Anger about convertntfs - suggestions and discussion desired

[Guest]

I built a network based unattended installation of Windows 2003 Server.
Everything works fine...DOS boot, create FAT-partition, connect network,
start setup. The disk is converted to NTFS during the setup in addition with
the ExtendOEMPartition- switch.

But: Everybody disadvises me strongly to use the convertntfs-switch, because
this results in a fragmented, not optimal clustered disk with low
performance. You find hundreds of articles about this "feature" in several
forums in the internet.

So what to do now?

- prepare the disk in another server with an adequate NTFS-partition and
then put
it back into the new server? This cannot be a solution for the setup of a
modern OS.
And what should one do, who only has one server?

- start a "normal" setup from Windows2003-CD and prepare the partition; then
cancel this setup and start the DOS-boot-installation? Again: This cannot be
a solution for the setup of a modern OS - and it wastes time

- format the NTFS-partition from my DOS-bootdisk? I haven't found any tools,
which could do that

So - any other suggestions? Any statements from the Microsoft guys, who
possibly read this?

I need to have an acceptable solution because my servers need to be high
performance and I do not pay hundreds of dollars for a suboptimal operating
system.

 

[Gerry Hickman]

Hi Ulrich,

But: Everybody disadvises me strongly to use the convertntfs-switch, because
this results in a fragmented, not optimal clustered disk with low
performance. You find hundreds of articles about this "feature" in several
forums in the internet.

I didn't know about this. Is it only Win2003 or all versions of NTFS
windows? Can you point me to any difinitive articles?

 

[Guest]

Hi Gerry,

the most articles I have are in German...here I found an article in English,
which describes one issue of convertntfs: The cluster size is to small
http://groups.google.com/groups?q=ntfs+convert&hl=de&lr=&[email protected]&rnum=8

The other issue is, that the master file table becomes fragmented during the
convert. I don't know, if these things have become better under Windows
2003...a statement from Microsoft and recommendations for best practice would
be useful.

If you start the defrag tool after an unattended setup of a Window
2003-Server, which had run through the convert ntfs during the installation,
you can definetely see the fragmentation.

So again the question directed to Microsoft: How should we deal with this
problem?

 

[Gerry Hickman]

Hi Ulrich,

the most articles I have are in German...here I found an article in English,
which describes one issue of convertntfs: The cluster size is to small
http://groups.google.com/groups?q=ntfs+convert&hl=de&lr=&[email protected]&rnum=8

OK I see it, they imply it's fixed in XP? I have dozens of Win2k
machines that were built using conversion so I can have a look at them.

How can I get a reading of my cluster size?

 

[Guest]

Hi Gerry,

if you run the defragmentation tool in your computer management and view the
analyze report, you can see the cluster size there. You should also see some
red areas in the graphic, if your harddisk is fragmented. Running a
defragmentation does not defragment the master file table, because it is in
use.

Greetings from Hamburg / Germany
Ulrich

 

[Gerry Hickman]

if you run the defragmentation tool in your computer management and view the
analyze report, you can see the cluster size there.

OK I see it, is there any way to get it at the command prompt?

You should also see some
red areas in the graphic, if your harddisk is fragmented. Running a
defragmentation does not defragment the master file table, because it is in
use.

OK, checking my home computer I see the C: drive is 512bytes, but data
drives are 4kb. The C: drive was created by ConvertNTFS and the data
drives were created manually from within Windows after I built the
machine. It probably means all my machines at work are the same, even
the production servers! C: drives tend to have a lot of small files and
data drives tend to have bigger files. I wonder what the implications are?

Thing is, all these computers run faster and better than any other
computers I've ever sat in front of, so I've never really thought
anything could be wrong. I'm not so concerned now, that it appears to
affect only C: drives - if it was the data drives on the big RAID
subsystems it would be a serious problem.

 

[Guest]

Of course you can discuss about wether this is a problem or not. Fact is,
that the cluster size is not the optimal size. which is recommended by
Microsoft. If this affects the performance or not requires some extended
testing, I think.

What my main problem is: I am responsible for a large number of servers,
which have been roled out with convertntfs. If one of these servers has a
performance problem, I want to be on the "save side" against the people, who
I have to report to and be able to say to them: All these servers have been
set up in a way, that is fully supported by Microsoft.

So what I still hope is, that someone from Microsoft reads this topic and
tells me, that I am on the right way and that the small cluster size and the
fragmented MFT are no problem at all.

Regards
Ulrich

 

[Gerry Hickman]

Hi Ulrich,

Of course you can discuss about wether this is a problem or not. Fact is,
that the cluster size is not the optimal size. which is recommended by
Microsoft. If this affects the performance or not requires some extended
testing, I think.

Can you point me to where Microsoft recommends this size?

What my main problem is: I am responsible for a large number of servers,
which have been roled out with convertntfs.

I agree, I'm in the same boat.

So what I still hope is, that someone from Microsoft reads this topic

Unfortunately, Microsoft don't respond in this group. It's also not a
"managed group". The only hope of getting a response from Microsoft is
either to use PSS (if you have it), or to try one of the "managed"
groups via MSDN subscription.

 

[Gerry Hickman]

Unfortunately, Microsoft don't respond in this group.

Oops, spoke to soon! I've just seen a post in another thread from
Rebecca Chen!

 

[Guest]

Hi Gerry,

if you owe a technet, you can find an article called "Optimizing NTFS",
which deals with the problems of this topic.

The essential statements in this article are: Converted NTFS-volumes are
much slower than NTFS volumes, you create from scratch. The only suggestion
from Microsoft for optimzing these volumes: "Religiously" run commercial
disk-defragmentation utilities.

Greetings
Ulrich

 

[Patrick J. LoPresti]

(Sorry for the delay; I am finally get caught up on this newsgroup.)

Hi, Ulrich. Microsoft addresses this issue here:

http://www.microsoft.com/whdc/system/winpreinst/ntfs-preinstall.mspx

The summary is that you can use the "oformat" and "cvtarea" tools to
prepare the FAT partition for conversion to an efficient NTFS
implementation.

I use this approach in my own system, and I can confirm that it
produces 4K clusters both in Windows XP and Server 2003. See also
<http://unattended.sourceforge.net/faq.php#5_3>.

- Pat