RE: [SLE] swap space question in 9.3 install
On Sunday, July 17, 2005 @ 10:45 PM I wrote:
On Sunday, July 17, 2005 @ 7:06 PM, Nick Jones wrote:
Sony Lies. Thats the max they sold it with, so as far as they are concerned its the max. Its well known their machines take more memory than they say.
Go to crucial.com and enter the machine model, and get a quote.
Firstly, let me apologize to all for the empty post. Lesson learned: don't use a Yahoo account on this list, Yahoo accounts suck for lists. I've changed email addresses.
Anyhoo, Crucial doesn't seem to sell RDRAM memory. The cheapest I can find for a 128mb stick is $44 at Newegg. Still not worth it. According to Sony's website, the max memory this machine can take is 512MB (4 x 128MB sticks) so is that figure wrong? Also, I think I figured out my problem, SUSE wants the boot partition to be at the beginning of the disk, but right now the swap space is there. I think I'll boot Knoppix back up and put an ext3 partition at the beginning of the disk and put the swap after that. Unless I can do it with the 9.3 network install disk...
Thanks -Nick
The first partition on my SuSE Linux box is a swap partition.
Greg Wallace
I think you always want your swap partition on high disk. One disk revolution moves more data on the outer part of the disk than on the inner part. Like being on the edge of a merry go-around instead of at the center, it's moving faster. For swapping in and out of memory, I would think that would be the best setup. Greg W
Greg Wallace wrote:
The first partition on my SuSE Linux box is a swap partition.
Greg Wallace
I think you always want your swap partition on high disk. One disk revolution moves more data on the outer part of the disk than on the inner part. Like being on the edge of a merry go-around instead of at the center, it's moving faster. For swapping in and out of memory, I would think that would be the best setup.
How does one know where on the platter "high disk" is?
On 18/07/05, James Knott
I think you always want your swap partition on high disk. One disk revolution moves more data on the outer part of the disk than on the inner part. Like being on the edge of a merry go-around instead of at the center, it's moving faster. For swapping in and out of memory, I would think that would be the best setup.
How does one know where on the platter "high disk" is?
I would imagine that is a job left to the operating system. Hence there being a specific 'swap' partition in Linux. -- Take care. Kevan Farmer 34 Hill Street Cheslyn Hay Staffordshire WS6 7HR
On Mon, 2005-07-18 at 11:43 +0100, Kevanf1 wrote:
On 18/07/05, James Knott
wrote: I think you always want your swap partition on high disk. One disk revolution moves more data on the outer part of the disk than on the inner part. Like being on the edge of a merry go-around instead of at the center, it's moving faster. For swapping in and out of memory, I would think that would be the best setup.
How does one know where on the platter "high disk" is?
I would imagine that is a job left to the operating system. Hence there being a specific 'swap' partition in Linux.
I always thought the track 0 was at the inner most part of the disk. Keep in mind that on older systems they had a problem if the boot partition was more than 1024 sectors, hence the reason most people got into the habit of creating a small boot partition at the beginning of the disk. As the OP stated this is an older system perhaps this is needed here. -- Ken Schneider UNIX since 1989, linux since 1994, SuSE since 1998 "The day Microsoft makes something that doesn't suck is probably the day they start making vacuum cleaners." -Ernst Jan Plugge
Ken Schneider wrote:
How does one know where on the platter "high disk" is?
I would imagine that is a job left to the operating system. Hence there being a specific 'swap' partition in Linux.
I always thought the track 0 was at the inner most part of the disk. Keep in mind that on older systems they had a problem if the boot partition was more than 1024 sectors, hence the reason most people got into the habit of creating a small boot partition at the beginning of the disk. As the OP stated this is an older system perhaps this is needed here.
Which track is closest to the hub or outside is determined by the manufacture. I've seen drives with track 0 on the outside. Also, with modern drives, there's no 1:1 mapping of tracks that the computer sees, to the tracks actually on the platters.
Kevan, On Monday 18 July 2005 03:43, Kevanf1 wrote:
On 18/07/05, James Knott
wrote: I think you always want your swap partition on high disk. One disk revolution moves more data on the outer part of the disk than on the inner part. Like being on the edge of a merry go-around instead of at the center, it's moving faster. For swapping in and out of memory, I would think that would be the best setup.
How does one know where on the platter "high disk" is?
I would imagine that is a job left to the operating system. Hence there being a specific 'swap' partition in Linux.
No. the OS does not make such a determination. The disk partitions (the number in, e.g., /dev/hda0) represent contiguous portions of the disk in a strict order from inside to outside or vice versa. (I now forget which the convention is, but I think it's the opposite of CDs / DVDs. They start at the inside and go outward. Hard drives, if I'm not mistaken, start at the outside with partition 0 and move inward) The fact that a partition is labelled a swap partition is mostly redundant (not strictly necessary), but it prevents the system from swapping on a file systerm partition or one of the mkfs series of programs from creating a file system in a swap partition. But where the swap goes (which disk and which partition) is up to the person configuring the system.
-- Take care. Kevan Farmer
Randall Schulz
-----Original Message----- On Monday, July 18, 2005 @ 5:40 AM, Randall Schulz wrote:
Kevan,
On Monday 18 July 2005 03:43, Kevanf1 wrote: On 18/07/05, James Knott
wrote: I think you always want your swap partition on high disk. One disk revolution moves more data on the outer part of the disk than on the inner part. Like being on the edge of a merry go-around instead of at the center, it's moving faster. For swapping in and out of memory, I would think that would be the best setup.
How does one know where on the platter "high disk" is?
I would imagine that is a job left to the operating system. Hence there being a specific 'swap' partition in Linux.
No. the OS does not make such a determination.
The disk partitions (the number in, e.g., /dev/hda0) represent contiguous portions of the disk in a strict order from inside to outside or vice versa. (I now forget which the convention is, but I think it's the opposite of CDs / DVDs. They start at the inside and go outward. Hard drives, if I'm not mistaken, start at the outside with partition 0 and move inward)
I would have thought CDs and DVDs would work the same as hard drives (working from the outside in).
The fact that a partition is labelled a swap partition is mostly redundant (not strictly necessary), but it prevents the system from swapping on a file systerm partition or one of the mkfs series of programs from creating a file system in a swap partition.
So the space is set aside for memory swapping, thus "swap", right? And, as you say, the os won't store any permanent files there. The os can then dump RAM buffers into it in whole, starting right at the beginning of the swap space and moving forward, without worrying about stepping on any files.
But where the swap goes (which disk and which partition) is up to the person configuring the system.
At installation time, where Nick (Nick being the guy who started this thread, back in the day) is right now, right?
-- Take care. Kevan Farmer
Randall Schulz
Greg W
Greg, On Monday 18 July 2005 07:27, Greg Wallace wrote:
-----Original Message-----
On Monday, July 18, 2005 @ 5:40 AM, Randall Schulz wrote: ...
The disk partitions (the number in, e.g., /dev/hda0) represent contiguous portions of the disk in a strict order from inside to outside or vice versa. (I now forget which the convention is, but I think it's the opposite of CDs / DVDs. They start at the inside and go outward. Hard drives, if I'm not mistaken, start at the outside with partition 0 and move inward)
I would have thought CDs and DVDs would work the same as hard drives (working from the outside in).
And yet, that's how it is. Look at some partially recorded CDs and you'll see that the recorded area is always at the center and proceeding varying distances out from there, depending on how much is recorded on the disc.
The fact that a partition is labelled a swap partition is mostly redundant (not strictly necessary), but it prevents the system from swapping on a file systerm partition or one of the mkfs series of programs from creating a file system in a swap partition.
So the space is set aside for memory swapping, thus "swap", right? And, as you say, the os won't store any permanent files there. The os can then dump RAM buffers into it in whole, starting right at the beginning of the swap space and moving forward, without worrying about stepping on any files.
Yes. They're segregated. In fact, in the old days (and quite possibly on other, non-Linux, Unix-like systems, still today) swap area could share a partition with a file system. The file system would be formatted for less than the full capacity of the partition and the swap space would be created beginning after the end of the file system area).
But where the swap goes (which disk and which partition) is up to the person configuring the system.
At installation time, where Nick (Nick being the guy who started this thread, back in the day) is right now, right?
True. And with a modicum of care, you can reconfigure later, too.
Greg W
Randall Schulz
On Mon, 2005-07-18 at 06:27 -0800, Greg Wallace wrote:
-----Original Message----- On Monday, July 18, 2005 @ 5:40 AM, Randall Schulz wrote:
Kevan,
The disk partitions (the number in, e.g., /dev/hda0) represent contiguous portions of the disk in a strict order from inside to outside or vice versa. (I now forget which the convention is, but I think it's the opposite of CDs / DVDs. They start at the inside and go outward. Hard drives, if I'm not mistaken, start at the outside with partition 0 and move inward)
I would have thought CDs and DVDs would work the same as hard drives (working from the outside in).
No. Next time you burn a partial CD take a look at were it put the data. It starts at the inner most part and works toward the outside. -- Ken Schneider UNIX since 1989, linux since 1994, SuSE since 1998 "The day Microsoft makes something that doesn't suck is probably the day they start making vacuum cleaners." -Ernst Jan Plugge
On Monday, July 18, 2005 @ 6:59 AM, Ken Schneider wrote:
On Mon, 2005-07-18 at 06:27 -0800, Greg Wallace wrote:
-----Original Message----- On Monday, July 18, 2005 @ 5:40 AM, Randall Schulz wrote:
Kevan,
The disk partitions (the number in, e.g., /dev/hda0) represent contiguous portions of the disk in a strict order from inside to outside or vice versa. (I now forget which the convention is, but I think it's the opposite of CDs / DVDs. They start at the inside and go outward. Hard drives, if I'm not mistaken, start at the outside with partition 0 and move inward)
I would have thought CDs and DVDs would work the same as hard drives (working from the outside in).
No. Next time you burn a partial CD take a look at were it put the data. It starts at the inner most part and works toward the outside.
-- Ken Schneider UNIX since 1989, linux since 1994, SuSE since 1998
"The day Microsoft makes something that doesn't suck is probably the day they start making vacuum cleaners." -Ernst Jan Plugge
Sure enough. I had only payed casual attention to the surface of CDs before. As you say, you can see the area that has been burned (on the inside of the CD). Not sure what the rational would be for that, but maybe the speed at which it could pull and push data isn't a factor for CDs (or is so insignificant that other reasons for putting it on the inside would override them). Or maybe the data is packed tighter on the inside to offset the shorter tracks. Anyway, it seems counter-intuitive. Greg Wallace
Greg, On Monday 18 July 2005 16:34, Greg Wallace wrote:
...
Sure enough. I had only payed casual attention to the surface of CDs before. As you say, you can see the area that has been burned (on the inside of the CD). Not sure what the rational would be for that, but maybe the speed at which it could pull and push data isn't a factor for CDs (or is so insignificant that other reasons for putting it on the inside would override them). Or maybe the data is packed tighter on the inside to offset the shorter tracks. Anyway, it seems counter-intuitive.
CDs are recorded at constant linear velocity, so the number of sectors per linear cm of recording is fixed, but the number of sectors per revolution is not (that would be constant angular velocity, formerly common among hard drives, but no longer so, I don't believe). As such, it's irrelevant and arbitrary where recording starts. On the other hand, damage to the surface is much more likely near the outer edge because that's where people handle them mostly, so by starting recording at the inside, only the fullest discs will experience the maximum likelihood of damage to recorded areas.
Greg Wallace
Randall Schulz
On Monday, July 18, 2005 @ 9:06 PM, Randall Schulz wrote:
Greg,
On Monday 18 July 2005 16:34, Greg Wallace wrote:
...
Sure enough. I had only payed casual attention to the surface of CDs before. As you say, you can see the area that has been burned (on the inside of the CD). Not sure what the rational would be for that, but maybe the speed at which it could pull and push data isn't a factor for CDs (or is so insignificant that other reasons for putting it on the inside would override them). Or maybe the data is packed tighter on the inside to offset the shorter tracks. Anyway, it seems counter-intuitive.
CDs are recorded at constant linear velocity, so the number of sectors per linear cm of recording is fixed, but the number of sectors per revolution is not (that would be constant angular velocity, formerly common among hard drives, but no longer so, I don't believe).
As such, it's irrelevant and arbitrary where recording starts.
On the other hand, damage to the surface is much more likely near the outer edge because that's where people handle them mostly, so by starting recording at the inside, only the fullest discs will experience the maximum likelihood of damage to recorded areas.
Greg Wallace
Randall Schulz
Just to make sure I'm understanding this, are you saying that if I'm burning on the inside of a CD or DVD, the disk is actually rotating faster than if I'm burning on the outside? Greg W
Greg, On Monday 18 July 2005 22:42, Greg Wallace wrote:
...
Just to make sure I'm understanding this, are you saying that if I'm burning on the inside of a CD or DVD, the disk is actually rotating faster than if I'm burning on the outside?
Indeed. To learn more, search the Web for "CD DVD CLV CAV". E.g.: http://a9.com/CD%20DVD%20CLV%20CAV
Greg W
RRS
On Monday, July 18, 2005 @ 10:13 PM, Randall Schulz wrote:
Greg,
On Monday 18 July 2005 22:42, Greg Wallace wrote:
...
Just to make sure I'm understanding this, are you saying that if I'm burning on the inside of a CD or DVD, the disk is actually rotating faster than if I'm burning on the outside?
Indeed.
To learn more, search the Web for "CD DVD CLV CAV".
Greg W
RRS
Interesting. I have no idea whether I have a CLV or CAV CD, but since it's several years old, it's probably CLV (sounds like CAV is somewhat new on the scene). Greg Wallace P. S.: Speaking of new technology, did you hear about the new TV's that show two programs simultaneously on the same screen? If you're sitting to the left of center, you see one program. If you're sitting to the right of center, you see another. I guess if two people were watching different programs they'd have to be wearing headphones so they could hear the audio for just the one they're seeing.
Greg Wallace wrote:
Just to make sure I'm understanding this, are you saying that if I'm burning on the inside of a CD or DVD, the disk is actually rotating faster than if I'm burning on the outside?
Assuming constant linear velocity, yes.
-----Original Message-----
From: James Knott
Greg Wallace wrote:
Just to make sure I'm understanding this, are you saying that if I'm burning on the inside of a CD or DVD, the disk is actually rotating faster than if I'm burning on the outside?
Assuming constant linear velocity, yes.
Actually this is pretty easy to see that CDs spin at different velocities. Back when I was in high school, I drove an old beater car (didn't we all?) and therefore couldn't afford an in-dash CD player. I had my handy portable unit (that I still have and it still works, go Pioneer!) connected through the tape deck to listen to music. The lid of the CD player was clear plastic. I know for a fact that if you start at track 1 (on the innermost portion of the disk) it spins much faster than if you progress to the last track on a disk. You can watch the spinning velocity change in this manner. I thought it was pretty neat when I discovered that 7 years ago, but that's far before I knew anything about computers or CDs for that matter. -Nick
On Tue, 2005-07-19 at 13:12 -0500, Nick Jones wrote:
-----Original Message----- From: James Knott
To: suse-linux-e@suse.com Date: Tue, 19 Jul 2005 05:55:19 -0400 Subject: Re: [SLE] swap space question in 9.3 install Greg Wallace wrote:
Just to make sure I'm understanding this, are you saying that if I'm burning on the inside of a CD or DVD, the disk is actually rotating faster than if I'm burning on the outside?
Assuming constant linear velocity, yes.
Actually this is pretty easy to see that CDs spin at different velocities. Back when I was in high school, I drove an old beater car (didn't we all?) and therefore couldn't afford an in-dash CD player.
I had a 45rpm record player in mine, wish I still had it. Sure fooled a lot of people wanting to know which station you had tuned in. :-)
I had my handy portable unit (that I still have and it still works, go Pioneer!) connected through the tape deck to listen to music. The lid of the CD player was clear plastic. I know for a fact that if you start at track 1 (on the innermost portion of the disk) it spins much faster than if you progress to the last track on a disk. You can watch the spinning velocity change in this manner. I thought it was pretty neat when I discovered that 7 years ago, but that's far before I knew anything about computers or CDs for that matter.
-Nick
-- Ken Schneider UNIX since 1989, linux since 1994, SuSE since 1998 "The day Microsoft makes something that doesn't suck is probably the day they start making vacuum cleaners." -Ernst Jan Plugge
Nick Jones wrote:
-----Original Message----- From: James Knott
To: suse-linux-e@suse.com Date: Tue, 19 Jul 2005 05:55:19 -0400 Subject: Re: [SLE] swap space question in 9.3 install Greg Wallace wrote:
Just to make sure I'm understanding this, are you saying that if I'm burning on the inside of a CD or DVD, the disk is actually rotating faster than if I'm burning on the outside? Assuming constant linear velocity, yes.
Actually this is pretty easy to see that CDs spin at different velocities.
While CDs generally spin at a constant linear velocity, the spec also allows for constant angular velocity. There is nothing to stop someone from making a CAV CD.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Monday 2005-07-18 at 15:34 -0800, Greg Wallace wrote:
Sure enough. I had only payed casual attention to the surface of CDs before. As you say, you can see the area that has been burned (on the inside of the CD). Not sure what the rational would be for that, but maybe the speed at which it could pull and push data isn't a factor for CDs (or is so insignificant that other reasons for putting it on the inside would override them). Or maybe the data is packed tighter on the inside to offset the shorter tracks. Anyway, it seems counter-intuitive.
The reason is very simple: there are different CD sizes (diameter). You can put a mini CD in a big player and it works, because it starts reading at the center. - -- Cheers, Carlos Robinson -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFC4tdVtTMYHG2NR9URAgS5AJsGtA+n0gN4opwAHgj0Ndzer9TVSACZAUuQ Ud/ePBnYCZNRVO6VLmiGb54= =iJhu -----END PGP SIGNATURE-----
So the space is set aside for memory swapping, thus "swap", right? And, as you say, the os won't store any permanent files there. The os can then dump RAM buffers into it in whole, starting right at the beginning of the swap space and moving forward, without worrying about stepping on any files. There are no permanent files ever stored in swap. In essence, the name swap is a misnomer. Your memory is divided up into pages. The page size is a function of the hardware and OS. The Linux kernel only needs to store
On Monday 18 July 2005 10:27 am, Greg Wallace wrote:
dirty pages in the swap space. During periods of high memory use, systems
may start to use the swap space excessively.
(I don't want to get into a detailed technical discussion of how paging
works).
Also note that Linux supports either a separate swap partition, a swap file,
or both. The swapon(8) command may be used to turn paging areas on or off.
--
Jerry Feldman
Greg Wallace wrote:
I would have thought CDs and DVDs would work the same as hard drives (working from the outside in).
No. CDs & DVDs start from the hub. You may recall the old vinyl records. The different sized disks had different start points, which made things more complex for turntables & changers. With all CDs & DVDs starting from the inside track, all disks, no matter what size, all start from the same point. With hard drives, it's irrelevant where "track 0" is.
On Monday, July 18, 2005 @ 12:57 PM, James Knott wrote:
Greg Wallace wrote:
I would have thought CDs and DVDs would work the same as hard drives (working from the outside in).
No. CDs & DVDs start from the hub. You may recall the old vinyl records. The different sized disks had different start points, which made things more complex for turntables & changers. With all CDs & DVDs starting from the inside track, all disks, no matter what size, all start from the same point. With hard drives, it's irrelevant where "track 0" is.
It may be true that it's irrelevant where "track 0" is, but, physically, you can move data to and from a disk faster from the outside of it than from the inside. That's just pure physics at work. Norton builds a utility called "Speed Disk" for Windows machines. That utility takes data that has been plopped toward the inside of a disk and moves it toward the outside of the disk. It significantly increases file access speed. I believe it actually also uses last access date, among maybe other factors, to decide the order in which to place the files on the outside of the disk. You can also tell it to move something to the inside of the disk if it's a file you use on a very infrequent basis, which leaves more available outer disk space for more frequently used files. Greg Wallace
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Monday 2005-07-18 at 16:53 -0800, Greg Wallace wrote:
It may be true that it's irrelevant where "track 0" is, but, physically, you can move data to and from a disk faster from the outside of it than from the inside. That's just pure physics at work.
That may be true. But...
Norton builds a utility called "Speed Disk" for Windows machines. That utility takes data that has been plopped toward the inside of a disk and moves it toward the outside of the disk. It significantly increases file access speed. I believe it actually also uses last access date, among maybe other factors, to decide the order in which to place the files on the outside of the disk. You can also tell it to move something to the inside of the disk if it's a file you use on a very infrequent basis, which leaves more available outer disk space for more frequently used files.
That utility (and similar uts) moved files not to place them near the edge or the center of the disk, but to have them near the start of the "logical disk" in dos/windows. The rationale is that the FAT table and the root directory are at that logical disk start, and that any file access needed first to access the directory tables and fat tables. Judiciously placing files on disk optimized head movement, that is the biggest time waster in disk access. Remember also that those tables were not cached by default (you could remove the floppy any time). - -- Cheers, Carlos Robinson -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFC4tm1tTMYHG2NR9URAjj6AJ4jmIQY4hxzxbryRrod7yuj51IvpQCfQEGc 3HCaW5aY/2WEwgwUmiiQpzA= =Sg4t -----END PGP SIGNATURE-----
Kevanf1 wrote:
On 18/07/05, James Knott
wrote: I think you always want your swap partition on high disk. One disk revolution moves more data on the outer part of the disk than on the inner part. Like being on the edge of a merry go-around instead of at the center, it's moving faster. For swapping in and out of memory, I would think that would be the best setup. How does one know where on the platter "high disk" is?
I would imagine that is a job left to the operating system. Hence there being a specific 'swap' partition in Linux.
My point was, how does one know what disk addresses are on the outer (faster) tracks. The manufacturer can have the 1st track be either inside or outside and with logical sector addressing, the sectors could literally be anywhere on the disk.
The first partition on my SuSE Linux box is a swap partition.
Greg Wallace
I think you always want your swap partition on high disk. One disk revolution moves more data on the outer part of the disk than on the inner part. Like being on the edge of a merry go-around instead of at the
On Monday, July 18, 2005 @ -----Original Message----- From: James Knott [mailto:james.knott@rogers.com] Sent: Monday, July 18, 2005 1:55 AM To: suse-linux-e@suse.com Subject: Re: [SLE] swap space question in 9.3 install Greg Wallace wrote: center,
it's moving faster. For swapping in and out of memory, I would think that would be the best setup.
How does one know where on the platter "high disk" is?
I was under the impression that installation started at the outside of the disk and worked inward. Based on the order in which you specify them in your initial partitioning at installation time, I recall seeing offsets under partitioning that indicated the partitions started at low numbers for partition 1 and went up for each subsequent partition. Surely Linux would not, by default, start installing on the inside of the disk and work outward(?)! Isn't that the way all operating systems are installed, starting on the outside of the disk? If I am not correct here, I would be interested in knowing about it. Greg Wallace
I always try to allocate my sawp partition to try to minimize head movement.
While, the disk geometry is relevant, head movement is a very significant
factor. Most home users are going to allocate 1 partition for swap and 1
partition for Linux. In this case, placing swap before the Linux partition
might be better.
But, if you want to do an analysis, take a look at where your busiest
partition is. That could be /var because /var contains all the log files
and spools.
If you have 2 drives, and one is a Linux drive and the other is a Windows
drive, placing the swap on the Windows drive will have the advantage of not
affecting head movement on the Linux drive, and if it is on a separate IDE
channel, swap I/O can be overlapped.
However, if your system is swapping excessively, then you need to add more
memory.
--
Jerry Feldman
On Monday, July 18, 2005 @ 4:47 AM, Jerry Feldman wrote:
I always try to allocate my sawp partition to try to minimize head movement. While, the disk geometry is relevant, head movement is a very significant factor. Most home users are going to allocate 1 partition for swap and 1 partition for Linux. In this case, placing swap before the Linux partition might be better.
But, if you want to do an analysis, take a look at where your busiest partition is. That could be /var because /var contains all the log files and spools.
If you have 2 drives, and one is a Linux drive and the other is a Windows drive, placing the swap on the Windows drive will have the advantage of not
affecting head movement on the Linux drive, and if it is on a separate IDE channel, swap I/O can be overlapped.
However, if your system is swapping excessively, then you need to add more memory.
-- Jerry Feldman
Boston Linux and Unix user group http://www.blu.org PGP key id:C5061EA9 PGP Key fingerprint:053C 73EC 3AC1 5C44 3E14 9245 FB00 3ED5 C506 1EA9
From above -- While, the disk geometry is relevant, head movement is a very significant factor. Most home users are going to allocate 1 partition for swap and 1 partition for Linux. In this case, placing swap before the Linux partition might be better.
That was really the point I was trying to get across; i. e., that it does make a difference where you put swap. Linux defaults swap to the first partition and there is actually a reason for putting it there. Greg Wallace
On Monday 18 July 2005 9:12 am, Greg Wallace wrote:
That was really the point I was trying to get across; i. e., that it does make a difference where you put swap. Linux defaults swap to the first partition and there is actually a reason for putting it there. In the case where the user takes the defaults, I agree with this.
On my home system I have 2 disks, both have swap. disk 1 has /home
and /usr/local as well as the Windows partition. disk2 has SuSe 9.3 root.
(disk1 has the older SuSE 9.2). Since the root partition tends to be more
active than the /home partition, I have the primary swap allocated on
disk1. But, I generally don't do much swapping.
--
Jerry Feldman
On Monday 18 July 2005 15:12, Greg Wallace wrote:
While, the disk geometry is relevant, head movement is a very significant factor. Most home users are going to allocate 1 partition for swap and 1 partition for Linux. In this case, placing swap before the Linux partition might be better.
That was really the point I was trying to get across; i. e., that it does make a difference where you put swap. Linux defaults swap to the first partition and there is actually a reason for putting it there.
Greg Wallace
That was really the point I was trying to get across; i. e.,
does make a difference where you put swap. Linux defaults swap to
that it the
first partition and there is actually a reason for putting it there.
Greg Wallace
On this particlar machine, it wanted the boot partition as the first partition on the disk because the BIOS was four years old and it needed it there for booting purposes (or at least that's the way I interpret the messages that it spit out at me, I should have written them down, whoops). Also, for some reason, it wanted 900+MB of swap for a machine with 128mb of RAM. Who am I to argue? I went ahead with the reccomended setup and all went just fine. This box is only a test machine anyway for some web development work. Chances are it'll be reformatted and reinstalled many times over it's remaining lifetime. Thanks -Nick
On Monday, July 18, 2005 @ 4:28 PM, Nick Jones wrote:
That was really the point I was trying to get across; i. e.,
does make a difference where you put swap. Linux defaults swap to
that it the
first partition and there is actually a reason for putting it there.
Greg Wallace
On this particlar machine, it wanted the boot partition as the first partition on the disk because the BIOS was four years old and it needed it there for booting purposes (or at least that's the way I interpret the messages that it spit out at me, I should have written them down, whoops). Also, for some reason, it wanted 900+MB of swap for a machine with 128mb of RAM. Who am I to argue? I went ahead with the reccomended setup and all went just fine. This box is only a test machine anyway for some web development work. Chances are it'll be reformatted and reinstalled many times over it's remaining lifetime.
Thanks -Nick
If your machine does indeed requires the first boot partition to be the first one on the disk, then I guess you have no choice. In any event, the setup you said you used "-Delete partition /dev/hda1 509.9 MB (Linux Swap) -Create boot partition 64.4 MB (/dev/hda1 with ext2) -Create swap partition 901.1 MB on /dev/hda2 -Create root partition 73.6 GB (/dev/hda3 with reiser)" should be fine. You only have swap offset 64.4 MB from the beginning, and that's a pretty insignificant offset from the beginning of the disk. You just want to, all other things being equal, try to keep swap toward the top of the disk. By the way, why did you use ext2 for boot and reiser for root? Just curious. Can you not have two reiser partitions on one system? Greg Wallace
-----Original Message-----
From: "Greg Wallace"
If your machine does indeed requires the first boot partition to be the first one on the disk, then I guess you have no choice. In any event, the setup you said you used "-Delete partition /dev/hda1 509.9 MB (Linux Swap) -Create boot partition 64.4 MB (/dev/hda1 with ext2) -Create swap partition 901.1 MB on /dev/hda2 -Create root partition 73.6 GB (/dev/hda3 with reiser)" should be fine. You only have swap offset 64.4 MB from the beginning, and that's a pretty insignificant offset from the beginning of the disk. You just want to, all other things being equal, try to keep swap toward the top of the disk. By the way, why did you use ext2 for boot and reiser for root? Just curious. Can you not have two reiser partitions on one system?
Greg Wallace
Well, those partition options were all the ones that SUSE reccomended for this system by default. It's a 4 year old Sony Vaio 1.3Ghz P4, 128MB PC800 RDRAM, and an 80GB Maxtor HDD. I guess it just wanted to be wierd. I know you can have two reiser partitions (there are 2 on my desktop system and 3 on my main web server). I've heard that before too, about how swap should be at the beginning of a disk for better performance, and that's how all my other systems are set up. I guess this one just had a BIOS that required that boot partition at the beginning of the disk. -Nick
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Monday 2005-07-18 at 17:27 -0800, Greg Wallace wrote:
By the way, why did you use ext2 for boot and reiser for root? Just curious. Can you not have two reiser partitions on one system?
The /boot partition is almost always ext2 because reiserfs requires at least 100MB. Also, the journal takes size, and it is pointless for a partition so small (it is always one or two tracks, 10 or 20 Mb or roundabouts). For the same reason, it is not ext3. - -- Cheers, Carlos Robinson -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFC4twTtTMYHG2NR9URApq2AJ4qAHpI8p98rgMAOrXoTOGDez/p/ACgiQBN zqto+B1Z8Vf4p20gAZh7HLw= =t+k+ -----END PGP SIGNATURE-----
On Tue, 19 Jul 2005 10:27 am, Nick Jones wrote:
On this particlar machine, it wanted the boot partition as the first partition on the disk because the BIOS was four years old and it needed it there for booting purposes
I still like to keep a separate boot partition the issue of a kernel upgrade taking my vmlinuz or initrd out of the "bootable zone" is a piece of randomness I can do without. Apparently a more current plus is that it makes suspend/resume work much better too.
Also, for some reason, it wanted 900+MB of swap for a machine with 128mb of RAM. This box is only a test machine anyway for some web development work. Chances are it'll be reformatted and reinstalled many times over it's remaining lifetime.
YOU want more than 128M of RAM. IT was just allowing for the day you get it. ;^) -- Michael James michael.james@csiro.au System Administrator voice: 02 6246 5040 CSIRO Bioinformatics Facility fax: 02 6246 5166 No matter how much you pay for software, you always get less than you hoped. Unless you pay nothing, then you get more.
-----Original Message-----
From: Michael James
YOU want more than 128M of RAM. IT was just allowing for the day you get it. ;^)
Eh, I'm not keen on spending nearly $200 to bring it up to it's 512mb max. I think this machine can just sit headless in the closet and serve out MP3s and be a test web server for development work. I installed a minimal 9.3 system with Apache and Samba, and that's all I need it for. Thanks! -Nick
On Monday 18 July 2005 15:12, Greg Wallace wrote:
While, the disk geometry is relevant, head movement is a very significant factor. Most home users are going to allocate 1 partition for swap and 1 partition for Linux. In this case, placing swap before the Linux partition might be better.
That was really the point I was trying to get across; i. e., that it does make a difference where you put swap. Linux defaults swap to the first partition and there is actually a reason for putting it there.
If your disk starts swapping so much that you can tell the difference between one location on the disk and another, it will be the least of your worries. Your system will be basically useless. Swap is good in three cases 1. An application uses so much memory that you need to go in to swap to be able to let it run. Some multimedia editing programs are like that. There, speed of swap is almost totally irrelevant 2. To make room for cache. The kernel will move parts of memory that haven't been accessed in a long time to swap, so that more internal memory can be used to cache disk accesses. The speed of the swap is again almost totally irrelevant 3. To store memory images when you suspend to disk. Here you will want to make sure you have enough swap available to fit your memory image. But the speed is relatively insignificant, since the access is on boot with nothing else running, there will be no massive seeking back and forth, and the relative difference in speed between inner and outer cylinders I seriously doubt will make much of an impact But swap used for the general running of the system? Forget it. If it gets that far, you might as well give up
On Monday 18 July 2005 15:12, Greg Wallace wrote:
While, the disk geometry is relevant, head movement is a very significant factor. Most home users are going to allocate 1 partition for swap and 1 partition for Linux. In this case, placing swap before the Linux
On Monday, July 18, 2005 @ 4:02 PM, Anders Johansson wrote: partition
might be better.
That was really the point I was trying to get across; i. e., that it does make a difference where you put swap. Linux defaults swap to the first partition and there is actually a reason for putting it there.
If your disk starts swapping so much that you can tell the difference between one location on the disk and another, it will be the least of your worries.
Your system will be basically useless.
Swap is good in three cases
1. An application uses so much memory that you need to go in to swap to be able to let it run. Some multimedia editing programs are like that. There, speed of swap is almost totally irrelevant
2. To make room for cache. The kernel will move parts of memory that haven't been accessed in a long time to swap, so that more internal memory can be used to cache disk accesses. The speed of the swap is again almost totally irrelevant
3. To store memory images when you suspend to disk. Here you will want to make sure you have enough swap available to fit your memory image. But the speed
is relatively insignificant, since the access is on boot with nothing else running, there will be no massive seeking back and forth, and the relative difference in speed between inner and outer cylinders I seriously doubt will make much of an impact
But swap used for the general running of the system? Forget it. If it gets that far, you might as well give up
All true, but is there any reason to purposely move swap down during installation? If you move it down after your regular Linux partitions and that ends up putting it 40, 50, 60 G further down, I think that might make a noticeable difference in some cases. Of course, other factors will be at play, such as the speed of your connection (if you're moving data to an external device), etc., but why not set up you partitions in the best possible manner to begin with. Greg Wallace
On Tuesday 19 July 2005 03:02, Greg Wallace wrote:
All true, but is there any reason to purposely move swap down during installation? If you move it down after your regular Linux partitions and that ends up putting it 40, 50, 60 G further down, I think that might make a noticeable difference in some cases. Of course, other factors will be at play, such as the speed of your connection (if you're moving data to an external device), etc., but why not set up you partitions in the best possible manner to begin with.
Feel free to experiment with it. I stand by what I said in my other mail, that if you swap so much that you see the difference in read speed, then the system will be so bogged down that the disk read speed will be the least of your worries
Greg Wallace wrote:
How does one know where on the platter "high disk" is?
I was under the impression that installation started at the outside of the disk and worked inward. Based on the order in which you specify them in your initial partitioning at installation time, I recall seeing offsets under partitioning that indicated the partitions started at low numbers for partition 1 and went up for each subsequent partition. Surely Linux would not, by default, start installing on the inside of the disk and work outward(?)! Isn't that the way all operating systems are installed, starting on the outside of the disk? If I am not correct here, I would be interested in knowing about it.
My point is, that the average user does not know which tracks correspond to the inner or outer areas of the disk. The manufacturer can have track 1 closest to the hub or to the outside. Add logical sector addressing, and a sector could theoretically be anywhere. I have even seen disks, where every other track, going into the hub was skipped, and then when the heads reached the inner track, the track count continued on the previously skipped tracks, so that the last track was physically right next to the first and the "middle" tracks, were actually right near the hub. I have also worked on "head per track" drives, where the tracks were not in sequential order.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Monday 2005-07-18 at 04:01 -0800, Greg Wallace wrote:
I was under the impression that installation started at the outside of the disk and worked inward. Based on the order in which you specify them in your initial partitioning at installation time, I recall seeing offsets under partitioning that indicated the partitions started at low numbers for partition 1 and went up for each subsequent partition. Surely Linux would not, by default, start installing on the inside of the disk and work outward(?)! Isn't that the way all operating systems are installed, starting on the outside of the disk? If I am not correct here, I would be interested in knowing about it.
The OS is not interested in knowing whether a partition is at the inside or the outside, or if track 0 is the innermost or not. It doesn't really matter, it doesn't really know. That's is the HD manufacturer choice. The OS starts partitioning at track 0 and end at the last. Notice, for example, that the numbers of heads, tracks and sectors the OS thinks the HD has can be false. My HD reports having 16 heads, ie, 8 platters, when I know for a fact it only has two platters and three heads (the fourth is not installed on my model: the four models are differentiated in size by installing 1, 2, 3 or 4 heads). The HD bios just does an appropriate translation from real placements to logical, outside world, placements. So... you only really know that partition 0 is at the tracks the HD says it is. You know that you are accessing track 0 or 10000, but you don't really know where those are in fact. If you want to know whether a partition placed here or there is faster or slower, measure it: don't make assumptions. Just divide the disk in several partitions of the same type and size, initialize them, and make speed measurements (for example with hdparm). The results might not be what you expected, and might be different from make to make. - -- Cheers, Carlos Robinson -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.0 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFC4utjtTMYHG2NR9URArj9AKCUfBLP/bJbN1sZUTg1LhHRzNQ5PgCgj9+J ZBMA1Vn8cqDTgjLoH0t1Zqw= =Se4i -----END PGP SIGNATURE-----
participants (10)
-
Anders Johansson
-
Carlos E. R.
-
Greg Wallace
-
James Knott
-
Jerry Feldman
-
Ken Schneider
-
Kevanf1
-
Michael James
-
Nick Jones
-
Randall R Schulz