In the same way the immune system doesn't produce antibodies until it has something to attack, the Linux distributions won't produce virus killers until we actually find a virus in the wild. If you know ahead of time how to identify and kill a virus then you've obviously got a better crystal ball than anyone else and should go and make a killing on the lottery. Either that or you're the chap who's just written one.
I haven't noticed a lot of "defragment" programs for Linux either, perhaps because the ext2 filesystem is inherantly less likely to become fragmented.
If necessary, market forces will bring required applications into being if there's a market for it. Hey, it even brings products into being even if they don't work but you can get suckers to buy it anyway.
The surfaces of the glass disks are not actually flat, but faintly speckled to prevent striction (where if the heads were to touch they would be both be so flat that they would actually stick together)
These bumps are very small, just enough to pop the heads back into a floating mode if they did happen to touch the disk surface.
The trouble with that plan is that the whole reason to bang on these pre-release kernels is so that faults can be caught and squashed when the real one is released.
If you tried to log a fault with a user-mode kernel you'd not get very far on the lkml. Worse, you're going to cause confusion and dilute the testing process.
The only benefit would be if you were an absolute testing god, and could weed implementation bugs out from the inherant kernel bugs, do some fancy debugging from the working system to the broken one and send the lkml a proper fault analysis and patched code. THEN I'd be impressed.
I had to do this to get that extra point (You couldn't accuse me of cheating)
Here is the result of your Slashdot Purity Test. You answered "yes" to 59 of 200 questions, making you 70.5% slashdot pure (29.5% slashdot corrupt). According to the scoring guide, your slashdot experience level is: JonKatz Wannabe
Is anyone going to collate these, perhaps this shouldn't have been a quickie but a full entry - I reckon it'll bust the 500 replies mark.
This looks like its done at installation, where you install a package and all it does is fill a tree full of \\host\share\whatever\path links apart from the read-write files. If you had a pervasive NFS mount you could do the same in UNIX, but NT's ability to make arbitary mounts just by specifying \\ at the the beginning of the path is adding the value here. Lets hope the server's reliable then, and that the system administrator doesn't move things around.
Remember that the tests were carried out by Network World Fusion... The following URL contains tuning details: http://www.nwfusion.com/reviews/2000/0124revtuni ng.html
I'm not a tuning expert, but it looks like they made some changes which wouldn't improve performance. (I support NetworkWorldFusion fully in carrying out the tests and making the tuning details obvious.) 1. If they were using RedHat 6.1 why did they have to apply RAID patches, especially raid-2.2.14-A2 when RH ships with 2.2.12. 2. Why increase EXT2_MAX_GROUP_LOADED and fiddle with the update parameters. 3. Samba configuration - turning off "read raw" - the smb.conf man pages says this provides a major performance benefit. (They also forgot about swat/linuxconf as samba gui's)
Its a shame they didn't carry out the performance modifications that came out of the Mindcraft 1 "study". One in my archives is: -----from a message Jeremy Allison posted---- b). What will give the most performance benefit is to tell Linux to use most of main memory for file cache and to keep it in memory for a long time. To do this add the line :
to your rc.local. This tells Linux to use 80% of memory for file system cache and to keep it around for as long as possible.
smb.conf tweaks from Juan Carlos Castro y Castro socket options = TCP_NODELAY SO_SNDBUF=8192 SO_RCVBUF=8192 read prediction = yes debug level = 0;yeah, explicit - the default seems not to be 0
I've put in RedHat linux servers for both Thumbswood and Blackthorn schools in Hertfordshire. Windows on the desktop, but the apps are loaded from a samba drive and the web is browsable through squid. Support costs - 1 day to set up, 1 hours support since the system went online 09/99. Can't say the same about supporting the W95 boxes though, ugh.
AFAIK the main problem with non-monolithic kernel designs is that the more you seperate the components, the worse the communication and syncronisation problems become.
It compares with SMP multiprocessing, you never
achieve 100% of each added processor and beyond a certain limit (the largest I've seen is 64-processor SUN e1000) they trip each other up to the point where you get lower performance by adding processors.
I've been tracking MS IE development on Solaris - there isn't any. When a new version of IE comes out they release a solaris version, but its so buggy you've only got a 50/50 chance of even opening the front page - it always dies within 2 minutes of browsing for me (Solaris 2.6, recent patches). Even with these problems I keep downloading from the MS site and guess what... the file I download doesn't change - between April and November 1999 the same unusable buggy IE5 was available.
This indicates that either no-one uses IE5 on Solaris, or haven't logged any faults or that MS don't care. Probably a bit of each.
The first thing about a hardware raid controller is that it hides failures from the operating system. With software RAID you have to manually carry out all sorts of tasks, and I'm sure we've all heard of the engineer who mirrored the new blank disk on top of the one remaining data disk of a mirror. Units such as SUN A1000 and Baydel connect via SCSI and you just watch for an orange light, even the part-time cleaner could pull out the correct disk and replace it and have the system back and running without the OS noticing. Storageworks and Clariion(EMC) do the same but over Fiber Channel. SCSI units tend to top out at 40Mb/s, Fiber Channel theoretically top out at 200Mb/s (they have two 100Mb/s loops) but since I only had a max of 30x18Gb disks to play with the disks were the bottleneck. Monster multi-scsi machines like EMC/IBM's can achieve whatever bandwidth you want by multiplexing SCSI connections. We've evaluated software RAID, Hardware RAID over SCSI, Hardware RAID over Fiber channel from EMC, IBM, SUN, Compaq(storageworks) and in our opinion a good smart raid controller with two data channels and load balancing software is impossible to beat. For Speed, stripe(0) mirrors together(1), in RAID 0+1, this allows reads at double speed because each mirrored disk can handle a request seperately, and slightly sped-up writes because you can write to the RAID controller's NV cache and carry on doing your work whilst that takes care of putting the data to media. This of course has only a 50% data efficiency. Using Raid 3 or 5 you lose one disk in a rank for parity, raid 6 (used by Network Appliances) use two disks for parity but have wider ranks of disks. This often means that sequential reads are fast, because a request for data wakes up all the disks in the rank, but therefore the whole rank can only handle one request at a time. Writes are slower because you have to read a stripe of data, calculate parity and write the whole stripe back again. RAID5 is really good for data which doesn't have to be the absolute fastest. Whilst we were doing performance tests, we measured a linear increase in speed up to 20 disks (in transactions/second), and there is a definite art in making sure that you spread the load over all the disks available so that a single disk doesn't get thrashed to death. In conclusion? well, that depends on your OS. For me, for a PC-based system I would choose a hardware RAID system with SCSI connection which let me choose the LUN sizes. 5 disks in a RAID5 configuration will only waste 1 disk in capacity. If you're finding your mail spool is being thrashed then I would build a 10 disk 0+1 raid and stripe the mail area across them, using the rest of the area for home areas or web areas or something else which has large storage requirements but doesn't get hit hard. Oops, this assumes that this REALLY is your problem, a lot of disk problems go away by adding more memory to the machine... I assume you have measured this by tracking the outstanding I/O queue.
I was told that a doctorate was awarded for expanding the field of human knowledge, this is the reason you have to choose a unique field when you do one at college. Linus has, (possibly inadvertantly at the time) changed how we look at computer software. I also think he has expanded the field of achievements in computer science. 7 years ago when I first saw the paradigm he was putting forward I knew this was something new which was only then starting to become reality due to the internet. I believe he may deserve a doctorate for three reasons... 1) He's a very good programmer and a smart cookie. 2) He's popularised a new way of producing software. 3) He's changed the course of history. Now he'll have to field silly phone calls from panicing babysitters who pick up the phone book and look at the first Dr. they see.
I would be more supportive of Microsofts freedom to innovate if microsoft was more supportive of my freedom to stay with a stable system. Incompatible file formats have forced me through various upgrades to remain able to share files with my colleagues for no benefit to me (I need no more from a word processor than I had with Microsoft Word 2) I also disagree with innovation when it gets in the way of shipping a stable, reliable, efficient product. Please remember that there are two kinds of innovation, the good and the bad.
Slashdot Mirror
In the same way the immune system doesn't produce antibodies until it has something to attack, the Linux distributions won't produce virus killers until we actually find a virus in the wild. If you know ahead of time how to identify and kill a virus then you've obviously got a better crystal ball than anyone else and should go and make a killing on the lottery. Either that or you're the chap who's just written one.
I haven't noticed a lot of "defragment" programs for Linux either, perhaps because the ext2 filesystem is inherantly less likely to become fragmented.
If necessary, market forces will bring required applications into being if there's a market for it. Hey, it even brings products into being even if they don't work but you can get suckers to buy it anyway.
--
The surfaces of the glass disks are not actually flat, but faintly speckled to prevent striction (where if the heads were to touch they would be both be so flat that they would actually stick together)
These bumps are very small, just enough to pop the heads back into a floating mode if they did happen to touch the disk surface.
--
If you're running a recent linux kernel or other production-quality OS then all the bios does is restrict where you can boot from.
Once you've got your OS up and running you can happily partition and use the rest of your disk.
I've got some big drives on a dell which can only see 8Gb of a disk, all it has done is cramp (!!) the size of my root filesystem.
--
The best thing about the 40Gb drive are the moves to make it quieter.
The whine in my ear from my wife would reduce if the whine from the computers reduced.
--
The trouble with that plan is that the whole reason to bang on these pre-release kernels is so that faults can be caught and squashed when the real one is released.
If you tried to log a fault with a user-mode kernel you'd not get very far on the lkml. Worse, you're going to cause confusion and dilute the testing process.
The only benefit would be if you were an absolute testing god, and could weed implementation bugs out from the inherant kernel bugs, do some fancy debugging from the working system to the broken one and send the lkml a proper fault analysis and patched code. THEN I'd be impressed.
I had to do this to get that extra point
(You couldn't accuse me of cheating)
Here is the result of your Slashdot Purity Test.
You answered "yes" to 59 of 200 questions, making you 70.5% slashdot pure (29.5% slashdot corrupt).
According to the scoring guide, your slashdot experience level is: JonKatz Wannabe
Is anyone going to collate these, perhaps this
shouldn't have been a quickie but a full entry -
I reckon it'll bust the 500 replies mark.
This looks like its done at installation, where you install a package and all it does is fill a tree full of \\host\share\whatever\path links apart from the read-write files.
If you had a pervasive NFS mount you could do the same in UNIX, but NT's ability to make arbitary mounts just by specifying \\ at the the beginning of the path is adding the value here.
Lets hope the server's reliable then, and that the system administrator doesn't move things around.
Remember that the tests were carried out by Network World Fusion...i ng.html
;yeah, explicit - the default seems not to be 0
The following URL contains tuning details:
http://www.nwfusion.com/reviews/2000/0124revtun
I'm not a tuning expert, but it looks like they made some changes which wouldn't improve performance.
(I support NetworkWorldFusion fully in carrying out the tests and making the tuning details obvious.)
1. If they were using RedHat 6.1 why did they have to apply RAID patches, especially raid-2.2.14-A2 when RH ships with 2.2.12.
2. Why increase EXT2_MAX_GROUP_LOADED and fiddle with the update parameters.
3. Samba configuration - turning off "read raw" - the smb.conf man pages says this provides a major performance benefit.
(They also forgot about swat/linuxconf as samba gui's)
Its a shame they didn't carry out the performance modifications that came out of the Mindcraft 1 "study".
One in my archives is:
-----from a message Jeremy Allison posted----
b). What will give the most performance benefit is to tell Linux to use most of main memory for file cache and to keep it in memory for a long time. To do this add the line :
echo "80 500 64 64 80 6000 6000 1884 2" >/proc/sys/vm/bdflush
to your rc.local. This tells Linux to use 80% of memory for file system cache and to keep it around for as long as possible.
smb.conf tweaks from Juan Carlos Castro y Castro
socket options = TCP_NODELAY SO_SNDBUF=8192 SO_RCVBUF=8192
read prediction = yes
debug level = 0
I've put in RedHat linux servers for both
Thumbswood and Blackthorn schools in Hertfordshire.
Windows on the desktop, but the apps are loaded
from a samba drive and the web is browsable through
squid.
Support costs - 1 day to set up, 1 hours support
since the system went online 09/99. Can't say
the same about supporting the W95 boxes though, ugh.
Tim
Yeah, or alternatively they saw it coming
and couldn't do anything and died.
I'd rather keep sending the unmanned probes.
AFAIK the main problem with non-monolithic kernel designs is that the more you seperate the components, the worse the communication and syncronisation problems become.
It compares with SMP multiprocessing, you never
achieve 100% of each added processor and beyond a certain limit (the largest I've seen is 64-processor SUN e1000) they trip each other up to the point where you get lower performance by adding processors.
I've been tracking MS IE development on
Solaris - there isn't any.
When a new version of IE comes out they
release a solaris version, but its so buggy
you've only got a 50/50 chance of even opening
the front page - it always dies within 2 minutes
of browsing for me (Solaris 2.6, recent patches).
Even with these problems I keep downloading from
the MS site and guess what... the file I download
doesn't change - between April and November 1999
the same unusable buggy IE5 was available.
This indicates that either no-one uses IE5 on
Solaris, or haven't logged any faults or that
MS don't care. Probably a bit of each.
The first thing about a hardware raid controller is that it hides failures from the operating system. With software RAID you have to manually carry out all sorts of tasks, and I'm sure we've all heard of the engineer who mirrored the new blank disk on top of the one remaining data disk of a mirror.
Units such as SUN A1000 and Baydel connect via SCSI and you just watch for an orange light, even the part-time cleaner could pull out the correct disk and replace it and have the system back and running without the OS noticing. Storageworks and Clariion(EMC) do the same but over Fiber Channel. SCSI units tend to top out at 40Mb/s, Fiber Channel theoretically top out at 200Mb/s (they have two 100Mb/s loops) but since I only had a max of 30x18Gb disks to play with the disks were the bottleneck. Monster multi-scsi machines like EMC/IBM's can achieve whatever bandwidth you want by multiplexing SCSI connections.
We've evaluated software RAID, Hardware RAID over SCSI, Hardware RAID over Fiber channel from EMC, IBM, SUN, Compaq(storageworks) and in our opinion a good smart raid controller with two data channels and load balancing software is impossible to beat.
For Speed, stripe(0) mirrors together(1), in RAID 0+1, this allows reads at double speed because each mirrored disk can handle a request seperately, and slightly sped-up writes because you can write to the RAID controller's NV cache and carry on doing your work whilst that takes care of putting the data to media.
This of course has only a 50% data efficiency.
Using Raid 3 or 5 you lose one disk in a rank for parity, raid 6 (used by Network Appliances) use two disks for parity but have wider ranks of disks. This often means that sequential reads are fast, because a request for data wakes up all the disks in the rank, but therefore the whole rank can only handle one request at a time. Writes are slower because you have to read a stripe of data, calculate parity and write the whole stripe back again.
RAID5 is really good for data which doesn't have to be the absolute fastest.
Whilst we were doing performance tests, we measured a linear increase in speed up to 20 disks (in transactions/second), and there is a definite art in making sure that you spread the load over all the disks available so that a single disk doesn't get thrashed to death.
In conclusion? well, that depends on your OS.
For me, for a PC-based system I would choose a hardware RAID system with SCSI connection which let me choose the LUN sizes. 5 disks in a RAID5 configuration will only waste 1 disk in capacity. If you're finding your mail spool is being thrashed then I would build a 10 disk 0+1 raid and stripe the mail area across them, using the rest of the area for home areas or web areas or something else which has large storage requirements but doesn't get hit hard.
Oops, this assumes that this REALLY is your problem, a lot of disk problems go away by adding more memory to the machine... I assume you have measured this by tracking the outstanding I/O queue.
I was told that a doctorate was awarded for
expanding the field of human knowledge, this is
the reason you have to choose a unique field
when you do one at college.
Linus has, (possibly inadvertantly at the time)
changed how we look at computer software.
I also think he has expanded the field of
achievements in computer science. 7 years ago
when I first saw the paradigm he was putting
forward I knew this was something new which was
only then starting to become reality due to the
internet.
I believe he may deserve a doctorate for three
reasons...
1) He's a very good programmer and a smart cookie.
2) He's popularised a new way of producing software.
3) He's changed the course of history.
Now he'll have to field silly phone calls from
panicing babysitters who pick up the phone book
and look at the first Dr. they see.
Here is the comment I sent to Microsoft...
I would be more supportive of Microsofts freedom to innovate if microsoft was more supportive of my freedom to stay with a stable system.
Incompatible file formats have forced me through various upgrades to remain able to share files with my colleagues for no benefit to me (I need no more from a word processor than I had with Microsoft Word 2)
I also disagree with innovation when it gets in the way of shipping a stable, reliable, efficient product.
Please remember that there are two kinds of innovation, the good and the bad.