N8800 Pro

[mikemsd]

We just ordered a N8800 Pro. We were hoping to use it for backups and secondary storage. After reading through this forum, it appears that the general take is that the N8800 is not that great. Is there anybody with positive impressions about this unit. It is sitting in box right now in our storage room. Should I just send it back as is, or is it worth trying out?

[Joe Wells]

If you can get a refund, that's what I'd do. We've had two fail already and have completely given up on SMB/CIFS, AD integration, and ZFS. Ours is used for iSCSI only now.

[Karniv4]

It's worth youre time giving it a go. We're using a 8800SAS here, ADS/CIFS/iSCSI/Webserver all work well. Using the XFS filesystem here with no issues at all.
We're using Seagate ST3100340NS 1TB disks working faultlessly.

In my opinion, its a good budget box, just keep youre expectation's in the right area. Thecus support is a little wishy washy, but they respond to tickets raise in a fairly prompt fashion.

Andy.

[remoh]

I got a N8800PRO some days ago and after some testing and benchmarking i am happy with it.
Performance compared to our FC SAN boxes is good if you consider the price.
Of course our FC SAN boxes with 4Gbit FC and 15K disks perform better but they cost several times more and aim to other use cases. The N8800PRO will be used for backups and for some and development VMs.

Actually i use it only as iSCSI device, here are my Benchmarks:

Since my switch does not support 802.3ad i configured "Load Balance" in N8800PRO to use both LAN Ports.
I used 2 concurrent connections to make sure i could use the whole bandwith.
2 ESXi4 Servers (RX300 S4) running 2 VM Windows 2003 Server with ATTO Disk Benchmark.
I started the benchmark simultaneous so you have to add both to get overall performance.





I was not happy with the write rates, so i installed the sshd module to get more information.
In the web frontend i configured jumbo frames, but the aggregated link bond0 show MTU 1500?
I changed that to 9000 to reflect my server/switch settings and to ensure jumbo frames are used.
Benchmark after that:







A lot better...

Then i figured out that the aggregation mode "Load Balance" means linux bonding mode "balance-rr". Since "balance-alb" seems to be more efficient i changed that too.
Benchmark after that:







Thats looks good...

Combined top rates:
READ more than 230 MB/s
WRITE more than 160 MB/s

I can't say anything to NFS or CIFS performance since i do not used it yet. Maybe i will use NFS later.

@Thecus-Ivan:
Is there any special reason why load balancing mode "balance-alb" is not implement?
Did you have any problems with it? If not, i think i will continue to use it. Maybe you can add an option for in in the next firmware release?
And what about jumbo frame support for the bond0 interface? I think that should be fixed.
My web interface shows firmware version 3.00.08.10 ? I thought 3.00.08.c is actual? Which one is newer?

[remoh]

Howto activate jumbo frame support for load balanced interface:

The change is not permanent. It's lost after reboot.
Activate Link Aggregation in Web interface:
System Network -> WAN
Choose "Load Balance" (this activates the "balance-rr" mode)
reboot the system

Install and activate SSHD 2.00.00 & SYSUSER 2.00.02 Modules in Web interface
onbeat.dk/thecus/index.php/N5200_Module_List
Module Management -> Module Installation

login with SSH (user: sys password: sys)

Write 2 scripts, put them in /raid0/data/module (for example, there they are not lost after reboot).
First one is a script that reboots the system after 3 minutes if anything goes wrong.

> cd /raid0/data/module
> vi rb.sh
Put the following in:

sleep 180
reboot

Exit with ESC :wq
make it executable:net
> chmod 777 rb.sh

Write the script that changes the mtu (replace <IP gateway> with the IP adress of your gateway):
> vi changemtu.sh

ifconfig bond0 mtu 9000
ifconfig bond0 down
sleep 1
ifconfig bond0 up
route add default gw <IP gateway>


Exit with ESC :wq
make it executable:
> chmod 777 changemtu.sh
start the reboot script rb.sh:
> /raid0/data/module/rb.sh &
Note the Task id, you will need it to kill the task if the change works
execute the script that changes the bonding mode:
> /raid0/data/module/changemtu.sh
wait and pray
if the System comes back kill the rb.sh script
kill <Task id of rb.sh>
check if the mode has changed:
> ifconfig bond0

There you should see MTU 9000

[remoh]

Howto activate the port trunking/bonding mode "balance-alb" on N8800PRO (no switch support needed for this link aggregation):

This setting is not permanent! After every reboot you have to make the change.
Activate Link Aggregation in Web interface:
System Network -> WAN
Choose "Load Balance" (this activates the "balance-rr" mode)
reboot the system

Install and activate SSHD 2.00.00 & SYSUSER 2.00.02 Modules in Web interface
onbeat.dk/thecus/index.php/N5200_Module_List
Module Management -> Module Installation

login with SSH (user: sys password: sys)

Write 2 scripts, put them in /raid0/data/module (for example, there they are not lost after reboot).
First one is a script that reboots the system after 3 minutes if anything goes wrong.

> cd /raid0/data/module
> vi rb.sh
Put the following in:

sleep 180
reboot

Exit with ESC :wq
make it executable:net
> chmod 777 rb.sh
Write the script that changes the bonding mode:
> vi bonding.sh
Put the following in (replace <IP gateway> with the IP adress of your gateway, if jumbo frames are not enabled delete the "ifconfig bond0 mtu 9000" line):

chmod 777 /sys/class/net/bond0/bonding/mode
ifconfig bond0 down
sleep 3
echo balance-alb 6 > /sys/class/net/bond0/bonding/mode
ifconfig bond0 mtu 9000
ifconfig bond0 up
route add default gw <IP gateway>

Exit with ESC :wq
make it executable:
> chmod 777 bonding.sh
start the reboot script rb.sh:
> /raid0/data/module/rb.sh &
Note the Task id, you will need it to kill the task if the change works
execute the script that changes the bonding mode:
> /raid0/data/module/bonding.sh
wait and pray
if the System comes back kill the rb.sh script
kill <Task id of rb.sh>
check if the mode has changed:
> cat /proc/net/bonding/bond0
there you should see "Bonding Mode: adaptive load balancing" -> bonding mode is now "balance-alb" (formerly "Bonding Mode: load balancing (round-robin)")
If the system does not come back wait about 5 minutes, it will reboot and revert to the "Load Balance" Konfiguration because the change is not permanent.

[remoh]

For those who want to know how big the difference between the thecus and an entry class fibre channel system is, i used iometer with the profile iomix (from german c't magazin) to test:
The iomix profile simulates everday use cases.

N8800 PRO



Fujitsu FibreCAT SX88 (expect it to perform better because it was not completely without I/O for the test. It runs several VMs)



And here a high end system, not I/O free for the test too. So expect better values (5 year old EMC CLARiiON CX600)




OK, not fair. The EMC cost many many many many times the price of the N8800 PRO...

The N8800 performs well compared to the entry level FC system which costs a lot more (but has a lot more reliability features like write cache with BBU).