Should I use a VPN with rclone?

Help and Support
21

So this part is only for the true paranoids :slight_smile:

Without TLS interception then the ISP doesn't know anything about the transfer except for volume of data. They might be able to guess "this is mostly upload" based on volume, but that's about it. They can't easily guess how many files (if you're doing large files) because of multipart interleaved uploads.

Now someone with a MITM proxy that can see the data can not see the raw filename nor the raw file contents in a properly configured crypt wrapper. But there is still side channel leakage; they'd see the http headers and so would know how many files, the size of each file, the directory tree structure (even if not the directory names). They might be able to make inferences ("Hmm, he's always uploading 4-9Gb files; these are probably DVD ISO rips").

We lose a level of protection.

Is it something you need to be worried about in America? Probably not. PRISM likely wouldn't flag anything because of this.

If you're in China it might be enough to warrant state scrutinity.

Kazakhstan which is forcing government CAs on all devices? I dunno...

As I said...it depends on your level of paranoia :slight_smile:

1 Like
22

good points,

i guess i should not take if for granted i am here in free america.

someone would have to be an experienced expert in the use of vpn stealth to avoid china's great firewall.
seems to me the china can easily detect the use of a vpn, by deep packet inspection and ip endpoints.
and that would perhaps attract even more attention.

23

Yes. Same config file. Same commands. Have tried messing around with --drive-chunk-size but it doesn't improve more than 20MB/s in Windows.

24

curios to compare this for both win and wsl2

https://www.speedtest.net/apps/cli

i get the same speeds for both

Latency: 5.97 ms (0.17 ms jitter)
Download: 832.49 Mbps (data used: 1.1 GB)
Upload: 922.87 Mbps (data used: 1.1 GB)
Packet Loss: 0.0%

25

With Windows:

Latency:     1.31 ms   (0.05 ms jitter)
Download:   737.47 Mbps (data used: 829.0 MB)
Upload:   932.34 Mbps (data used: 873.6 MB)
Packet Loss: Not available.

With WSL2:

Download: 674.78 Mbit/s
Upload: 594.98 Mbit/s

So Windows is actually faster. Still, rclone runs slower speeds. Not sure..

26

i am doing some testing on my win10 and wsl2.

can you post your rclone command?

27

I do

rclone copy --verbose --transfers 1 --checkers 10 --stats 1s --fast-list --tpslimit 10 --drive-chunk-size 128M

28

that is not a valid command?
there is no source.
there is no dest

can you post the real commands for win and for wsl2?
is the source the exact same set of files for both?

29

Yeah sorry I didn't put the actual source and dest. But it is the same for both Windows and WSL2.

WSL2

rclone copy --verbose --transfers 1 --checkers 10 --stats 1s --fast-list --tpslimit 10 --drive-chunk-size 128M "/mnt/e/Backup/H23450.mp4" "gdrive-crypt:/Backup"

Windows CMD

C:\rclone>rclone copy --verbose --transfers 1 --checkers 10 --stats 1s --fast-list --tpslimit 10 --drive-chunk-size 128M "e:/Backup/H23450.mp4" "gdrive-crypt:/Backup"

30

thanks, that looks good.
what is the size of the .mp4?

i will create a dummy file of random data the exact size as that .mp4
i will run the same exact commands on win and wsl2.

31

These files are usually between 10 to 200GB.

The one I just tested was 49GB.

32

what are the versions of rclone for both?

33

WSL:

  • rclone v1.52.3
  • os/arch: linux/amd64
  • go version: go1.14.7

Windows:

  • rclone v1.52.3
  • os/arch: windows/amd64
  • go version: go1.14.7
34

My only experience is with Windows, but I can tell you that using "transfers 1", I would not be able to max out my speed, even with a very large "drive-chunk-size". I use at least "transfers 4". Have you tried simply increasing transfers?

35

long story short, in all my testing, win and wsl2 are behaving as expected.
i tested using wasabi a s3 rclone and gdrive, both using crypted remotes.
upload speed were about the same for win and wsl2.

with rclone, depending on the remote, comparing average speeds is often not the best to figure out differences.

for wasabi s3 backend
i uploaded a 35GB .vhdx file to a crypted remote in wasabi, using these flags for both win and wsl2

i tried to mimic your settings tho most of them as not needed for your test and my test using a single file
--verbose --transfers=1 --checkers=10 --stats=1s --fast-list --s3-chunk-size=128M --progress

the first table, for wasbi s3, is how much time rclone spent in minutes, rounded to nearest minute

  • checksum hashing the file before upload
  • uploading the file

wsl2 was slower due to overhead of hashing the file but the actual upload time was the approx. the same

times are minutes

        wsl2	win
hash 	8	    2
upload  7	    6
total	14	    8

here the upload speeds, keep in mind that average final upload speed is not a good indicator of upload speeds as it counts the time spent doing the initial hash checksum.
and these values, based on peak speeds, as expected, matched the chart above, as wsl2 used 7 minutes uploading and win used 6 minutes uploading

speeds are MB/s

        wsl2    win
avg     41	    76
peak    90	   100

i see that gdrive is different then s3 in terms of delay on initial calculation of hash before upload.
there is no delay with gdrive before upload.
i have a free gdrive account, with max of 15GB.
so i create 14GB file of random, non-compressible data.
times are in minutes
speeds are in MB/s
without the initial delay for hashing on gdrive, it is all about upload speeds and the speeds are about the same
but somehow wsl2 was a little faster

        wsl2     win
time    04:15    03:20
speed   58       75
36

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.