Deleted to avoid pop-up ads.

I have the same issue with the device. In my environment with multiple servers and multiple devices ( workstations, laptops, smart phone, consoles) I would prefer to be able to do this from NFS via my main server so that access rights are preserved across the board with my current in-house accounts. From a security perspective this also allows for more direct access instead of tunneling to their website for control of the drive hence being more secure and more efficient.

Bottom line, I shouldn't have to modify/hack the box to do something for rudimentary functionality.

As am I. This is a linux box without the most basic of file functionality. Can I hack the box, yeah. Is it a pain to do, yeah. Do I really want to alter it, no. Will it void the warranty, yes.

NFS would have been so much better for LAN functionality instead of the client. Less headache, and I would almost bet better performance.

Not sure if this will help,But I figured out how to run a website from my PogoPlug with it's own domain I started off with a .tk domain which is free than I bought a .com domain they both work great,I didn't do anything as far as hacking the PoGoPlug.This device is simply amazing.

My website running on my PogoPlug
http://www.just25watts.com

Please add support for direct disk sharing. I have my pogoplug on a public subnet, and my computer which is connected to the SAME Gigabit Switch, when I connect to it, has to go out to the pogoplug servers and back again. The performance drags.

DreamDemon said:

Pogo 朱 said: Please add support for direct disk sharing. I have my pogoplug on a public subnet, and my computer which is connected to the SAME Gigabit Switch, when I connect to it, has to go out to the pogoplug servers and back again. The performance drags.

I have to agree with this.  If my ISP takes a dump I dont have acess to my drives.  There is also the issue of performance across the network/internet since it has to do a round trip.

Here's the sad thing.  I'm taking all my old data on old laptops that I have lying around, and I'm transferring them to the pogoplug.  I'm on a gigabit connection but the upload is going over encrypted connection so I'm seeing 200 to 300 kilobytes per second.  If I connect to the Pogoplug with winscp (directly) I'm seeing 1800kB per second, if I lower the encryption (to blowfish instead of AES) I get to 2500kB per second.  And mind you, this is still encrypted and the computers and Pogoplug are on the same gigabit switch (!) but just doing the roundtrip to pogoplug and back loses 90% of the performance.  Just give me the ability to mount the drive directly. Even 2.5 megs per second is too slow on the same gigabit lan.  

I decided to do some benchmarking so I can put some scientific numbers onto my anecdotal experiences.

Test Setup:

I have a 5 drive RAID 5 array (connected via USB) to the Pogoplug. When I connect that same drive directly to a PC, I max out the USB 2.0 link easily (forget the exact numbers... 50MB/s..?). The array also has a eSata port, so overall the rate data can be transfered is not the issue here. Also, upstream from the Gigabit switch is a 300Mbps symmetric (upload/download = same) connection directly to the Internet. So that's not the bottleneck either.

Timing Tests

I ssh'd into the pogoplug to test the unencrypted raw download performance by issuing the following command on my ext3 partition:

time wget http://download.fedoraproject.org/pub/fedora/linux/releases/14/Live/i686/Fedo...
-- 1m15.198s to transfer 719323136 bytes = 9.1MB/s (73Mbps) -- this means without encryption, it's possible to upload data to a pogoplug ext3 usb drive at 73Mbps on a 300Mbps WAN link

I repeated this identical test but this time transferring the file to my ntfs raid5 usb drive: 2m1.172s or 5.7MB/s (45Mbps) -- this means without encryption, it seems like an ext3 usb partition is 40% faster than an ideal ntfs partition.

Now moving to the local lan (everything plugged into the same gigabit switch)
wget from local lan web server to Pogoplug ntfs raid5: 6.1MB/s (49Mbps)
wget from local lan web server to Pogoplug ext3: 8.73MB/s (70Mbps)

So what this says is there is a bottleneck in the Pogoplug ntfs driver reducing performance by 30-40% versus ext3.

I also used the Pogoplug web interface with full security disabled to download the same file from the ntfs raid5 drive to local 7.36MB/s 59Mbps.
Then I uploaded from local to ntfs raid5 (via web interface): 4.5MB/s 36Mbps

This tells me there're some significant performance issues in the web-client upload code (!). That's pretty bad.

So now I'm turning on "full security" and verifying https:// in the url. Uploading the file from local to pogoplug ntfs raid5: 4.5MB/s 36Mbps (same performance with or without security for uploading!) Why is this?
Then i reversed this and got: 6.1MB/s (49Mbps) ~20% reduction in performance on full security.

But basically this also says that when uploading to pogoplug via web client the bottleneck is not the encryption but something else in between.

So now I'm going to run a timed copy from ext3 to ntfs raid5 to remove the network effects and check the data transfer limits: 8.37MB/s (67Mbps)
Then copying back from ntfs raid5 to ext3: 9.74MB/s (78Mbps)

What this says is the ntfs raid5 can read data about 20% faster than ext3 single drive and there is a performance penalty writing to a raid5... but overall the natural limit seems to be in the 70 to 80Mbps range.

So some questions here: why is the performance so slow internal to the pogoplug? And the conclusion here is that moving unencrypted data directly into the pogoplug on the local gigabit lan is negligible in performance to moving data without the network.

Using the same 1 gigabyte file:
Windows Client from pogoplug ntfs raid 5 to local drive: 8576KB/s (67Mbps)
Windows Client from local drive to pogoplug ntfs raid5: 3399KB/s (27Mbps)
Windows Client from pogoplug ext3 to local drive: 6250KB/s (49Mbps)
Windows Client from local drive to pogoplug ext3: 4085KB/s (32Mbps)

This result is good and bad. Here having an ntfs raid5 for transferring files OUT of the pogoplug has advantages but moving stuff INTO the pogoplug shows there is a bottleneck at the Windows Client and whatever else is in between the Windows Client and the pogoplug. There's a huge penalty (about half!) when uploading to pogoplug versus downloading.

Now I'll switch over to winscp and fiddle with various settings:

Winscp from pogoplug ntfs raid5 to local drive, aes encrypted: 2116KB/s (17Mbps)
Winscp from local drive to pogoplug ntfs raid5, aes encrypted: 1990KB/s (16Mbps)
Winscp from local drive to pogoplug ntfs raid 5, blowfish encrypted: 2260KB/s (18Mbps)
Winscp from pogoplug ntfs raid5 to local drive, blowfish encrypted: 3125KB/s (24Mbps)
Winscp from local drive to pogoplug ext3, blowfish encrypted: 1450KB/s (11Mbps)
Winscp from pogoplug ext3 to local drive, blowfish encrypted: 2843KB/s (22Mbps)
Winscp from local drive to pogoplug ext3, aes encrypted: 1233KB/s (10Mbps)
Winscp from pogoplug ext3 to local drive, aes encrypted: 2186KB/s (17Mbps)

So the conclusions are: When connecting by Windows Client it's the fastest way to transfer large files from pogoplug to local...

In general uploading files to the pogoplug presents huge performance issues (this is what I do on a regular basis).

In general I'm puzzled why performance maxes out at ~9MBps in all the best scenarios.

Any thoughts...?

Actually considering that wget is built into pogoplug, it should be easy enough to implement a feature to have pogoplug directly download files on its own. Perhaps in conjunction with a browser plug-in.