, people rarely write open source software unless they need it themself. And it is true that when you have written a program you don't need documentation to use that program. However the author of the program is the one, who knows best how it works. Of course that doesn't always mean the author of the program is also the best person to write the documentation, but at least whoever wants to write documentation should communicate with the author of the program.
I have written software with way too litle documentation, but I don't want to spend time writing documentation if nobody would ever use that documentation. I know at least one person is going to use the software, so writing the code isn't a waste of time. Should anybody actually ask me how to use a program I have written, I would help them as much as I can. If that kept happening, I would keep answering. At first I would surely be happy about people finding my program useful, later I might feel a bit bad about the time I spend answering peoples questions. But at that time I would know it was because I hadn't written documentation. So rather than blaming people about the questions, I would blame myself about the documentation I hadn't written. And then I might get started writing some documentation.
Probably not all people writing software are like me. But if you find an open source program without documentation, and you need a bit of help. Try asking the author, he might actually be a nice person, who are willing to help you.
The 'open' one downlodable from whatever Free Software web or FTP site that is linked.
On that FTP server you will surely find sourcecode. You might also find a compiled version. Which one would you use.
any of the dozens of other binary diff/patch utilities?
But what binaryfiles are you going to compare with this program?