Speed capped on macOS App Store Tailscale?

macOS
1

I am experiencing low speeds capped at around 50mbits/s on my Mac at home, tested with iperf.

The problem seems related to the App Store version of Tailscale.
On my office Mac, with a tailscaled installation, I’ve been able to reach much higher speeds during tests with other machines and friends.

Is there any limitation in the App Store version of Tailscale? something related to Apple policies or limitations perhaps?
Anyone else has limited speeds on the App Store Tailscale?

Tailscale is updated to 1.34.2, on a Mac Pro 2013 running Monterey 12.6.2

2

Oh this depends on many factors. E.g. current bandwidth that is used in the networks, current CPU resources used (wireguard still need to decrypt the traffic etc.), Maybe running over a tailscale relay server etc. Etc.

I have not compared the Mac AppStore and the “classic” version of tailscale myself but I don’t think there should be any difference in speed comparison.

3

thanks, here’s a few info:

  • network is 1000/60 at home, and 1000/1000 at the office, so that should be enough
  • tailscale is running directly between the devices I’m testing (checked with the “status” command)
  • CPU on both ends is more than enough, and I tested while everything else was idle or stopped

Meanwhile I uninstalled the App Store version, and I installed the standalone from the official tailscale Packages page.
I uninstalled every file related to the prior installation as per the official instructions, and did a fresh install.
Unfortunately, I’m experiencing the same cap also with this version.

Any ideas?

4

That’s good Infos for Better troubleshooting.

But i’ll stick with the network. I’m going through some basic things here since I don’t know your networking experience. So, sorry if this is more than basic for you and you’re actually more experienced in networking.

Your upload at home is around 60mbit, which represents the speed you’re getting with iperf I guess.

In what direction do you test iperf? Upload / download? Can you post a screenshot of the iperf output?

Also remember that your friends etc. Also need a good connection.

E.g. when you make a download test from your friends computer to your Mac, you will actually download a file from him. So he’s uploading to you (in your case the upload would be: 50-60mbit).

Your office speed is 1gbit symmetrical. So there shouldn’t be any limits you could encounter (if your office isn’t capping the speed ofc)

5

thank you for your reply.
I’m not a complete stranger to networking but I’m not an expert, so thank you for your approach.

I’m running iperf with the -P 8 (or 16) option. I also tried the -R option from time to time.

My friend’s connection is a 1000/300.

Here’s a new test (with -P 8) that I just run from the Mac at the office towards the Mac at home:

[SUM]   0.00-10.00  sec   316 MBytes   265 Mbits/sec                  sender
[SUM]   0.00-10.00  sec   302 MBytes   253 Mbits/sec                  receiver

If I’m not wrong, the office connection could send 1000 in upload (theoretical) and the home connection would receive a 1000 in download (theoretical).
The office connection is a dedicated corporate fiber and it’s only used by my office – so in the weekend there’s no other traffic. The home connection usually stabilizes around ~600/700mbit in normal usage.

So even if the home connection wouldn’t be capable of downloading a full 1000, I guess there’s a lof of difference from the iperf results – even considering wireguard, CPUs, etc.

Also, this is a test (with -P 8) from the Mac at home towards the Mac at the office (so the opposite of the test above):

[SUM]   0.00-10.00  sec  55.7 MBytes  46.7 Mbits/sec                  sender
[SUM]   0.00-10.00  sec  55.1 MBytes  46.2 Mbits/sec                  receiver

The Mac at the office is running tailscaled, and the Mac at home is running the official tailscale package (no App Store)
To me, these numbers don’t add up, but again I’m not an expert and perhaps I’m missing something.
Thanks for helping!

6

Yes, this is correct.

Just did some testing myself and found similiar drops in speed. Never really had an eye on it cause it was always “fast” enough for me.

Office 300/300
Home 120/25

Home → Office (thats ok since 25mbit is my upload speed)
image

Office → Home (not that ok, my Download at home is 120mbit which i usually get)
image

Speedtest at Office says 280/280.
Speedtest at Home says 120/25.

Interestingly, if i download something from my office public webserver (hosted at the office) i only get about 8MB/s (~64Mbit). I think our Office Firewall is scanning the traffic (our firewall is not the fastest to be honest).

copying a file from office to home over SMB i only get 4-5 MB/s (from theoretical 16MB/s)

Also did the same thing with two cloud servers, and got similar results.

I cant test the real wireguard client to check what the speed are with this, but i assume there is some bottlenecking in the tailscale client.

7

thank you for your tests.
I hope someone else will jump in on this and will shed some light

8

The first thing that comes to mind when I read this was are you using WiFi when connecting to both networks with your Mac? If so I would assume the bottleneck is with whatever wireless standard your access points are supporting.

9

I see something similar to what OP reported. My test:

  • Tailscale Version 1.38.2 (App Store) on my macOS laptop
  • Tailscale 1.30.2-300027 on my Synology NAS
  • Throughput test from the laptop to the Synology’s 192.168.X.X address: ~320 Mbps
  • Throughput test from the laptop to the Synology’s 100.X.X.X address: ~17 Mbps
  • Throughput test when the laptop uses the Synology as an exit node (to the Synology’s 192.168.X.X address): ~312 Mbps

The bottleneck seems to be either the macOS Tailscale app or the Synology one (not the network nor the devices).