Horribly slow disk access

[Guest]

Hi all,

Scenario: We are currently consolidating 130 fileservers (also acting dc’s),
to a 2-node Windows 2000 file cluster (SP4). On all 130 fileservers, the
folder structure is identical; three folders in the root hierarchy named
“Usersâ€, “Common†and “Projectsâ€.

“Users†contain all home folders, which only the specific user and
administrators has access to. The security on “Common†is set to Change for
all users. Finally, “Projects†contain several sub folders with different
security lists on each folder. There are usually 10-30 users on these lists,
which have read, write or change access to the folders.

Now the problem: After the consolidation, users have no problems accessing
their home folders or the Common folder. But when they are accessing the
Projects folder (or any sub folder) they are experiencing horribly slow disk
access! It may take up to 45 seconds to open just a normal word-file.

What can cause this problem? The only thing I can think about is the file
security which is used very frequently in the “Projects†folder. Hardware is
not an issue, nor is simultaneous users.

Cheers,
Johan Ã…kerlund

 

[Andrew Hodes]

Have you checked the event log for any clues? This seems quite odd... is
the projects folder on the same partition as the rest of the folders?

 

[wanderer]

Not sure why you would think hardware or
users would not be a issue with
consolidating 130 servers
into a single two node cluster.

I can see your setup just getting bogge
down in authenications let alone
serving up resources.

Might want to run some performance monitoring and check out your usage
on a number of fronts.

 

[Jobro]

Not sure why you would think hardware or
users would not be a issue with
consolidating 130 servers
into a single two node cluster.

I can see your setup just getting bogge
down in authenications let alone
serving up resources.

Might want to run some performance monitoring and check out your usage
on a number of fronts.

[/QUOTE]

Hi again,

Thanks for your replies. We solved the problem by copying all data in
the Projects folder to a temporary server. Then we checked the disk for
errors (we didn't find any), and copied all data back to the cluster
again. Performance is now better then imaginable. Sometimes there are
odd solutions to odd problems

Wanderer, before such a consolidation we did of course estimate a
baseline of what hardware we should use. The cluster servers are
actually over dimensioned for this job. We continuously measure the
performance on both cluster nodes, and none are heavily used. The
number of simultaneous users never exceeds 2500. Hence, hardware or no
of users was not the issue in this case.

Take care,
Johan

 

[wanderer]

Sounds like you did an excellent job on your hardware consolidation
considerations. I am still boggled why coping off and back would speed
up access but am glad it worked for you. I wonder if it was a
fragmentation problem.