Is it still correct that when partitioning a (magnetic) hard drive, the later partitions are faster because they're located near the outside of the disk?
I thought the disk slows down as the read/write head moves outwards to maintain a constant read speed. But if the spool was a constant speed the read times would be shorter on the outer parts of the platter.
I would assume non primary partitions would suffer a hit due to the file table being in some arbitrary location.
I'm only taking a stab, no facts unfortunately.
pYro_65:
I thought the disk slows down as the read/write head moves outwards to maintain a constant read speed. But if the spool was a constant speed the read times would be shorter on the outer parts of the platter.
This is only true of CD/DVD drives. Magnetic drives spin at a constant speed. Data is written from the center out (which is the same for CD/DVD).
And while read times are shorter on the outside ring, the lookup table is still on the inner circles, so the head will be traveling back and forth anyway. When measured from the moment the request comes in to the time the data is send back, it's actually shorter to read the inner circles than it is to read the outside circles because the reading head isn't traveling that far. However, that time difference is so freaking miniscule, it's not worth sweating over it.
As far as partitioning goes, the real reason here is why? Consider that the same reading head has to travel across the same platter, regardless of what partition the data is in. In some ways, partitioning a drive slows it down because the reading head is now traveling back and forth several times. And partitioning really doesn't gain you any advantage anymore. If the drive is going to fail, it's going to fail. No amount of partitions will save your data. Drives are cheap, get a second one. Or if you're like me, a 7th one.
KirAsh4:
As far as partitioning goes, the real reason here is why?
The point of partitioning isn't to protect data (although it can do this; recovery, though, may end up being expensive) - partitioning is useful in the instance where you want to keep your data (and possibly applications) separate from your system install; if you ever wanted to wipe your system and start fresh, you could just wipe the partition and start over.
Then again, what you say is true about drives being cheap: Have one drive for the system (and if you have the cash, make it an SSD), another for the data...
KirAsh4:
And while read times are shorter on the outside ring, the lookup table is still on the inner circles, so the head will be traveling back and forth anyway.
I thought the file system structures always stay within the partition.
I'm actually thinking about partitioning a 1TB drive so that the last 128GB or so would be the root filesystem, the first 16GB or so (depending on amount of RAM) for suspend to disk, and the rest for bulk storage.
NiHaoMike:
Is it still correct that when partitioning a (magnetic) hard drive, the later partitions are faster because they're located near the outside of the disk?
Wouldn't the inside be faster? Less distance to travel?
Anyway, modern operating system cache disk accesses, so I wouldn't worry too much.
cr0sh:
The point of partitioning isn't to protect data (although it can do this; recovery, though, may end up being expensive) - partitioning is useful in the instance where you want to keep your data (and possibly applications) separate from your system install; if you ever wanted to wipe your system and start fresh, you could just wipe the partition and start over.
Yes and no. It depends on the operating system. Under Win7, you will quickly discover that your files and folders are not accessible because the UUID is different. There are ways to fix that, but it's still a pain to have to deal with it. Been there, had to deal with it ... decided it wasn't worth and bought a $100 2TB drive (t'was on special.)
if your racing from the start of the track to the end of the track yes, if your data is longer than the track, no cause you have to switch tracks more often
This awakens my dormant past when I was young and hard drives came in MBs. That was the 90's by the way 8). From what I read during that time, a smart way to speed up hard drive access is to format a hard drive so that the outer track starting points chases the inner track ending points. This way for a large file stored on multiple tracks, the hard drive starts from the first track, finishes the entire track, moves read head to one track out, just in time to catch next track's starting point. If you radially format your tracks, you always miss your next track and have to wait for a whole rotation. Of course that is for storing large files such as page file or something at least MB size. I don't know if this technique is still in use and even an English term of it. Floppies probably have the same thing. These spinning discs have a servo to maintain speed, instead of varying them over different tracks.
I wonder if any multiple read head discs were designed to increase the output speed 
liudr:
I wonder if any multiple read head discs were designed to increase the output speed
I'm sure waaaaay back in the early days of disk/hard drives this was tried (but probably had issues due to greater mass on the arm); it was certainly done for drum memory systems, which were somewhat akin to predecessors of disk/hard drives. What ended up happening with hard drives was going to multiple platters...
And that would only work once, or if you consistently write the same exact size file back in after deleting another. Otherwise welcome to the world of fragmentation where you get a single file scattered all over the place. Single track files go to hell in a hand basket really fast that way.
what you describe should be in the low level format of a drive, which you dont do .. you could back in the day but as time marches on the machines that wrote the hard facts about the disc's were using much stronger magnetics so if you try to low level format them again it may only stick for a little while before getting all sorts of funny on you.
You are right Osgeld. Back in the 90's you can't low-level format the way I described in DOS. The bios or dos interrupts that you can use to "format" a disc can't do this. Some bios setup programs had low-level format functions. I bet some special software had to be used to mass format at the factory or somehow and end user had to access the IDE controller instead of going through bios interrupt calls to do it.
http://www.tomshardware.com/forum/259256-32-raid-questions-short-stroking-striping
Looks like it's true, but the other way around (first partition being the fastest). And if the bulk partition is not accessed very frequently, there's the benefit of decreasing the average seek time.
My next PC will hopefully have an SSD. Just not sure when that will happen as I'm not all that excited about leaving XP for windows 7 but I won't get a new machine until I'm ready to upgrade to W7, so it's kind of a mexican standoff inside by brain.
Lefty
liudr! you just made cold shudders go down my spine!
I still have a whole bunch of MFM drive controllers.
What a freakin pain they were/are.
The controller board is not just the interface to the PC bus, but has the head stepper drivers on it as well as the analogue electronics for the read/write heads.
to format the drive you run DEGUB.COM and execute a bit of code on the controller card.
I still get people asking me to recover data from MFM drives!
SCSI and to a lesser extent IDE made life much simpler!
As for the partitioning for speed, are we talking Linux?
I've been using a seperate drive for swap since day 1.
It makes a HUGE difference.
Under Debian on my K7, My drive map looks like
/dev/hda1 - boot (about 10M)
/dev/hda2 - root (about 8G)
/dev/hdb - DVD rom
/dev/hdc1 - swap (1G)
/dev/hdd - CDROM/burner
/dev/hde1 - work (160G)
/dev/hdf1 - work2 (160G)
/dev/hdg1 - videoscratch (320G)
/dev/hdi1 - audioscratch (320G)
on my K6, My drive map looks like
/dev/hda1 - boot (about 10M)
/dev/hda2 - DR-DOS(about 800M)
/dev/hdb1 - root (40G) - Debian
/dev/hdc1 - windose 98se(40G)
/dev/hdd - DVD
/dev/hde1 - swap (800M)
/dev/hdf1 - work(80G)
on my DX4, My drive map looks like
/dev/hda1 - boot (about 10M)
/dev/hda2 - DR-DOS(about 1G)
/dev/hdb - DVD rom
/dev/sda1 - root (100M) - Debian
/dev/sdb1 - swap(80M)
/dev/sdc1 - usr(240M)
/dev/sdd1 - work(240M)
Having a separate swap volume way bigger than my physical RAM makes things sooo much faster.
Especially on my DX4 that only has 96M of RAM.
But I can go from X to DR-DOS running under DOSEMU as quickly as I can on the 600Mhz K6 or 800Mhz K7.
The DX4 has a VLB MF card from hell, SCSI, IDE, 2xRS232, 2xParralell, Joystick port.
Having a pure SCSI DOS, Linux or Windose system is pretty fast.
Just a pity they are getting hard to find and have always been expensive.
Unless some unanticipated breakthrough happens, SSDs are not going to compete with magnetic disks in capacity anytime soon. Most likely you'll have a SSD for storing frequently used files and a magnetic disk for storing less used files.