| Author |
Message |
Jim Mack
Guest
|
 Posted:
Wed Oct 12, 2005 3:35 pm Post subject:
USB Drives, NFS performance |
|
|
Hi All -
We're evaluating a 6TB Dell box with WSS2k3 as a solution to replace several Snap servers, and a couple of questions have come up that I hope someone can shed light on.
This is a mixed Windows / Irix (SGI Unix) graphics environment. NFS performance between Irix and non-Irix NFS devices has been very poor in our experience. Using SFU1 and SFU2 in the past didn't give great results -- can anyone speak about SFNFS and how it performs relative to SFU2, and especially wrt Irix?
Second, we often exchange large numbers of large files using USB 'shuttle' drives of 200GB or so. Assuming that the Dell box can handle it, is there any reason why we can't use these directly in WSS2k3? Is FAT32 an acceptable FS for use in this way? We don't have to share these out directly (though we'd like to), but at least we have to be able to copy the files to a share. Is administration of plug/unplug + mount/unmount done in the web interface?
And this may be a naive question because I don't yet understand what VSS can do, but is it possible to keep VSS snapshots on the same sorts of USB drives?
Thanks for any info or insights.
--
Jim |
|
| Back to top |
|
 |
Eric Bursley [MVP]
Guest
|
| Posted:
Thu Oct 13, 2005 7:05 am Post subject:
Re: USB Drives, NFS performance |
|
|
There are tweaks that can be used to improve NFS performance. Windows
Storage Server 2003 that comes on the Powervault 745 includes Service for
NFS 3.0.
This document shows some tweak that can be used:
http://www.microsoft.com/technet/interopmigration/unix/sfu/perfnfs.mspx#EGAA
The PV 745 does accept a USB disk, in fact comes with a 64MB USB disk for
reimaging the appliance if needed.
FAT / FAT32 is typically fine for USB disks, but larger disks can support
NTFS, and may be a consideration for security purposes in some environments.
The web interface allows for the creation and removal of shares. As well as
configuring Service for NFS for authenication.
Eric Bursley
eric at bursley dot net
Microsoft MVP
RHCE, MCSE, BCFP, EEIE-CS, ESCE-CS
GPG Signature:
Key fingerprint = CEAE CF3A 3876 7ECE 9DA7 946F DA9F DDCA C392 6DCB
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:OFWKkixzFHA.2540@TK2MSFTNGP09.phx.gbl...
Hi All -
We're evaluating a 6TB Dell box with WSS2k3 as a solution to replace several
Snap servers, and a couple of questions have come up that I hope someone can
shed light on.
This is a mixed Windows / Irix (SGI Unix) graphics environment. NFS
performance between Irix and non-Irix NFS devices has been very poor in our
experience. Using SFU1 and SFU2 in the past didn't give great results -- can
anyone speak about SFNFS and how it performs relative to SFU2, and
especially wrt Irix?
Second, we often exchange large numbers of large files using USB 'shuttle'
drives of 200GB or so. Assuming that the Dell box can handle it, is there
any reason why we can't use these directly in WSS2k3? Is FAT32 an
acceptable FS for use in this way? We don't have to share these out
directly (though we'd like to), but at least we have to be able to copy the
files to a share. Is administration of plug/unplug + mount/unmount done in
the web interface?
And this may be a naive question because I don't yet understand what VSS can
do, but is it possible to keep VSS snapshots on the same sorts of USB
drives?
Thanks for any info or insights.
--
Jim |
|
| Back to top |
|
|
Jim Mack
Guest
|
| Posted:
Thu Oct 13, 2005 4:16 pm Post subject:
Re: USB Drives, NFS performance |
|
|
Eric Bursley [MVP] wrote:
| Quote: | There are tweaks that can be used to improve NFS performance. Windows
Storage Server 2003 that comes on the Powervault 745 includes Service
for NFS 3.0.
This document shows some tweak that can be used:
http://www.microsoft.com/technet/interopmigration/unix/sfu/perfnfs.mspx#EGAA
The PV 745 does accept a USB disk, in fact comes with a 64MB USB disk
for reimaging the appliance if needed.
FAT / FAT32 is typically fine for USB disks, but larger disks can
support NTFS, and may be a consideration for security purposes in
some environments.
The web interface allows for the creation and removal of shares. As
well as configuring Service for NFS for authenication.
|
Thanks for replying, good info on NFS. Looks like they made progress since SFU 2.0.
To summarize and extrapolate a bit from your answer -- I should be able to receive a USB disk formatted by another facility as FAT32, plug it in to the 745, share it via NFS, then later remove the share and unplug the disk, all via the web interface... correct? If so that's fantastic, just what we need.
--
Jim |
|
| Back to top |
|
|
Pat [MSFT]
Guest
|
| Posted:
Fri Oct 14, 2005 5:33 am Post subject:
Re: USB Drives, NFS performance |
|
|
Be very, very careful.
NFS (v2 & v3, not sure about v4, but I don't think so) does not have a
protocol command to do a cache flush and it is possible that data may not be
written before you do a removal of the disk (the lag can be quite a while
depending on usage scenarios). SMB does have a Cache_Flush command in the
protocol - and in fact the command can be disabled via the registry on the
server for certain scenarios.
Also IDE (which is probably the interface to the disk itself) does not
(always) respect the Write_Through command - though the disk may report to
the OS that it does. So your data that was flushed from the OS cache may be
sitting on the IDE drive cache. When the power goes away...
So you have 2 potential areas of corruption/data loss. If you want to use
this, you should always do a safe removal - not just pull the plug. That
will at least force a flush action and minimize the risk - but won't force
the interface to support the Write_Through command. If you wait at least 10
seconds (after the flush) you should have a very high probability (but no
guarantee) that the data you expect to be there will be.
It sounds like your scenario calls for a relatively rarely used disk (or is
primarily read only) so you will probably be OK, but just keep in mind what
the limitations of the protocol and interface are. If you must guarantee
the data writes, then you should look into a SCSI based device.
Pat
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:uvIVYt%23zFHA.1264@tk2msftngp13.phx.gbl...
Eric Bursley [MVP] wrote:
| Quote: | There are tweaks that can be used to improve NFS performance. Windows
Storage Server 2003 that comes on the Powervault 745 includes Service
for NFS 3.0.
This document shows some tweak that can be used:
http://www.microsoft.com/technet/interopmigration/unix/sfu/perfnfs.mspx#EGAA
The PV 745 does accept a USB disk, in fact comes with a 64MB USB disk
for reimaging the appliance if needed.
FAT / FAT32 is typically fine for USB disks, but larger disks can
support NTFS, and may be a consideration for security purposes in
some environments.
The web interface allows for the creation and removal of shares. As
well as configuring Service for NFS for authenication.
|
Thanks for replying, good info on NFS. Looks like they made progress since
SFU 2.0.
To summarize and extrapolate a bit from your answer -- I should be able to
receive a USB disk formatted by another facility as FAT32, plug it in to the
745, share it via NFS, then later remove the share and unplug the disk, all
via the web interface... correct? If so that's fantastic, just what we
need.
--
Jim |
|
| Back to top |
|
|
Jim Mack
Guest
|
| Posted:
Fri Oct 14, 2005 6:05 am Post subject:
Re: USB Drives, NFS performance |
|
|
Pat [MSFT] wrote:
| Quote: | Be very, very careful.
NFS (v2 & v3, not sure about v4, but I don't think so) does not have a
protocol command to do a cache flush and it is possible that data may
not be written before you do a removal of the disk (the lag can be
quite a while depending on usage scenarios). SMB does have a
Cache_Flush command in the protocol - and in fact the command can be
disabled via the registry on the server for certain scenarios.
Also IDE (which is probably the interface to the disk itself) does not
(always) respect the Write_Through command - though the disk may
report to the OS that it does. So your data that was flushed from
the OS cache may be sitting on the IDE drive cache. When the power
goes away...
So you have 2 potential areas of corruption/data loss. If you want
to use this, you should always do a safe removal - not just pull the
plug. That will at least force a flush action and minimize the risk
- but won't force the interface to support the Write_Through command.
If you wait at least 10 seconds (after the flush) you should have a
very high probability (but no guarantee) that the data you expect to
be there will be.
It sounds like your scenario calls for a relatively rarely used disk
(or is primarily read only) so you will probably be OK, but just keep
in mind what the limitations of the protocol and interface are. If
you must guarantee the data writes, then you should look into a SCSI
based device.
Pat
|
Thanks for the warnings. I definitely wouldn't remove a disk by just unplugging, and never until there had been no activity for a few minutes. As long as there's an interface available to do it we'll be unsharing the drive, then requesting a safe removal. And we wouldn't do any of that until after the file system had been umounted on the Irix side.
As you suspect, this is a read-only situation, so we should be safe from write cache issues at least. We really have no choice but to use an IDE-based USB drive, formatted FAT32, because that's the common denominator among graphics houses and platforms.
--
Jim
| Quote: |
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:uvIVYt%23zFHA.1264@tk2msftngp13.phx.gbl...
Eric Bursley [MVP] wrote:
There are tweaks that can be used to improve NFS performance.
Windows Storage Server 2003 that comes on the Powervault 745
includes Service for NFS 3.0.
This document shows some tweak that can be used:
http://www.microsoft.com/technet/interopmigration/unix/sfu/perfnfs.mspx#EGAA
The PV 745 does accept a USB disk, in fact comes with a 64MB USB disk
for reimaging the appliance if needed.
FAT / FAT32 is typically fine for USB disks, but larger disks can
support NTFS, and may be a consideration for security purposes in
some environments.
The web interface allows for the creation and removal of shares. As
well as configuring Service for NFS for authenication.
Thanks for replying, good info on NFS. Looks like they made progress
since SFU 2.0.
To summarize and extrapolate a bit from your answer -- I should be
able to receive a USB disk formatted by another facility as FAT32,
plug it in to the 745, share it via NFS, then later remove the share
and unplug the disk, all via the web interface... correct? If so
that's fantastic, just what we need. |
|
|
| Back to top |
|
|
Eric Bursley [MVP]
Guest
|
| Posted:
Fri Oct 14, 2005 6:32 am Post subject:
Re: USB Drives, NFS performance |
|
|
NFS requires the use of NTFS on the file system to export in SFU. This is
because of the permissions held on the file system. FAT32 as you know
doesn't hold these permissions. I didn't realize you wanted to export the
USB disk via NFS.
Eric Bursley
eric at bursley dot net
Microsoft MVP
RHCE, MCSE, BCFP, EEIE-CS, ESCE-CS
GPG Signature:
Key fingerprint = CEAE CF3A 3876 7ECE 9DA7 946F DA9F DDCA C392 6DCB
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:uvIVYt%23zFHA.1264@tk2msftngp13.phx.gbl...
Eric Bursley [MVP] wrote:
| Quote: | There are tweaks that can be used to improve NFS performance. Windows
Storage Server 2003 that comes on the Powervault 745 includes Service
for NFS 3.0.
This document shows some tweak that can be used:
http://www.microsoft.com/technet/interopmigration/unix/sfu/perfnfs.mspx#EGAA
The PV 745 does accept a USB disk, in fact comes with a 64MB USB disk
for reimaging the appliance if needed.
FAT / FAT32 is typically fine for USB disks, but larger disks can
support NTFS, and may be a consideration for security purposes in
some environments.
The web interface allows for the creation and removal of shares. As
well as configuring Service for NFS for authenication.
|
Thanks for replying, good info on NFS. Looks like they made progress since
SFU 2.0.
To summarize and extrapolate a bit from your answer -- I should be able to
receive a USB disk formatted by another facility as FAT32, plug it in to the
745, share it via NFS, then later remove the share and unplug the disk, all
via the web interface... correct? If so that's fantastic, just what we
need.
--
Jim |
|
| Back to top |
|
|
Eric Bursley [MVP]
Guest
|
| Posted:
Fri Oct 14, 2005 6:33 am Post subject:
Re: USB Drives, NFS performance |
|
|
Speaking of NFS v4, when does Microsoft plan on implementing it? I'm
presuming this would require joining the UNIX / Linux host to the Kerberos
Realm in Active Directory.
Eric Bursley
eric at bursley dot net
Microsoft MVP
RHCE, MCSE, BCFP, EEIE-CS, ESCE-CS
GPG Signature:
Key fingerprint = CEAE CF3A 3876 7ECE 9DA7 946F DA9F DDCA C392 6DCB
"Pat [MSFT]" <patfilot@online.microsoft.com> wrote in message
news:uYRRcbF0FHA.2652@TK2MSFTNGP14.phx.gbl...
| Quote: | Be very, very careful.
NFS (v2 & v3, not sure about v4, but I don't think so) does not have a
protocol command to do a cache flush and it is possible that data may not
be written before you do a removal of the disk (the lag can be quite a
while depending on usage scenarios). SMB does have a Cache_Flush command
in the protocol - and in fact the command can be disabled via the registry
on the server for certain scenarios.
Also IDE (which is probably the interface to the disk itself) does not
(always) respect the Write_Through command - though the disk may report to
the OS that it does. So your data that was flushed from the OS cache may
be sitting on the IDE drive cache. When the power goes away...
So you have 2 potential areas of corruption/data loss. If you want to use
this, you should always do a safe removal - not just pull the plug. That
will at least force a flush action and minimize the risk - but won't force
the interface to support the Write_Through command. If you wait at least
10 seconds (after the flush) you should have a very high probability (but
no guarantee) that the data you expect to be there will be.
It sounds like your scenario calls for a relatively rarely used disk (or
is primarily read only) so you will probably be OK, but just keep in mind
what the limitations of the protocol and interface are. If you must
guarantee the data writes, then you should look into a SCSI based device.
Pat
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:uvIVYt%23zFHA.1264@tk2msftngp13.phx.gbl...
Eric Bursley [MVP] wrote:
There are tweaks that can be used to improve NFS performance. Windows
Storage Server 2003 that comes on the Powervault 745 includes Service
for NFS 3.0.
This document shows some tweak that can be used:
http://www.microsoft.com/technet/interopmigration/unix/sfu/perfnfs.mspx#EGAA
The PV 745 does accept a USB disk, in fact comes with a 64MB USB disk
for reimaging the appliance if needed.
FAT / FAT32 is typically fine for USB disks, but larger disks can
support NTFS, and may be a consideration for security purposes in
some environments.
The web interface allows for the creation and removal of shares. As
well as configuring Service for NFS for authenication.
Thanks for replying, good info on NFS. Looks like they made progress since
SFU 2.0.
To summarize and extrapolate a bit from your answer -- I should be able to
receive a USB disk formatted by another facility as FAT32, plug it in to
the 745, share it via NFS, then later remove the share and unplug the
disk, all via the web interface... correct? If so that's fantastic, just
what we need.
--
Jim
|
|
|
| Back to top |
|
|
Jim Mack
Guest
|
| Posted:
Fri Oct 14, 2005 7:35 am Post subject:
Re: USB Drives, NFS performance |
|
|
Eric Bursley [MVP] wrote:
| Quote: | NFS requires the use of NTFS on the file system to export in SFU.
This is because of the permissions held on the file system. FAT32 as
you know doesn't hold these permissions. I didn't realize you wanted
to export the USB disk via NFS.
|
I though I'd read that, but earlier answers made it seem that it might not be so.
Well then, as I see it we have two options. The first would be to simply copy the files from the USB drive to a local NTFS share. The second would be to convert the disk to NTFS. Does WSS2k3 have the 'convert' utility?
Given that there could be dozens of gigabytes on the drive, in several thousand files, which do you suppose would be faster? My guess would be 'convert'.
--
Jim |
|
| Back to top |
|
|
Pat [MSFT]
Guest
|
| Posted:
Fri Oct 14, 2005 4:16 pm Post subject:
Re: USB Drives, NFS performance |
|
|
I can't speak to NFSv4 other than it won't be in the next release
(Win2k3R2), but there are a lot of updates to our NFS support coming with
the Windows Server 2k3-R2 release - available as a Release Candidate now
(http://www.microsoft.com/windowsserver2003/r2/default.mspx).
Highlights:
RFC2307 support (LDAP authentication)
NFS Boot
Improved Write performance - improvement is highly dependent on the NFS
client used - i.e. how strictly the client maintains sequential writes, so
Solaris (strict) clients won't see much (though some) whereas Fedora (very
loose) clients will see a significant improvement.
64bit support
Pat
"Eric Bursley [MVP]" <ebursley at swbell dot net> wrote in message
news:OTlVS9F0FHA.3780@TK2MSFTNGP12.phx.gbl...
| Quote: | Speaking of NFS v4, when does Microsoft plan on implementing it? I'm
presuming this would require joining the UNIX / Linux host to the Kerberos
Realm in Active Directory.
Eric Bursley
eric at bursley dot net
Microsoft MVP
RHCE, MCSE, BCFP, EEIE-CS, ESCE-CS
GPG Signature:
Key fingerprint = CEAE CF3A 3876 7ECE 9DA7 946F DA9F DDCA C392 6DCB
"Pat [MSFT]" <patfilot@online.microsoft.com> wrote in message
news:uYRRcbF0FHA.2652@TK2MSFTNGP14.phx.gbl...
Be very, very careful.
NFS (v2 & v3, not sure about v4, but I don't think so) does not have a
protocol command to do a cache flush and it is possible that data may not
be written before you do a removal of the disk (the lag can be quite a
while depending on usage scenarios). SMB does have a Cache_Flush command
in the protocol - and in fact the command can be disabled via the
registry on the server for certain scenarios.
Also IDE (which is probably the interface to the disk itself) does not
(always) respect the Write_Through command - though the disk may report
to the OS that it does. So your data that was flushed from the OS cache
may be sitting on the IDE drive cache. When the power goes away...
So you have 2 potential areas of corruption/data loss. If you want to
use this, you should always do a safe removal - not just pull the plug.
That will at least force a flush action and minimize the risk - but won't
force the interface to support the Write_Through command. If you wait at
least 10 seconds (after the flush) you should have a very high
probability (but no guarantee) that the data you expect to be there will
be.
It sounds like your scenario calls for a relatively rarely used disk (or
is primarily read only) so you will probably be OK, but just keep in mind
what the limitations of the protocol and interface are. If you must
guarantee the data writes, then you should look into a SCSI based device.
Pat
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:uvIVYt%23zFHA.1264@tk2msftngp13.phx.gbl...
Eric Bursley [MVP] wrote:
There are tweaks that can be used to improve NFS performance. Windows
Storage Server 2003 that comes on the Powervault 745 includes Service
for NFS 3.0.
This document shows some tweak that can be used:
http://www.microsoft.com/technet/interopmigration/unix/sfu/perfnfs.mspx#EGAA
The PV 745 does accept a USB disk, in fact comes with a 64MB USB disk
for reimaging the appliance if needed.
FAT / FAT32 is typically fine for USB disks, but larger disks can
support NTFS, and may be a consideration for security purposes in
some environments.
The web interface allows for the creation and removal of shares. As
well as configuring Service for NFS for authenication.
Thanks for replying, good info on NFS. Looks like they made progress
since SFU 2.0.
To summarize and extrapolate a bit from your answer -- I should be able
to receive a USB disk formatted by another facility as FAT32, plug it in
to the 745, share it via NFS, then later remove the share and unplug the
disk, all via the web interface... correct? If so that's fantastic, just
what we need.
--
Jim
|
|
|
| Back to top |
|
|
Eric Bursley [MVP]
Guest
|
| Posted:
Sat Oct 15, 2005 12:17 am Post subject:
Re: USB Drives, NFS performance |
|
|
I believe win2k3 does have the convert program, but if this is just a USB
drive, why not just reformat and restore the files?
Eric Bursley
eric at bursley dot net
Microsoft MVP
RHCE, MCSE, BCFP, EEIE-CS, ESCE-CS
GPG Signature:
Key fingerprint = CEAE CF3A 3876 7ECE 9DA7 946F DA9F DDCA C392 6DCB
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:%23XVi0fG0FHA.2312@TK2MSFTNGP14.phx.gbl...
Eric Bursley [MVP] wrote:
| Quote: | NFS requires the use of NTFS on the file system to export in SFU.
This is because of the permissions held on the file system. FAT32 as
you know doesn't hold these permissions. I didn't realize you wanted
to export the USB disk via NFS.
|
I though I'd read that, but earlier answers made it seem that it might not
be so.
Well then, as I see it we have two options. The first would be to simply
copy the files from the USB drive to a local NTFS share. The second would be
to convert the disk to NTFS. Does WSS2k3 have the 'convert' utility?
Given that there could be dozens of gigabytes on the drive, in several
thousand files, which do you suppose would be faster? My guess would be
'convert'.
--
Jim |
|
| Back to top |
|
|
Eric Bursley [MVP]
Guest
|
| Posted:
Sat Oct 15, 2005 12:17 am Post subject:
Re: USB Drives, NFS performance |
|
|
LDAP support would be great, because one of my biggest complaints in the IT
shops I support is configuring a NIS domain. Most large UNIX / Linux shops
use LDAP for authenication, and they don't want to configure NIS because of
security concerns.
Eric Bursley
eric at bursley dot net
Microsoft MVP RHCE, MCSE, BCFP, EEIE-CS, ESCE-CS
GPG Signature:
Key fingerprint = CEAE CF3A 3876 7ECE 9DA7 946F DA9F DDCA C392 6DCB
"Pat [MSFT]" <patfilot@online.microsoft.com> wrote in message
news:OADlfUM0FHA.2072@TK2MSFTNGP14.phx.gbl...
| Quote: | I can't speak to NFSv4 other than it won't be in the next release
(Win2k3R2), but there are a lot of updates to our NFS support coming with
the Windows Server 2k3-R2 release - available as a Release Candidate now
(http://www.microsoft.com/windowsserver2003/r2/default.mspx).
Highlights:
RFC2307 support (LDAP authentication)
NFS Boot
Improved Write performance - improvement is highly dependent on the NFS
client used - i.e. how strictly the client maintains sequential writes, so
Solaris (strict) clients won't see much (though some) whereas Fedora (very
loose) clients will see a significant improvement.
64bit support
Pat
"Eric Bursley [MVP]" <ebursley at swbell dot net> wrote in message
news:OTlVS9F0FHA.3780@TK2MSFTNGP12.phx.gbl...
Speaking of NFS v4, when does Microsoft plan on implementing it? I'm
presuming this would require joining the UNIX / Linux host to the
Kerberos Realm in Active Directory.
Eric Bursley
eric at bursley dot net
Microsoft MVP
RHCE, MCSE, BCFP, EEIE-CS, ESCE-CS
GPG Signature:
Key fingerprint = CEAE CF3A 3876 7ECE 9DA7 946F DA9F DDCA C392 6DCB
"Pat [MSFT]" <patfilot@online.microsoft.com> wrote in message
news:uYRRcbF0FHA.2652@TK2MSFTNGP14.phx.gbl...
Be very, very careful.
NFS (v2 & v3, not sure about v4, but I don't think so) does not have a
protocol command to do a cache flush and it is possible that data may
not be written before you do a removal of the disk (the lag can be quite
a while depending on usage scenarios). SMB does have a Cache_Flush
command in the protocol - and in fact the command can be disabled via
the registry on the server for certain scenarios.
Also IDE (which is probably the interface to the disk itself) does not
(always) respect the Write_Through command - though the disk may report
to the OS that it does. So your data that was flushed from the OS cache
may be sitting on the IDE drive cache. When the power goes away...
So you have 2 potential areas of corruption/data loss. If you want to
use this, you should always do a safe removal - not just pull the plug.
That will at least force a flush action and minimize the risk - but
won't force the interface to support the Write_Through command. If you
wait at least 10 seconds (after the flush) you should have a very high
probability (but no guarantee) that the data you expect to be there will
be.
It sounds like your scenario calls for a relatively rarely used disk (or
is primarily read only) so you will probably be OK, but just keep in
mind what the limitations of the protocol and interface are. If you
must guarantee the data writes, then you should look into a SCSI based
device.
Pat
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:uvIVYt%23zFHA.1264@tk2msftngp13.phx.gbl...
Eric Bursley [MVP] wrote:
There are tweaks that can be used to improve NFS performance. Windows
Storage Server 2003 that comes on the Powervault 745 includes Service
for NFS 3.0.
This document shows some tweak that can be used:
http://www.microsoft.com/technet/interopmigration/unix/sfu/perfnfs.mspx#EGAA
The PV 745 does accept a USB disk, in fact comes with a 64MB USB disk
for reimaging the appliance if needed.
FAT / FAT32 is typically fine for USB disks, but larger disks can
support NTFS, and may be a consideration for security purposes in
some environments.
The web interface allows for the creation and removal of shares. As
well as configuring Service for NFS for authenication.
Thanks for replying, good info on NFS. Looks like they made progress
since SFU 2.0.
To summarize and extrapolate a bit from your answer -- I should be able
to receive a USB disk formatted by another facility as FAT32, plug it in
to the 745, share it via NFS, then later remove the share and unplug the
disk, all via the web interface... correct? If so that's fantastic,
just what we need.
--
Jim
|
|
|
| Back to top |
|
|
Jim Mack
Guest
|
| Posted:
Sat Oct 15, 2005 7:08 am Post subject:
Re: USB Drives, NFS performance |
|
|
Eric Bursley [MVP] wrote:
| Quote: | I believe win2k3 does have the convert program, but if this is just a
USB drive, why not just reformat and restore the files?
|
The scenario is this: we send a blank, FAT32-formatted 250GB drive across the country, where 65mm film frames are scanned at very high resolution and copied to the drive. It must be FAT32 at this step, no exceptions.
When we receive the filled drive, we have to get the frames off of it and onto multiple SGI (Irix) systems. All these systems mount a common graphics server, which in the near future will be a PV745N, so this is where the frames are centralized.
Optimum workflow would have us copy these thousands of large files as little as possible, and get them into the Irix environment as quickly as possible.
Since we can't share the FAT32 drive directly via NFS, our choices seem to be as I outlined them: copy the files to a native NTFS share on the server, or convert the USB drive to NTFS so that it can be shared.
That's why I asked the question: which of those two scenarios would likely take less time -- copying the files from USB/FAT32 to SCSI/NTFS, or simply converting the USB drive in place?
I don't understand what other course you're suggesting, but if it involves more copying then it's suboptimal workflow. Given what I've described, if you still see another possibility I'd love to have it explained further.
Thanks
--
Jim |
|
| Back to top |
|
|
Eric Bursley [MVP]
Guest
|
| Posted:
Sun Oct 16, 2005 12:16 am Post subject:
Re: USB Drives, NFS performance |
|
|
Why not use a VPN over the Internet to copy the files between locations? It
seems to me that doing a VPN solution would be much faster than shipping a
USB drive cross country.
I can understand if one or both locations do not have Internet access, but
if you did have Internet access creating a secure VPN between both locations
would be easy, faster, and cheaper.
Without the Internet VPN as an option, your best bet would be to copy the
files from the USB disk onto the servers NTFS file system. The folder on
the NTFS file system can remain exported via NFS.
Eric Bursley
eric at bursley dot net
Microsoft MVP
RHCE, MCSE, BCFP, EEIE-CS, ESCE-CS
GPG Signature:
Key fingerprint = CEAE CF3A 3876 7ECE 9DA7 946F DA9F DDCA C392 6DCB
"Jim Mack" <jmack@mdxi.nospam.com> wrote in message
news:uqCyM1S0FHA.2924@TK2MSFTNGP15.phx.gbl...
Eric Bursley [MVP] wrote:
| Quote: | I believe win2k3 does have the convert program, but if this is just a
USB drive, why not just reformat and restore the files?
|
The scenario is this: we send a blank, FAT32-formatted 250GB drive across
the country, where 65mm film frames are scanned at very high resolution and
copied to the drive. It must be FAT32 at this step, no exceptions.
When we receive the filled drive, we have to get the frames off of it and
onto multiple SGI (Irix) systems. All these systems mount a common graphics
server, which in the near future will be a PV745N, so this is where the
frames are centralized.
Optimum workflow would have us copy these thousands of large files as little
as possible, and get them into the Irix environment as quickly as possible.
Since we can't share the FAT32 drive directly via NFS, our choices seem to
be as I outlined them: copy the files to a native NTFS share on the server,
or convert the USB drive to NTFS so that it can be shared.
That's why I asked the question: which of those two scenarios would likely
take less time -- copying the files from USB/FAT32 to SCSI/NTFS, or simply
converting the USB drive in place?
I don't understand what other course you're suggesting, but if it involves
more copying then it's suboptimal workflow. Given what I've described, if
you still see another possibility I'd love to have it explained further.
Thanks
--
Jim |
|
| Back to top |
|
|
Jim Mack
Guest
|
| Posted:
Sun Oct 16, 2005 6:29 am Post subject:
Re: USB Drives, NFS performance |
|
|
Eric Bursley [MVP] wrote:
| Quote: | Why not use a VPN over the Internet to copy the files between
locations? It seems to me that doing a VPN solution would be much
faster than shipping a USB drive cross country.
I can understand if one or both locations do not have Internet
access, but if you did have Internet access creating a secure VPN
between both locations would be easy, faster, and cheaper.
Without the Internet VPN as an option, your best bet would be to copy
the files from the USB disk onto the servers NTFS file system. The
folder on the NTFS file system can remain exported via NFS.
|
Thanks for your input. |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in