make tape drive on hyper-v host available within vm guest

[tmichiels]

Hi,

Just to be sure, is the following possible: we have a server (hp proliant ml 350 g8) with an hp ultrium 3000 LTO-5 tape drive in it. We would very much like to use virtualization on it but need to access the tape drive from a guest (running Windows server 2008 with backup exec 2012).

We are considering to use Hyper-v server 2012 as hypervisor.

Would this be possible with the tape redirector software?

thanks
Tom

[anton (staff)]

Yes, that's exactly the case for that product. Just make sure you've limited I/O transfer size on your initiator as it's discussed on forums.

[tmichiels]

thus you can install the tape redirector software on the host running windows 2012 core (hyper-v) ? Is there some sort of procedure for, since you don't gave a GUI?

thanks
Tom

[anton (staff)]

You install a service app where tape drive resides. And run a GUI from where you want. Just install TR and select what do you want to install - service, management console (GUI) or both.

thus you can install the tape redirector software on the host running windows 2012 core (hyper-v) ? Is there some sort of procedure for, since you don't gave a GUI?

thanks
Tom

[tmichiels]

Hi,

I've installed the service app on the hyper-v core server. I'm using the Microsoft iSCSI initiator to connect to the tape drive which succeeded without any serious problems.

Anyway, question: the backup is very very slow... I only get a throughput of 8 MB/sec while this tape drive can do a few 100 MB/sec - any idea?

rgds
Tom

[anton (staff)]

Are you running everything in a loopback? Or using physical network? Get nttcp and iperf and check how fast can TCP do between target and initiator.
Let us know the numbers and we'll proceed. Thanks!

Hi,

I've installed the service app on the hyper-v core server. I'm using the Microsoft iSCSI initiator to connect to the tape drive which succeeded without any serious problems.

Anyway, question: the backup is very very slow... I only get a throughput of 8 MB/sec while this tape drive can do a few 100 MB/sec - any idea?

rgds
Tom

[tmichiels]

Hi,

From target (core server with tape) -> initiator (guest os with backup software):

C:\>iperf -c x.x.x.8
------------------------------------------------------------
Client connecting to x.x.x.8, TCP port 5001
TCP window size: 64.0 KByte (default)
------------------------------------------------------------
[192] local x.x.x.12 port 49163 connected with x.x.x.8 port 5001
[ ID] Interval Transfer Bandwidth
[192] 0.0-10.0 sec 2.94 GBytes 2.52 Gbits/sec

And vice versa:

C:\>iperf -c x.x.x.12
------------------------------------------------------------
Client connecting to x.x.x.12, TCP port 5001
TCP window size: 8.00 KByte (default)
------------------------------------------------------------
[196] local x.x.x.8 port 12912 connected with x.x.x.12 port 5001
[ ID] Interval Transfer Bandwidth
[196] 0.0-10.0 sec 2.61 GBytes 2.24 Gbits/sec

I should expect more since the hyper-v guest is connected thru a 10 gbit virtual NIC with the host and this is a new gen 8 HP server...

But still, I don't understand why the backup throughput is only <100 mbit

thanks for your help!

[tmichiels]

I agree tape drives are not the fastest devices, but this one is an HP Ultrium LTO-5 3000 tape drive and when we do a backup directly on it, we get speeds up to 15 times what we achieve now using the tape redirector software?

Thus is there any fine tuning to do? Can we optimize things? Because you asked me to give iperf stats before looking further?

[anton (staff)]

With 2.5Gbps you can get your physical bandwidth with your tape (ignore 10 Gbps virtual network settings as they mean nothing really).

Can you move VM to another physical host and check how well it will perform over network? There are loopback issues with MS network stack solved with Windows Server 2012.

I agree tape drives are not the fastest devices, but this one is an HP Ultrium LTO-5 3000 tape drive and when we do a backup directly on it, we get speeds up to 15 times what we achieve now using the tape redirector software?

Thus is there any fine tuning to do? Can we optimize things? Because you asked me to give iperf stats before looking further?

[tmichiels]

With 2.5Gbps you can get your physical bandwidth with your tape (ignore 10 Gbps virtual network settings as they mean nothing really).

Can you move VM to another physical host and check how well it will perform over network? There are loopback issues with MS network stack solved with Windows Server 2012.

I agree tape drives are not the fastest devices, but this one is an HP Ultrium LTO-5 3000 tape drive and when we do a backup directly on it, we get speeds up to 15 times what we achieve now using the tape redirector software?

Thus is there any fine tuning to do? Can we optimize things? Because you asked me to give iperf stats before looking further?

No unfortunately I don't have any additional server with hyper-v.

About your remark about loopback issues : We are using Windows server 2012 as hyper-v server with windows 2011 abs (2008 R2) as guest...

[anton (staff)]

It can be any Windows machine on network just fine. We're not talking about production, we're talking about experiment.

I know. We did a job @ our side so now waiting for MS to fix their initiator (or we'll end with a workaround solution) to speed up the things in 5x times at least. But it SHOULD NOT work THAT slow right now.

With 2.5Gbps you can get your physical bandwidth with your tape (ignore 10 Gbps virtual network settings as they mean nothing really).

Can you move VM to another physical host and check how well it will perform over network? There are loopback issues with MS network stack solved with Windows Server 2012.

I agree tape drives are not the fastest devices, but this one is an HP Ultrium LTO-5 3000 tape drive and when we do a backup directly on it, we get speeds up to 15 times what we achieve now using the tape redirector software?

Thus is there any fine tuning to do? Can we optimize things? Because you asked me to give iperf stats before looking further?

No unfortunately I don't have any additional server with hyper-v.

About your remark about loopback issues : We are using Windows server 2012 as hyper-v server with windows 2011 abs (2008 R2) as guest...

[tmichiels]

Ok I did another test: I installed hyper-v server 2008 with the same tape and guest configuration. Now the backup is 5-10 x faster.

So could it be that the issues lies with Hyper-v server 2012?

[anton (staff)]

Actually Windows 2012 was supposed to be better (some tricks applied to make loopback work faster). Could you please stick with a working config and we'll try Windows Server 2012 with a tape in a loopback?

Ok I did another test: I installed hyper-v server 2008 with the same tape and guest configuration. Now the backup is 5-10 x faster.

So could it be that the issues lies with Hyper-v server 2012?

[tmichiels]

In the meanwhile, are these errors in the server log thing we have to worry about:

Code: Select all

10/28 0:53:02.140 7b8 SPTI: '\\?\scsi#sequential&ven_hp&prod_ultrium_5-scsi#5&35356f5e&0&000400#{53f5630b-b6bf-11d0-94f2-00a0c91efb8b}': IOCTL_SCSI_PASS_THROUGH_DIRECT (after):
 0000  38 00 00 00 00 00 06 40 01 00 00 00 00 00 04 00 8......@........ 
 0010  e0 a5 01 00 00 00 00 00 90 49 4f 02 00 00 00 00 à¥......IO..... 
 0020  38 00 00 00 08 00 04 00 00 00 00 00 00 00 00 00 8............... 
 0030  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0040  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0050  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0060  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0070  00 00 00 00 00 00 00 00                         ........         
10/28 1:21:57.298 7b8 SPTI: '\\?\scsi#sequential&ven_hp&prod_ultrium_5-scsi#5&35356f5e&0&000400#{53f5630b-b6bf-11d0-94f2-00a0c91efb8b}': scsiPassThrough failed: The parameter is incorrect. (code: 87).
10/28 1:21:57.298 7b8 SPTI: '\\?\scsi#sequential&ven_hp&prod_ultrium_5-scsi#5&35356f5e&0&000400#{53f5630b-b6bf-11d0-94f2-00a0c91efb8b}': IOCTL_SCSI_PASS_THROUGH_DIRECT (before):
 0000  38 00 00 00 00 00 06 40 01 00 00 00 00 00 04 00 8......@........ 
 0010  e0 a5 01 00 00 00 00 00 90 49 4f 02 00 00 00 00 à¥......IO..... 
 0020  38 00 00 00 08 00 04 00 00 00 00 00 00 00 00 00 8............... 
 0030  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0040  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0050  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0060  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0070  00 00 00 00 00 00 00 00                         ........         
10/28 1:21:57.298 7b8 SPTI: '\\?\scsi#sequential&ven_hp&prod_ultrium_5-scsi#5&35356f5e&0&000400#{53f5630b-b6bf-11d0-94f2-00a0c91efb8b}': IOCTL_SCSI_PASS_THROUGH_DIRECT (after):
 0000  38 00 00 00 00 00 06 40 01 00 00 00 00 00 04 00 8......@........ 
 0010  e0 a5 01 00 00 00 00 00 90 49 4f 02 00 00 00 00 à¥......IO..... 
 0020  38 00 00 00 08 00 04 00 00 00 00 00 00 00 00 00 8............... 
 0030  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0040  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0050  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0060  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ 
 0070  00 00 00 00 00 00 00 00                         ........   

[anton (staff)]

I have an impression you did not patch the xfer size on your initiator box as you shoud be:

http://www.starwindsoftware.com/forums/ ... t1089.html

Could you please double-check this?