Optimal hard drive partitioning??

[Guest]

1. It’s mentioned anywhere that hd performance is better when magnetic head
is at the outer radios of the magnetic disk, but I never met numbers. Does
somebody know benchmark numbers for hard drive performance along magnetic
disks?

2. If there is serious difference in performance along a hard drive, there
is a sense to place operation system partition at the hard drive optimal
place. But I’m not sure where is this “place�?
I heard, but not sure, that an ordinary hard drive has two magnetic disks
(two sided both, I believe). If cylinder (sectors) numbering is straight
from the first disk inner radius to outer radius, and back on the other side,
and repeated again on the second disk, the disk performance should be
distributed along the disk as two sinusoidal half-waves, with two maximum at
¼ and ¾. If so, there is a sense to divide the disk for 5 partitions: two at
¼ and ¾ for operation systems (winxp and Linux in my case), and the disk
beginning, middle, and the end for storage. Is it correct, or cylinder
numbering has other order??

3. After disk defragmenter I see one blue bunch of files at the partition
beginning, next big free gap, and a blue bunch of file at the second part of
the partition (with unmovable green pagefile somewhere between). What type of
files in the second bunch? I believed that winxp moves less used files to
the partition end. But may be (only may be) the “clever†winxp moves most
frequently used files to place with most effective disk performance (as in my
disk)?? What do you think about this?

I’ll appreciate any information.
Best, Alex

 

[Carey Frisch [MVP]]

Please read the following:

Benchmarking on Windows XP
http://www.microsoft.com/whdc/system/sysperf/benchmark.mspx

NTFS Preinstallation and Windows XP
http://www.microsoft.com/whdc/system/winpreinst/ntfs-preinstall.mspx

<snip>

Microsoft implemented certain disk-layout optimizations in Windows XP.
To perform this optimization, during idle time Windows XP moves pages
used for booting the system and launching frequently used applications to
ensure these files are laid out contiguously on the hard disk. The contiguous
disk layout of these pages results in reduced disk seeks and improved disk I/O,
contributing to improved boot time and application launch time.

Windows XP does not perform these optimizations across volumes. Therefore,
for this optimization to be available to users, the hard disk must be partitioned
as a single volume.

<end of snip>

--
Carey Frisch
Microsoft MVP
Windows XP - Shell/User
Microsoft Newsgroups

Be Smart! Protect Your PC!
http://www.microsoft.com/athome/security/protect/default.mspx

------------------------------------------------------------------------------

:

| 1. It’s mentioned anywhere that hd performance is better when magnetic head
| is at the outer radios of the magnetic disk, but I never met numbers. Does
| somebody know benchmark numbers for hard drive performance along magnetic
| disks?
|
| 2. If there is serious difference in performance along a hard drive, there
| is a sense to place operation system partition at the hard drive optimal
| place. But I’m not sure where is this “place�?
| I heard, but not sure, that an ordinary hard drive has two magnetic disks
| (two sided both, I believe). If cylinder (sectors) numbering is straight
| from the first disk inner radius to outer radius, and back on the other side,
| and repeated again on the second disk, the disk performance should be
| distributed along the disk as two sinusoidal half-waves, with two maximum at
| ¼ and ¾. If so, there is a sense to divide the disk for 5 partitions: two at
| ¼ and ¾ for operation systems (winxp and Linux in my case), and the disk
| beginning, middle, and the end for storage. Is it correct, or cylinder
| numbering has other order??
|
| 3. After disk defragmenter I see one blue bunch of files at the partition
| beginning, next big free gap, and a blue bunch of file at the second part of
| the partition (with unmovable green pagefile somewhere between). What type of
| files in the second bunch? I believed that winxp moves less used files to
| the partition end. But may be (only may be) the “clever†winxp moves most
| frequently used files to place with most effective disk performance (as in my
| disk)?? What do you think about this?
|
| I’ll appreciate any information.
| Best, Alex

 

[James Hahn]

1. If the drive implements zoned bit recording, then the data on the outer
tracks is not only read more rapidly, but there is more data per track so
fewer head seeks are required for a given amount of data. The disk drive
manufacturers publish some of this data:
http://www.storagereview.com/guide2000/ref/hdd/perf/perf/spec/transMedia.html

2. Seek times are much more significant than any difference in internal data
transfer rates, so relative location of files is much more significant than
absolute location. Some optimizing to take advantage of the increased speed
of the outer tracks is possible, but only at the operating system level, so
that both issues are considered simultaneously.

 

[Guest]

Microsoft implemented certain disk-layout optimizations in Windows XP.
To perform this optimization, during idle time Windows XP moves pages
used for booting the system and launching frequently used applications to
ensure these files are laid out contiguously on the hard disk. The contiguous
disk layout of these pages results in reduced disk seeks and improved disk I/O,
contributing to improved boot time and application launch time.

Windows XP does not perform these optimizations across volumes. Therefore,
for this optimization to be available to users, the hard disk must be partitioned
as a single volume.

This is very interesting information. Before reading your post, I just
assumed that it was an old wive's tale among some people that partitioning a
physical hard drive helped performance, but no one had any actual evidence to
support the claim. So much for that.

Even leaving aside this point, I think it is also important to remember that
HD access time is most important when code is first loaded into RAM from the
HD. My understanding is that the part of this process attributable to the
access arm of the HD moving along the disk involves fractions of seconds, or
at worst a handful of seconds if the files are very large or the drive is
heavily fragmented (which it shouldn't be). Once in RAM, these files
essentially resides in the much faster system cache anyway -- and the more
memory a computer has, the longer it stays there, until the system is
rebooted.

Partitioning is good if you have only a single physical hard drive and want
to keep your data separate program and system files (but even better for this
purpose is a second physical hard drive, such as an external USB drive), but
I am unaware of a single shred of evidence that a NTFS volume with more
thanone partition outperforms a NTFS volume with a single large partition
(FAT32 volumes may be a different story, but I digress). Now you are
pointing out that the opposite, if anything, is much closer to the truth --
for best hard drive NTFS performance, use a single partition regardless of
the size of the disk. And I haven't even mentioned the enormous time that
gets lost when an attempt to partition a hard drive goes bad.

Ken

 

[Guest]

Please read the following:

Third times you simply copy me part of your sites without answering for my
questions; and I post mainly waiting help from guy such as James Hahn
(thanks) who gave me the right reference and now everything is clear for me.
Before I was not familiar with hd design and made silly suggestion, but now I
know what to do with my partitions. For case if somebody will read this post,
I place below information I looked for:
In fact, some demanding customers used to partition hard disks and use only
a small portion of the disk, for exactly this reason: so that seeks would be
faster.
For example, the IBM Deskstar 34GXP (model DPTA-373420) has a media transfer
rate of between approximately 171 Mb/s and 284 Mb/s depending where on the
disk you are reading: that drive has 12 different zones. This drive has 272
sectors in its innermost zone, and 452 sectors on its outside tracks.

Partition Placement: Most hard disks use zoned bit recording , which means
they hold more data per track at the outermost edge of the disk than they do
at the innermost edge; as a result, the outer tracks tend to deliver better
performance than the inner tracks do. Since the outer tracks are used first,
this means that the first partition on a physical disk volume will be
slightly faster than subsequent ones. If you have certain files that require
higher performance than others, placing them in a partition at the beginning
of the disk is preferred.
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

Best, Alex

 

[Alex Nichol]

1. It’s mentioned anywhere that hd performance is better when magnetic head
is at the outer radios of the magnetic disk, but I never met numbers. Does
somebody know benchmark numbers for hard drive performance along magnetic
disks?

That is obsolete advice. These days disks are 'zoned' with varying
density of data according to zone, and there is little difference. The
overall rotation speed is some guide to disk speed - but the big factor
in disk access is seek time getting to the right track

 

[James Hahn]

snip <
That is obsolete advice. These days disks are 'zoned' with varying
density of data according to zone, and there is little difference. The
overall rotation speed is some guide to disk speed - but the big factor
in disk access is seek time getting to the right track

It's the other way around. If the drive doesn't implement zoned bit
recording then the data transfer rate on all tracks is the same, so the only
criteria for file placement is seek times. When zoned bit recording is
implemented (as it is for most modern drives) then there are many more
sectors per track on the outer tracks, and as each track is still read in
the same number of revolutions (often, one revolution) then the data
transfer rate for the outer tracks is much higher than for inner tracks.
This means that physical placement of files now has to be added into the
considerations, complicating the process considerably.

A zoned disk actually has _less_ variation in storage density than an
unzoned disk.

 

[Guest]

That is obsolete advice. These days disks are 'zoned' with varying
density of data according to zone, and there is little difference. The
overall rotation speed is some guide to disk speed - but the big factor
in disk access is seek time getting to the right track

Please, look at this:
( http://www.storagereview.com/guide2000/ref/hdd/geom/tracksZBR.html ),
and compare MS MVP and hd professional site advices. Whom should I believe
to??
Simple logic says that there is no sense to decrease storage density when
everybody figt for density increasing.
Comment. please,
Best, Alex

: