StarPort AoE initiator - compatibility with Linux targets ?

Initiator (iSCSI, FCoE, AoE, iSER and NVMe over Fabrics), iSCSI accelerator and RAM disk

Moderators: anton (staff), art (staff), Max (staff), Anatoly (staff)

Post Reply
maraz
Posts: 4
Joined: Tue Sep 18, 2007 4:03 pm

Tue Sep 18, 2007 4:29 pm

Hello,

I am experiencing problems while using StarPort (v. 3.5 - trial version) on a Windows XP Professional (SP2) workstation in combination with Linux-based AoE targets.

I have tried the vblade and the aoeserver targets (kernel modules), on a server running Linux kernel v.2.6.10.

- With the vblade target, I see a '[E] Unknown ATA command 0x40' message in the log file of the AoE server.

- With the aoeserver target, I see no message logged in the AoE server.

In both tests, the StarPort AoE Initiator reports a driver failure with the following lines appearing in StarPort.log :
16:35:01 - CStarPortAPI::GetAoEDevices: Get AoE targets: #aoe:\DEVICE\{E471A580-D558-45D0-944D-41425824341D}
16:35:02 - CStarPortAPI::GetAoEDevices: AoE device: Device=0.0,00:0E:0C:DE:E9:FE,/dev/sdb2
16:35:02 - Exiting func: CStarPortAPI::GetAoEDevices()
16:35:06 - CWPPAoEDeviceSelect::OnWizardNext: Adding a AoE device.
16:35:06 - Entering func: CStarPortDoc::CreateAoEDevice()
16:35:06 - CStarPortAPI::InsertAoEDevice: #aoe:\DEVICE\{E471A580-D558-45D0-944D-41425824341D} [0.0] <DSL8>
16:35:07 - CStarPortAPI::InsertAoEDevice: Session failed, status == 31 !
16:35:07 - Exiting func: CStarPortDoc::CreateAoEDevice()
16:35:07 - CWPPAoEDeviceSelect::OnWizardNext: Exception caught: Driver failure.

In these tests, /dev/sdb2 is exported by the AoE server. This is a partition marked to be of type NTFS, with a size of ~13 GB.

My question for this what does 'status == 31' mean. I am looking for any clue that will allow me to overcome the problem, therefore I would appreciate any assistance that you might be able to provide.

My final question is if you have any specific (tested) recommendations on which Linux AoE target to use.

Best regards,
Manolis Marazakis.
User avatar
anton (staff)
Site Admin
Posts: 4010
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Tue Sep 18, 2007 4:57 pm

The only certified AoE target is CoRAID hardware. Other implementations may work or may not work. Their authors don't bother themselves to run cross-tests with the EXISTING solutions so we find the problems only occasionally (like in your case). Let us take a closer look at the issue to find out what's wrong and who's guilty :)
maraz wrote:Hello,

I am experiencing problems while using StarPort (v. 3.5 - trial version) on a Windows XP Professional (SP2) workstation in combination with Linux-based AoE targets.

I have tried the vblade and the aoeserver targets (kernel modules), on a server running Linux kernel v.2.6.10.

- With the vblade target, I see a '[E] Unknown ATA command 0x40' message in the log file of the AoE server.

- With the aoeserver target, I see no message logged in the AoE server.

In both tests, the StarPort AoE Initiator reports a driver failure with the following lines appearing in StarPort.log :
16:35:01 - CStarPortAPI::GetAoEDevices: Get AoE targets: #aoe:\DEVICE\{E471A580-D558-45D0-944D-41425824341D}
16:35:02 - CStarPortAPI::GetAoEDevices: AoE device: Device=0.0,00:0E:0C:DE:E9:FE,/dev/sdb2
16:35:02 - Exiting func: CStarPortAPI::GetAoEDevices()
16:35:06 - CWPPAoEDeviceSelect::OnWizardNext: Adding a AoE device.
16:35:06 - Entering func: CStarPortDoc::CreateAoEDevice()
16:35:06 - CStarPortAPI::InsertAoEDevice: #aoe:\DEVICE\{E471A580-D558-45D0-944D-41425824341D} [0.0] <DSL8>
16:35:07 - CStarPortAPI::InsertAoEDevice: Session failed, status == 31 !
16:35:07 - Exiting func: CStarPortDoc::CreateAoEDevice()
16:35:07 - CWPPAoEDeviceSelect::OnWizardNext: Exception caught: Driver failure.

In these tests, /dev/sdb2 is exported by the AoE server. This is a partition marked to be of type NTFS, with a size of ~13 GB.

My question for this what does 'status == 31' mean. I am looking for any clue that will allow me to overcome the problem, therefore I would appreciate any assistance that you might be able to provide.

My final question is if you have any specific (tested) recommendations on which Linux AoE target to use.

Best regards,
Manolis Marazakis.
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
maraz
Posts: 4
Joined: Tue Sep 18, 2007 4:03 pm

Tue Sep 18, 2007 5:17 pm

Hello,

Thank you for your previous reply.

- At this point in time, CoRAID hardware is not an option for us, so I am trying my luck with other alternatives.

- Could you please explain what 'status == 31' - just in case this gives me an idea of the root-cause of the problem ?

- By the way, I have also tested the StarPort AoE Initiator in combination with the vblade AoE target running on newer Linux servers with the 2.6.16 and 2.6.18 kernel versions, and I found no problems.


Thank you for your time.

Best regards,
Manolis Marazakis.
User avatar
anton (staff)
Site Admin
Posts: 4010
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Tue Sep 18, 2007 6:03 pm

1) In such a case I think you can perfectly install iSCSI Enterprise Target and free MSiSCSI or StarPort iSCSI initiators. And go for "all-free" solution. I know it may be sound stupid if company CEO tells you "don't buy from us" but saving money on target and paying for initiators (StarPort AoE is NOT free) is far from the best solution... IMHO of course :)

2) This is "general failure" Windows error and it can mean whole everything :)

3) If newer version works fine I'd recommend you to stick with it. Probably CoRAID people who maintain vblade had fixed some bugs :)
maraz wrote:Hello,

Thank you for your previous reply.

- At this point in time, CoRAID hardware is not an option for us, so I am trying my luck with other alternatives.

- Could you please explain what 'status == 31' - just in case this gives me an idea of the root-cause of the problem ?

- By the way, I have also tested the StarPort AoE Initiator in combination with the vblade AoE target running on newer Linux servers with the 2.6.16 and 2.6.18 kernel versions, and I found no problems.


Thank you for your time.

Best regards,
Manolis Marazakis.
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
Val (staff)
Posts: 496
Joined: Tue Jun 29, 2004 8:38 pm

Tue Sep 18, 2007 7:55 pm

maraz wrote:- At this point in time, CoRAID hardware is not an option for us, so I am trying my luck with other alternatives.

- Could you please explain what 'status == 31' - just in case this gives me an idea of the root-cause of the problem ?

- By the way, I have also tested the StarPort AoE Initiator in combination with the vblade AoE target running on newer Linux servers with the 2.6.16 and 2.6.18 kernel versions, and I found no problems.
Hi,

Use the most recent vblade or vblade-kernel AoE emulators.
They works with StarPort without problems.

StarPort is compatible with AoE v9, so any AoE device supporting the same or later AoE protocol version will work with it.
Best regards,
Valeriy
maraz
Posts: 4
Joined: Tue Sep 18, 2007 4:03 pm

Thu Sep 20, 2007 2:59 pm

Hello,

Just for the record, I tested successfully the StarPort AoE initiator (v. 3.5, running on Windows XP - SP2) with a Linux-based target (kernel version: 2.6.10). The kernel module used was kvblade-apha-3 from coraid.com (available via http://sourceforge.net/project/showfile ... _id=130453)

I saw the following warning in the log file of the target:
kvblade: ata: unknown ATA command 0xE5

The performance level was not too great, but at least it seems to be stable.

I am posting this report just in case somebody else finds it useful while testing StarPort AoE initiator in combination with Linux-based targets.

Best regards,
Manolis Marazakis.
User avatar
anton (staff)
Site Admin
Posts: 4010
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Thu Sep 20, 2007 6:41 pm

Can you share some performance numbers? :)
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
maraz
Posts: 4
Joined: Tue Sep 18, 2007 4:03 pm

Fri Sep 21, 2007 3:19 pm

Hello,

Here is a representative result:

- writing 4 GBytes of data, sequentially:
$ time dd if=/dev/zero of=zeros.bin bs=1k count=4194304
4194304+0 records in
4194304+0 records out
4294967296 bytes (4.3 GB) copied, 4633.48 s, 927 kB/s

real 77m13.816s
user 0m31.905s
sys 1m59.062s

- reading 4 GBytes of data, sequentially:
$ time dd if=zeros.bin of=/dev/null bs=1k count=4194304
4194304+0 records in
4194304+0 records out
4294967296 bytes (4.3 GB) copied, 3259.34 s, 1.3 MB/s

real 54m20.041s
user 0m30.280s
sys 1m37.405s

- AoE target: kvblade-alpha-3 kernel module, running on an XScale-based system with Linux kernel 2.6.10. This is basically a 'storage appliance' built using an XScale development board.

- The exported volume is a dedicated SATA disk (raw sequential I/O rate: ~60 MB/sec, but only ~ 17 MB/sec rate is achievable as measured by executing the dd commands above on the XScale-based host, which runs at a clock frequency of 400 MHz).

- AoE initiator: RocketDivision StarPort, v.3.5 running on Windows XP (SP2), where the dd command above is executed from a cygwin terminal session. In this experiment, the AoE-exported volume was formatted as an NTFS filesystem.

- network: FastEthernet (100 Mbps)

I would appreciate any comment that you might have. The AoE target that I use is admittedly 'under-powered' but that is what I have to work with ...

Best regards,

Manolis Marazakis.
User avatar
anton (staff)
Site Admin
Posts: 4010
Joined: Fri Jun 18, 2004 12:03 am
Location: British Virgin Islands
Contact:

Fri Sep 21, 2007 4:20 pm

1) Dump 100 Mbit network. It's unusable for SAN. Dirt cheap GbE NICs are already mainstream.

2) Use Intel I/O Meter as the measurement tool on the initiator side.

With the proper PCI-X or PCEe NICs and fast CPU on both sides you should be pretty close to the wire speed. ~100 MB/sec.
Regards,
Anton Kolomyeytsev

Chief Technology Officer & Chief Architect, StarWind Software

Image
Post Reply