• 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

_Gea said:
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
Click to expand...

Uptime ? What do you mean ?
 
if your os-drive is not mirrored and fails,
you have to reinstall your system on a new drive and
you must reconfigure all needed settings.

if you system is on a mirrored raid and one drive fail,
your system keeps running in most cases, no service interrupt.

Gea
 
sub.mesa said:
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?
Click to expand...

I have to check what electrical slot configurations the boards I have lying around actually have, there are like 5 cheap Intel boards, but no server class hardware. I also have a SASUC8i controller left. If you come up with some test scheme, I may be able to reproduce this.

But I would not care to much about these bandwidth limitations, as most SAS controllers are x8. Are there some specific bandwidth limitations you encountered? Also, inceasing the payload size may not be the best idea as long as the bandwidth is sufficient, as this will have negative impact on latency.

EDIT: The most noteable thing about Sandy Bridge is, that they not only made the PCIe interface at the chipset comply with PCIe 2.0 but also doubled the DMI throughput. So you will no longer run these interfaces into saturation if there is a SAS controller with 8 drives connected through the chipset (as long as they are not all SSDs).
 
Last edited:
Well if you would connect such a controller to PCI-express x4 electrical slot, like on AMD Zacate motherboards then you might run into some bottlenecks potentially, especially if like your document implies you cannot (always) utilize most of the PCI-express bandwidth.

But just out curiosity, i would love to test both sequential and random 4K I/O on a controller connected to PCI-express 1.0 x1 so having 250MB/s theoretical bandwidth (not throughput; as your document describes).
 
Personally I doubt that Zacate can process 1 GiB/s of data. And by process I mean like extracting files from file system blocks and package them into TCP/IP frames.

You can "simulate" a PCIe x1 connection by isolating all but one lane with sticky tape.
 
If you throw a few SSDs on Zacate that would be able to achieve those speeds. Whether it can achieve that in HDD RAIDZ2 configuration for example; i'm not sure! But with 8GiB RAM the Zacate platform is able to perform better than Atom platforms for ZFS workloads. The out-of-order CPU should also make it faster than the Atom; though not as much as i would have liked.
 
After two months, I have received all hardware and I use napp-it on Open Indiana under ESXi :cool:

By the way, I've didn't easy choice between Napp-it and ZFSGuru ! :)

So for a test, I've did a pool raidz with 4 x Samsung 2T.



Are the benchmarks are good ?
 
Last edited:
sequential (block) read is much better than write
I would suggest to compare with a raid-10 pool

raid-10 is much faster but you have less space
if you do not need the space, optimize for speed
not for max capacity, especially if its a ESXi datastore

Gea
 
Thank you Gea for your quick reply, I appreciate :)

I think to stand on RaidZ because of space and it's fast enough...

Can I read temperatures of HDD in Napp-it ?

And what about power economy of HDD (spin-off) ?
 
BonOeil said:
Thank you Gea for your quick reply, I appreciate :)

I think to stand on RaidZ because of space and it's fast enough...

Can I read temperatures of HDD in Napp-it ?

And what about power economy of HDD (spin-off) ?
Click to expand...

about temperature:
possible with smartmontools (installed with napp-it installer)
basic settings only, see menu disks-smartinfo

about hd spin-down
could be set with power management
see menu system - power management

Gea
 
Baddreams said:
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.
Click to expand...

I use WHS in Hyper-V on Server 2008 r2, and it has worked just fine. The host has 4GB RAM, and the virtual machine has 1GB.

HOWEVER, I only use WHS for backups. The host machine handles file storage. If I were to run more virtual machines, I would probably install more RAM. I might, anyway!
 
You must log in or register to reply here.
Back
Top