• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Operating System

B

BonOeil

n00b
Joined
Dec 12, 2010
Messages
32
Hi,

I'm new in this forum :cool:

My server is compose of:

¤ Norco 4220
¤ Fortron Green 400W
¤ P5N32-E SLI plus
¤ Pentium e2160
¤ 2 x 2Go DDR2 OCZ Reaper
¤ 2 x Perc 5/i +BBU
¤ 1 x 80Go 2.5" (OS)
¤ 8 x WD Green 1To
¤ 4 x Samsung 2To

It's actually on FreeNas and I think to migrate to WS2008 R2 with WHS on a VM.

Good idea? Because I want more control on my Raid cards, doing 2 VM and others stuffs...

What do you think ?

Thank you !
 
Last edited:
I'd put in more RAM if you are running several virtual machines.
Won't you risk losing your files if you migrate them from FreeNAS to Windows?
 
Okay for the memory, I'll think about it.

And for the migration, I still don't know how I'm doing it, perhaps with 2 pc and a Perc5i on each.

There are only one Perc used on FreeNas and one array of 8 x 1To so I can use the other Perc and a array of 4 x 2To on a other pc for the migration...

But I wonder if WHS is better than FreeNas ?
 
Chimel said:
I'd put in more RAM if you are running several virtual machines.
Click to expand...

I would also replace the processor with a much beefier CPU upgrading to an i3 or i5 system.

Good idea?
Click to expand...

Not with your current hardware.

But I wonder if WHS is better than FreeNas ?
Click to expand...

Better. I am not sure of that. As far as a file server probably not. For a backup server probably yes. This part may not be easy on freenas unless you understand non windows operating systems.
Easier to use for windows users. Yes.
 
Last edited:
My roommate did WHS as a guest under esxi and was not terribly thrilled with it. He also tried FreeNAS under esxi and wasn't happy there either. Eventually he just made a seperate physical SAN via iscsi with bonded gigabit. Food for thought.
 
BonOeil said:
But I wonder if WHS is better than FreeNas ?
Click to expand...

it is different, for sure. "better" is subjective.

i would certainly upgrade the CPU like suggested, before putting server 2008 on that computer. WHS would probably struggle as well on the current hardware.


what is your reasoning for dumping FreeNas?
 
Try Solaris 11 Express on the hardware. Should give you performance above anything else. ZFS is easy. WHS is a pain.
 
Flintstone said:
Try Solaris 11 Express on the hardware. Should give you performance above anything else. ZFS is easy. WHS is a pain.
Click to expand...

+1

ZFS gives you the added bonus of data integrity. Everything written to your storage pool is checksummed on the fly, so if data is silently corrupted, this can be detected and repaired. In the long run, how important is your data?
 
ghost6303 said:
it is different, for sure. "better" is subjective.

i would certainly upgrade the CPU like suggested, before putting server 2008 on that computer. WHS would probably struggle as well on the current hardware.


what is your reasoning for dumping FreeNas?
Click to expand...

My reasons are multiples but the most important is I want more support of my Perc5i because there are no smart on my disks and no hard disk standby time.

I want to be able to install some WM.

And I would like to simply install some addins like Sabnzbd, CouchPotato,.. because on FreeNas is a pain to do there.

Flintstone said:
Try Solaris 11 Express on the hardware. Should give you performance above anything else. ZFS is easy. WHS is a pain.
Click to expand...

I don't know it. Your server is running under Solaris ?

tormentum said:
+1

ZFS gives you the added bonus of data integrity. Everything written to your storage pool is checksummed on the fly, so if data is silently corrupted, this can be detected and repaired. In the long run, how important is your data?
Click to expand...

My data is not very important but I don't want to lose it :D
 
Okay I think about all this and the best solution is to change my configuration with a server motherboard and IPMI with KVM support.

I would take :

¤ SuperMicro X8SI6-F
¤ Intel Core i3-550
¤ 4x2Gb DDR3 ECC
¤ HP Sas Extender

But I wonder if software Raid5 doesn't take much time in reconstruction with the builtin LSI SAS 2008 ?

What do you think ?
 
Interesting but I would like to use Bittorrent + Sabnzbd in my server.

Is it possible to use them in VM under VMWare ESXi ?

One other thing, the Core i3 doesn't have VT-d...
 
Bittorrent can be used on pretty much any operating system with native applications. I am not sure what Sabnzbd is.
 
about my All-In-One concept without vti-d/ pass-through on i3

in this case, you can do it "as usual"

install ESXi hypervisor on your ahci-Sata disk (if sata chipset s supported)
Create a datastore on this disk
Install all VM's like Windows, Linux, BSD, Solaris* on this disk
Use whatever OS for its best use case

This is quite ok to work with a few VM's or to try them.
But access to your VM's via ESXI-filebrowser for cloning or backups is painfully slow.
You can have only a few snaps and with each snap it will become slower and slower.


With vti-d / pass-through
you will get the following additions:

You can Install a virtualized ZFS NAS/SAN with native disk-access via pass-through
-> much more performance and data-security due to ZFS features

You can install VM's on this embedded SAN with all its snapshot capabilities
and easy/ fast access to your VM's via iSCSI, NFS, SMB and Windows previous version.

You will get one of the most sophisticated ZFS-NAS boxes with SMB, AFP, Encryption, Deduplication and a lot more (depending on ZFS-OS) based on a type-1 barebone hypervisor. So if you want to have all these nice features and your mainbard supports vti-d its always suggested to buy a i7 or Xeon.

Beside low power/ low price i would never miss these feaurtures

gea
 
BonOeil said:
Interesting but I would like to use Bittorrent + Sabnzbd in my server.

Is it possible to use them in VM under VMWare ESXi ?

One other thing, the Core i3 doesn't have VT-d...
Click to expand...

I have Sabnzbd running on my Solaris 11 Express server without any problems. I would probably prefer running bittorrent on a Windows VM.

And yeah you need a Xeon to have VT-d...
 
Thanks everybody and Gea ;)

So my config will be :

¤ Norco 4220
¤ Corsair GX 800W
¤ SuperMicro X8SI6-F
¤ X3430
¤ 16Gb DDR3 ECC
¤ SAS Expander
¤ 1 x 160 Go 2.5"
¤ 8 x Western Digital Green 1To
¤ 4 x Samsung 2To

(not yet in bold)

One more thing, if I run FreeNAS (virtualized) under ESXi, can I access to my disks because of the builtin LSI SAS 2008 ?
 
FreeNAS doesn't support LSI SAS2008. My ZFSguru project does, however. And Solaris should also support this controller.
 
If the passthrough happens so that FreeNAS has direct access the same hardware; it still would need a driver for that hardware. If you emulate the controller so FreeNAS reads an IDE/SCSI/AHCI controller instead of your real hardware, then it could work on FreeNAS, but very slow and possibly with risk to your files, unless ESXi is set to honor buffer flush commands.
 
BonOeil said:
Thanks everybody and Gea ;)

So my config will be :

¤ Norco 4220
¤ Corsair GX 800W
¤ SuperMicro X8SI6-F
¤ X3430
¤ 16Gb DDR3 ECC
¤ SAS Expander
¤ 1 x 160 Go 2.5"
¤ 8 x Western Digital Green 1To
¤ 4 x Samsung 2To

(not yet in bold)

One more thing, if I run FreeNAS (virtualized) under ESXi, can I access to my disks because of the builtin LSI SAS 2008 ?
Click to expand...


i would avoid the expander, especially with SATA drives. Its cheaper, has more performance and less problems to add a SAS Controller, best based on LSI 1068e (SAS1) or LSi 2008 (SAS2).

I would prefer LSI 1068 based with your drives, because LSI 2008 gives you WWN-Ids insted of simple controller i's

read about:
http://www.nexenta.org/boards/1/topics/823

gea
 
Can a fan-in cable be used to connect a 4-disk miniSAS backplane to 4 SATAII mainboard ports? If not, there is no way of feeding 20 drive bays from the onboard controllers of the X8SI6-F and an additional LSI 8-port card without an expander. That board only has one 8x/16x PCIe slot. On the other hand, the 4x slot should be enough for 4 drives. Two 1068e cards are still less expensive than an expander.
 
i suppose that won't work.
(i have tried such a thing with a supermicro backplane).

It had worked only when connecting the backplane with miniSAS directly to the 1068 Controller, not when connecting via a miniSAS-> Sata Cable, connectet to Onboard-Sata.
So your Option may be a Expander or a Controller > 8 Ports or a new Mainboard

edited
http://www.supermicro.com/xeon_3400/Motherboard/X8SIE.cfm?IPMI=Y&TYP=SAS
the 4x slot is fast enough for a SAS controller even with 8 drives (but prefer the others first)
so you could have 8 drives on LSI 2008 and add two 1068 controller
no problem up to 24 drives


Gea
 
Last edited:
So what doing :confused:

And where to buy a 1068 controller ? I don't find them on eBay and between us I prefer to use an expander with the motherboard because of the price and the other hand I still have my 2 Perc5i, maybe I keep them ?

I can change too the motherboard (X8SIL-F) and going to an Areca 1680ix-16...

Seriously, I being crazy :D :D
 
The ix series have a port multiplier/expander, right? You may want to avoid these series; use only full-bandwidth 'pure HBAs'; the LSI 1068E is such a chip and works on virtually all OS.

You may look for Intel SASUC8i; they should work fine on your system. All you may need to do is flash to SuperMicro/LSI IT-firmware if they are shipped in IR-mode by default. Some people say they by default come in IT-mode, while others state they received theirs in IR-mode. Either way, you want IT-mode; which means non-RAID SAS/SATA controller; exactly what you want for ZFS, which i assume is the path you're investigating right now?
 
Yes, I want to make ZFS array.

So with the SuperMicro X8SIL-F and 3* IBM Br10i, I'm on good way ?
 
_Gea said:
the 4x slot is fast enough for a SAS controller even with 8 drives (but prefer the others first)
so you could have 8 drives on LSI 2008 and add two 1068 controller
Click to expand...
Please remember that PCIe 1.0 only has an effective data rate of around 150 MB/s per lane after overhead. While a x4 slot it will work with 8 drives, it will get saturated if you read sequential data from all drives simultaneously, which ZFS most likely will. Further, the x4 slot shares the DMI bandwidth, which also is only a x4 PCIe 1.0 bus, with all other components in/on the PCH like SATA controllers, the graphics controller and most importantly the 2 gigabit ethernet controllers.


BonOeil said:
Yes, I want to make ZFS array.

So with the SuperMicro X8SIL-F and 3* IBM Br10i, I'm on good way ?
Click to expand...
Sounds reasonable. I cannot comment on the IBM controllers, i only used SASUC8i (relabeled LSI) and LSI 9211-8i controllers. Best would be to connect the least amount of drives to the controller in the x4 slot. Another option would be the LSI 9201-16i, still looking for someone who tried this controller.
 
Last edited:
We can read smart but don't know for power management.

However, there is one problem with 3 HBAs because EXSi can connect 2 devices maximum to a VM :(
 
So I should buy an Sas expander if I can't use 3 devices on ESXi and keep a BR10i...
 
i have
2 x LSI 1068e
1 x LSI 2008
running with passthrough ESXI 4.1

in vmware reference you will find:
you can only use max 2 adapter with pass-through per virtual machine:


with my own esxi 4.1 i could add all three adapter to the same machine
but had not yet tried all three adapters if they will work.
i will check and report

i looked at the online help under add pci-adapter:
"Die VMDirectPath-E/A ermöglicht einem Gastbetriebssystem einer virtuellen Maschine den direkten Zugriff auf physische PCI- und PCIe-Geräte, die mit einem Host verbunden sind.
Jede virtuelle Maschine kann mit bis zu sechs PCI-Geräten verbunden werden."

So with ESXi 4.1 you can add up to 6 Adapter per VM
 
Last edited:
Personally I would avoid expanders but I'm not trying to talk you out of using them. While they seem to work fine for a lot of people there is a reason why the HP SAS expander thread is so long.
 
Thanks Gea & Omniscence !

I've read again the docs and I've found this :

Code: 
                             [CENTER]VMDirectPath limits

VMDirectPath PCI/PCIe devices per host                         8
VMDirectPath PCI/PCIe devices per virtual machine              4[/CENTER]

So It's Okay for me :cool:
 
Hello,

I'm waiting for my orders and I'm wonder if I buy an another 2.5" 160Go for making a Raid1 OS ?

It's better for ZFS, no ?
 
BonOeil said:
Hello,

I'm waiting for my orders and I'm wonder if I buy an another 2.5" 160Go for making a Raid1 OS ?

It's better for ZFS, no ?
Click to expand...

Its's not better for ZFS, its better for uptime.
Usually all important data are on a datapools, not on syspool,
so a reinstall to a new disk, if you have a hd crash, is not a problem.

Gea
 
omniscence said:
Please remember that PCIe 1.0 only has an effective data rate of around 150 MB/s per lane after overhead.
Click to expand...
PCI-express uses 8/10-bit encoding meaning that 2,5Gbps per lane means 250MB/s of bandwidth with signalling overhead already included. It is also full-duplex, meaning 250MB/s in and 250MB/s out at the same time; so 500MB/s aggregate.

So you get:

PCI-express x1
1.0 = 250MB/s
2.0 = 500MB/s
3.0 = 1000MB/s

PCI-express x4
1.0 = 1000MB/s
2.0 = 2000MB/s
3.0 = 4000MB/s

PCI-express x8
1.0 = 2000MB/s
2.0 = 4000MB/s
3.0 = 8000MB/s

PCI-express x16
1.0 = 4000MB/s
2.0 = 8000MB/s
3.0 = 16000MB/s

PCI-express is a modern point-to-point interface, sharing nothing with other devices; all dedicated links and full-duplex so the last example of 16GB/s bandwidth is actually 16GB/s in and 16GB/s out.

I think PCI-express is a very good interface, even the slowest PCIe 1.0 x1 shouldn't give you excessive performance issues like old PCI does due to its half-duplex and shared-access nature.
 
I'm aware of the theoretical bandwidth. This does not give a complete answer about what can be achieved practically. Since PCIe is a packet based communication standard, the net data rate does depend on the packet payload and header size.

Ten seconds of google yielded the following paper: http://www.plxtech.com/files/pdf/technical/expresslane/Choosing_PCIe_Packet_Payload_Size.pdf

Some years ago I did find a more complete paper I (if I remember correctly, it was from an FPGA manufacturer, I think Xilinx) but I lost the link. However, the abovementioned paper should be sufficient to prove my point. I checked my fileserver and the maximum payload size for the SAS and ethernet controllers is 128, which results in an absolute maximum net data rate of 216 MiB/s for PCIe 1.0 not considering flow control and handshaking. Increasing the signaling rate to PCIe 2.0 will not increase the packet efficiency but the overall efficiency by decreasing handshaking latency.

EDIT: The situation is similar with ethernet packets and the reason some people here suggest jumbo frames for higher throughput, increasing the packet efficiency. But increasing packet sizes also has drawbacks, it increases latencies (this is a reason some people still prefer PCI sound cards over PCIe for real-time recording) and it requires larger buffers.
 
Last edited:
Very interesting, omniscence!

However, i wonder about current chipset payload size, if AMD and other chipsets support 512-bytes but Intel limits at 64-bytes then this may explain difference in PCI-express performance between such chipset platforms.

I may be able to test this, by using a SuperMicro USAS-L8i controller with some SSDs in RAID0 and test throughput when the controller is connected to PCI-express 1.0 x1 slot. I can't test this easily on Intel chipset, though.

And perhaps the current generation Sandy Bridge chipsets have improvements in this area?
 
You must log in or register to reply here.
Back
Top