WNDR3800 new firmware

This old blog post has gathered some interest lately.
Maybe it was spot on, because Netgear released a new firmware for the WNDR3800 that promise some NTFS improvements, as the readme says.
So, here are the new tests with firmware version V1.0.0.32:

root@pc:~# dd if=/dev/zero of=t/testfile bs=4096 count=102400
102400+0 records in
102400+0 records out
419430400 bytes (419 MB) copied, 27.8515 s, 15.1 MB/s

root@pc:~# dd if=t/testfile of=/dev/null
819200+0 records in
819200+0 records out
419430400 bytes (419 MB) copied, 41.313 s, 10.2 MB/s

and the updated table:

Format	Read MB/s	Write MB/s
local USB	27.7	28.7
NTFS	10.2	15.1
ext2	12.0	19.6
FAT	10.2	14.6

NTFS now has reasonable performance, but ext2 is still the best choice.

WNDR3800 NAS performance

UPDATE: there are some news on the subject.

The WNDR3800 has an USB 2.0 port for connecting an external storage.
I've done some testing with an external disk.
The disk used was a 120GB Packard Bell Store and Save 2400 bought back in 2008.
Firmware version of the WNDR3800 was V1.0.0.24
Since the test disk was primarily used to share files between several machines, it was already formatted with NTFS for windows compatibility.

The WNDR3800 exports USB storage by an SMB share and by HTTP: I mounted the share under /mountpoint

So, let's check write performance by creating a test file:

# dd if=/dev/zero of=/mountpoint/testfile bs=4096 count=102400

102400+0 records in

102400+0 records out

419430400 bytes (419 MB) copied, 94.4866 s, 4.4 MB/s

And now read performance:

# dd if=/mountpoint/testfile of=/dev/null

819200+0 records in

819200+0 records out

419430400 bytes (419 MB) copied, 42.9255 s, 9.8 MB/s

Writing is often slower than reading, but 9.8MB/s is the throughput of a 100Mb Ethernet, and we're using Gigabit here.
Let's try with HTTP:

# wget -O /dev/null http://192.168.1.1/shares/USB_Storage/testfile
--2012-01-13 19:08:42--  http://192.168.1.1/shares/USB_Storage/testfile

Connecting to 192.168.1.1:80... connected.

HTTP request sent, awaiting response... 200 OK

Length: 419430400 (400M) [application/download]

Saving to: `/dev/null'

100%[======================================>] 419,430,400 13.6M/s   in 30s    

2012-01-13 19:09:12 (13.4 MB/s) - `/dev/null' saved [419430400/419430400]

HTTP has lower overhead than CIFS, but we're still far from the high performance we can expect.
Maybe the disk is the bottleneck?
Let's try connecting the disk directly to the USB port of the pc:

# dd if=/media/data_ntfs/testfile of=/dev/null

819200+0 records in

819200+0 records out

419430400 bytes (419 MB) copied, 15.1453 s, 27.7 MB/s

It's not the disk: it's able to reach nearly 30MB/s
And now the write performance:

# dd if=/dev/zero of=/media/data_ntfs/testfile bs=4096 count=102400

102400+0 records in

102400+0 records out

419430400 bytes (419 MB) copied, 14.5965 s, 28.7 MB/s

So, this disk is not slower while writing: it seems that the WNDR3800 slows things down.
Could it be the NTFS format? Let's try with a linux filesystem that the Netgear firmware may natively support.
I reformatted the disk with ext2 and retried with the WNDR3800 connection:

# dd if=/dev/zero of=/mountpoint/testfile bs=4096 count=102400

102400+0 records in

102400+0 records out

419430400 bytes (419 MB) copied, 21.423 s, 19.6 MB/s

That's a huge gain versus NTFS.
Read time:

# dd if=/mountpoint/testfile of=/dev/null

819200+0 records in

819200+0 records out

419430400 bytes (419 MB) copied, 35.0745 s, 12.0 MB/s

Again, better than NTFS but strangely enough, lower than writing.
Let's check HTTP:

# wget -O /dev/null http://192.168.1.1/shares/T_Drive/testfile

--2012-01-14 14:24:24--  http://192.168.1.1/shares/T_Drive/testfile

Connecting to 192.168.1.1:80... connected.

HTTP request sent, awaiting response... 200 OK

Length: 419430400 (400M) [application/download]

Saving to: `/dev/null'

100%[======================================>] 419,430,400 18.9M/s   in 22s    

2012-01-14 14:24:46 (18.2 MB/s) - `/dev/null' saved [419430400/419430400]

HTTP confirms its lower overhead, but again, lower performance than writing.
So partition format has a considerable impact on WNDR3800 NAS performance.
One last test may be the use of a really simple filesystem: FAT.
Back to square one with a FAT32 partition:
Write:

# dd if=/dev/zero of=/mountpoint/testfile bs=4096 count=102400

102400+0 records in

102400+0 records out

419430400 bytes (419 MB) copied, 28.7441 s, 14.6 MB/s

Read:

# dd if=/mountpoint/testfile of=/dev/null

819200+0 records in

819200+0 records out

419430400 bytes (419 MB) copied, 41.2611 s, 10.2 MB/s

# wget -O /dev/null http://192.168.1.1/shares/USB_Storage/testfile

--2012-01-14 14:38:28--  http://192.168.1.1/shares/USB_Storage/testfile

Connecting to 192.168.1.1:80... connected.

HTTP request sent, awaiting response... 200 OK

Length: 419430400 (400M) [application/download]

Saving to: `/dev/null'

100%[======================================>] 419,430,400 13.1M/s   in 29s    

2012-01-14 14:38:57 (13.8 MB/s) - `/dev/null' saved [419430400/419430400]

FAT does not boost performance, and even if it did, I wouldn't use FAT as a filesystem for anything serious.

Format	Read MB/s	Write MB/s
local USB	27.7	28.7
NTFS	9.8	4.4
ext2	12.0	19.6
FAT	10.2	14.6

UPDATE: there is a new firmware available that changes some of this results.

Motorola Defy wireless performance

Following my previous post, I've done some testing on my phone wi-fi performance compared to a full-blown PC.
Setup is the same as in the previous post, even the phone location is the same as the previous laptop location.
Data transfer was between the Defy and the desktop PC, wire-connected at gigabit speed to the Netgear WNDR3800.

Speed	MB/s
2.4Ghz@54Mbs	2.5
2.4Ghz@130Mbs	2.6
2.4Ghz@300Mbs	2.6
5Ghz@300Mbs	---

So, while the Defy supports 801.11n, it takes no advantage of it.
The 5Ghz network is not even recognised and the 300Mbs connection was not very reliable, I had to retry a couple of times before getting a full transfer.

WNDR3800 wireless performance

After looking at wired performance, I've got a spin at wireless:
Test conditions were as follows:

Router Firmware Version: V1.0.0.24
PC 1: notebook with Intel PRO/Wireless 4965 AG
PC 2: desktop with nVidia Corporation CK804 Ethernet (Gigabit speed)
PC 1 was within one meter from the router

On PC 1:
# dd if=/dev/zero bs=4096 count=102400 | nc 192.168.1.31 1999

On PC 2:

# nc -l 1999 | dd of=/dev/null

Speed reported is from dd output on PC 1:

Speed	MB/s
2.4Ghz@54Mbs	2.6
2.4Ghz@130Mbs	8.1
2.4Ghz@300Mbs	9.3
5Ghz@300Mbs	9.0

WNDR3800 backdoor

The Netgear WNDR3800 firmware is a version of OpenWRT, and that's one of the reasons I've bought this router.
While being open is a good thing, being wide open maybe it's not.

This firmware has a backdoor, enabled by a simple utility (the windows version is available directly from the Netgear support site).
Once executed, it gives root access to the router without any authentication.
Consequences may vary from simple denial of service (you can reboot at will), or something more elaborate: download the /etc/shadow file, run John the Ripper, and get the admin password. Then, logging on with the web interface, you can flash a full version of OpenWRT, install tcpdump and capture all unencrypted traffic.

Problem is that the backdoor can be enabled by any device connected to the router, even via wireless; so be aware that allowing someone to simply use your connection, means giving them full root access.
The backdoor is not accessible from the Guest Network, so if you are going to allow someone to use your wireless, at least give them only the Guest Network.

Netgear Genie... oh really?

A runner up for the first place in the User Interface Bad Design Contest: the Netgear Genie.
I was setting up wifi scheduling as I did on my old router, but I've found that the schedules were off by one hour.
No big deal, maybe the router defaulted to GMT, but where did they put the time settings?
Maybe under Administration?
Even Advanced Setup would be somewhat logical...
But no... it seems that to Netgear engineers, time belongs to security.

Netgear WNDR3800 switch performance

I'm trying the WNDR3800.
Some reviews are reporting less than stellar performance on the gigabit switch.
I've done some testing myself.
Using 2 linux pcs, with gigabit ethernet ports, I've put the first in receiving mode:

# nc -l 1999 >/dev/null

and then I started sending data from the second one:

# cat /dev/zero | nc 192.168.1.2 1999

The result is well over 110 MB/s:

Then, I've repeated the test using a straight cable connection between the two machines:

The result is pretty much the same, so the switch part of the WNDR3800 does not significantly impact performances.