Re: How to align the hard disk



Rayees wrote:
Hi

I have a IO performance problem while I'm using SAP frontend at my Windows XP PC.

While contacting SAP, they said that this was due to misaligned disk.

They said. Go to "Windows System Information" (command line: winmsd). Open
"Components" / "Storage" / "Disks". You will find a "Partition starting
offset" or "Starting offset" in the list. The value has to be 64 * 512
(32,768), the default is unfortunately 63 * 512 (32,256)."

My question is how to change Partition starting offset value to 32768 instead of 32256.

Regards

Rayees

This is the manual, for a tool previously known as PowerQuest PartitionMagic.
Symantec now owns it.

ftp://ftp.symantec.com/public/english_us_canada/products/norton_partitionmagic/npmagic_8/manuals/norton_partitionmagic_8.pdf

Note some of the comments here:

http://en.wikipedia.org/wiki/Talk:Cylinder-head-sector

"The History section here is obviously a bit out of date, since CHS tuples
haven't actually corresponded to disk hardware in a very long time!"

I doubt very much, that you can establish a relationship between physical
cylinders on a disk, and the logical addressing used to access them.

From this page, you can download PartInNT.zip . That program displays
disk geometry. The "NT" version would be good for WinXP/Win2K.

http://service1.symantec.com/Support/powerquest.nsf/pfdocs/2004073190203662
( ftp://ftp.symantec.com/public/english_us_canada/tools/pq/utilities/PartInNT.zip )

If you unzip the file, and execute the program, it will display the partitioning
info for the disk. The "sectors per track" is defined as 63 (as part of being a
"large drive placeholder" for escaping the arithmetic bounds of CHS geometry). If
the declaration is 63 sectors per track, then the first track contains sectors
0 through 62, and the next cylinder will begin at 63. It is possible, even when
using PartitionMagic, that it will not allow you to start the partition offset
from the cylinder boundary. So, I guess my issue with the information you've
been given is "64 * 512" is actually "63 * 512" as far as the large drive
placeholder values of geometry are concerned. The physical disk does not have
63 actual sectors per track, but some other value. And that value, may not
even be stated in the available information for the hard drive.

To prove the point, if I use PartInNT to look at the "total sectors" in my C:
partition on my 80GB drive, the sector count is 156280257. That number is
evenly divisible by 63, not 64. 63 * 2480639 = 156280257, so there are
2480639 tracks in my C partition, exactly. Thus the starting sector number
of 63, is not an error, and not something to fool with. In my opinion.

I think the information you've been given, might have been valid
20 years ago, but not now. But I'm not an expert. You should visit
comp.sys.ibm.pc.hardware.storage and ask someone there to review
the advice you've been given and give comments. I think SAP is putting
you to a lot of trouble for nothing.

Paul
.



Relevant Pages

  • Re: grub error 18
    ... because I thought the EXTEND partition would be the second partition in ... Selected cylinder exceeds maximum supported by BIOS." ... the disk using CHS addressing (that is: telling the BIOS a head number, ... sectors per track gave 32k bytes per track, ...
    (uk.comp.homebuilt)
  • Re: BAD SECTORS ERROR
    ... > It says it has 8 bad sectors on it. ... > primary partition c:\ drive to correct this error. ... > 22368404 KB available on disk. ... > I removed the Hardrive from the 1st server and placed it ...
    (microsoft.public.win2000.general)
  • Re: today the ext3 on my fc6 T43 crashes :(
    ... It recognize some errors, fix it, but the system is not usable. ... Partition 1 does not end on cylinder boundary. ... Sometimes this is indicative that the disk is beginning to fail - especially if more and more sectors show up as bad over time. ...
    (Fedora)
  • Re: How to align the hard disk
    ... I downloaded PartinNT and say the disk ... I have 60GB HDD and I have two partition. ... disk geometry. ... the declaration is 63 sectors per track, ...
    (microsoft.public.windowsxp.hardware)
  • Re: Linux community software-update-anarchy polemic
    ... Remember, I'm just a monkey. ... That implies that if you have one disk per partition, ... Trust the kernel or don't trust the kernel, but either way, both ...
    (comp.os.linux.misc)