FYI: NTFS Cluster Size /3rd Party Disk Mgmt Tools

[R. McCarty]

Here is some information on Service Pack 2 for XP, related
to Disk partitioning and Cluster sizes. This is complicated, but
I'll try and describe this as best as I can.

This came to light when doing a SP2 "Repair Install". The target
system used a cluster size greater than the default value of 4-K.
The repair install ran successfully. However, at the first reboot
the system halted with "Disk Read Error" (Black screen). After
some investigation, I used Partition Magic 8.01 to resize the XP
partition back to 4-K Clusters. XP then booted normally.

Next I restored the Sp1a image to the PC. As a test case, I used
an Acronis Disk Management tool to resize the partition clusters
up to 64-K (Largest). I then installed SP2 but in a standard update
mode. System booted normally and used the 64-K clusters.
Then I did a "Repair" install on the partition. At reboot the Disk
read error re-appeared. I resized the clusters to 32,16,8-K all w/
the same Disk error. Finally, I resized back to 4-K and the
system booted normally.

I've verified this on 3 different systems. Apparently, there is some
change to NTFS Version 3.1 that prevents new installs from using
cluster size other than the default of 4-K. However, if the partition
is already ~+4-K XP will run after a SP2 update - but not if a in
place or Repair install (Sp2) is used.

I've tested this with Partition Magic, Acronis & BootIT-NG so the
changes are not specific to any one 3rd party disk mgmt tool.

If anyone has any input I would appreciate it. I've contacted MS
but haven't obtained any useful feedback or acknowledgement that
NTFS has changed with Service Pack 2.

 

[SlowJet]

The 4K is the only cluster size that works with Folder /
Folie compression on NTFS including Encrytion.

Maybe something is compressed in the repair process of
the CD?

SJ

SO, did you do any read bench marks while you were at
it?

SJ

 

[SlowJet]

There could a few fatctoers involved in that problem.

The MFT and overallocated files (before and after OA is
released?


Interesting,

SJ

 

[Alex Nichol]

I've verified this on 3 different systems. Apparently, there is some
change to NTFS Version 3.1 that prevents new installs from using
cluster size other than the default of 4-K. However, if the partition
is already ~+4-K XP will run after a SP2 update - but not if a in
place or Repair install (Sp2) is used.

I would strongly advise not using a cluster size other than 4K on the
partition where the system resides, or on any other that has a page file
on it. This is important for efficiency; as that cluster matches the
internal hardware page used by the Intel-architecture paging hardware,
program loads and page operations can then be performed directly between
RAM and disk without need for additional buffering

Very large clusters should be reserved for separate partitions used
exclusively for very large data files, eg movie editing

 

[SlowJet]

This is important for efficiency; as that cluster
matches the
internal hardware page used by the Intel-architecture
paging hardware,
program loads and page operations can then be performed
directly between
RAM and disk without need for additional buffering<<<<

My gut tells me that's not quit true.
I humbbly disagree.

SJ

 

[Alex Nichol]

This is important for efficiency; as that cluster
matches the
internal hardware page used by the Intel-architecture
paging hardware,
program loads and page operations can then be performed
directly between
RAM and disk without need for additional buffering<<<<

My gut tells me that's not quit true.
I humbbly disagree.

Your gut may, but a senior developer in Redmond found no cause to
disagree

This direct transfer between RAM and 'aligned' code files came in in
Windows 98