Virgin media throttling openvpn & p2p any SH3.0 work arounds ?

[mav2k]

Hi all, I found this thread when searching for VM throttling my connection when downloading torrents. I would appreciate any help or insight you guys can provide on the following please, and whilst I am by no means a novice, I do not consider myself an expert or advanced in anyway on this topic.

I have a 350mbit Virgin Media broadband cable connection through their 3.0 hub in modem mode and then connected directly to my TP-Link Archer C9 cable router.

This has been my setup for the last 2 years and have not had any issues with speed capping, slow down or any dropped connections ever.

That is until recently, where I have now noticed if I attempt to download ANY torrent, my connection is throttled to less than 1% of my speed, in actual fact when I do a speed test with the speedtest app on my phone, it shows as between 512kbps and 1mbps bandwidth.

However, as soon as I stop the torrent, within 10 seconds my speed returns to normal and have no issues. I first thought my router had broke in some way, but having rebooted it, checked I was using the latest firmware which I was, and after monitoring the CPU and memory usage of the router whilst performing this test, the router is not stressed in anyway, so I do not believe my router is to blame.

I have been using qBittorrent for a long time, with no setting change, but just to be sure, I reduced the number of global maximum connections to just 150 from the default 500, which had no effect. This was also just trying to download a single torrent, with around 10 seeds and a few leeching peers.

Today after being very frustrated at this, I decided to try using a VPN, just to see what would happen. Guess what, speed remains high when downloading torrents, no slow down, no throttling, perfect as it has always been.

Now am I missing something here, but does this not suggest VM are currently severley throttling my connection if they detect a torrent being downloaded? Which they say they do not do?

Any ideas or help would be much appreciated.

Thank you.

 

[Kashinoda]

If you still have Mullvad, use WireGuard instead. It’s much faster than OpenVPN, especially on Linux but on any system really. It’s multithreaded for a start, unlike OpenVPN.

Bit of a late reply but this is indeed the ticket, the overhead using Wireguard is significantly lower than other solutions. I was having a nightmare with OpenVPN and IKEv2 with IPSec, torrenting with anything more than 50 connections and Superhub 3 would **** the bed.

With Wireguard it flies. For anyone wanting a simple install script you can grab one here:
https://github.com/angristan/wireguard-install

That said it's relatively easy to manually setup, assuming you have your own VPS or dedicated host somewhere ourside the UK.

 

[Avalon]

Bit of a late reply but this is indeed the ticket, the overhead using Wireguard is significantly lower than other solutions. I was having a nightmare with OpenVPN and IKEv2 with IPSec, torrenting with anything more than 50 connections and Superhub 3 would **** the bed.

With Wireguard it flies. For anyone wanting a simple install script you can grab one here:
https://github.com/angristan/wireguard-install

That said it's relatively easy to manually setup, assuming you have your own VPS or dedicated host somewhere ourside the UK.

OK i’ll play along... Since when did the SH3 have known issues with 50+ connections? Let alone 50+ connections over what it only sees as one connection because it’s running over a VPN tunnel to an end point?

 

[Kashinoda]

OK i’ll play along... Since when did the SH3 have known issues with 50+ connections? Let alone 50+ connections over what it only sees as one connection because it’s running over a VPN tunnel to an end point?

I have limited knowledge in this area but it's not seeing a 'single' connection, packets are sent to and from as normal when using a VPN tunnel. The difference being the packets are encrypted and sent to the VPN host.

Depending on the technology used the overhead (I.e. the encryption added to the header / footer of the packet) can vary dramatically. Wireguard is extremely light compared to traditional solutions.

I might off a little on that but that's my understanding. What is true is that SuperHub 3 falls down when using OpenVPN with torrenting, plenty about that online.

 

[Avalon]

I have limited knowledge in this area but it's not seeing a 'single' connection, packets are sent to and from as normal when using a VPN tunnel. The difference being the packets are encrypted and sent to the VPN host.

If you read that back, you may see a contradiction. Let’s say you have 50 peers, you send packets to each of them, they are encrypted locally and sent over a VPN tunnel to the VPN end point. The SH3 literally sees data packets to the end point, it doesn’t know what they are, where they’re ultimately going etc. as it’s encrypted data.

As to SH3 issues with OpenVPN, again I can’t see a technical reason for what you suggest you’ve seen, more importantly I don’t see it when I use OpenVPN or Wireguard over a SH3 at near line speed and the people I know using said connection now have reported zero issues. I’m not saying they don’t exist, but I would question if it’s an SH3 issue or a user issue, OpenVPN is single threaded, without using hardware acceleration, it doesn’t scale well at all, and as this thread shows, people are quick to come up with ‘theories’ and jump to illogical conclusions.

 

[Kashinoda]

If you read that back, you may see a contradiction. Let’s say you have 50 peers, you send packets to each of them, they are encrypted locally and sent over a VPN tunnel to the VPN end point. The SH3 literally sees data packets to the end point, it doesn’t know what they are, where they’re ultimately going etc. as it’s encrypted data.

As to SH3 issues with OpenVPN, again I can’t see a technical reason for what you suggest you’ve seen, more importantly I don’t see it when I use OpenVPN or Wireguard over a SH3 at near line speed and the people I know using said connection now have reported zero issues. I’m not saying they don’t exist, but I would question if it’s an SH3 issue or a user issue, OpenVPN is single threaded, without using hardware acceleration, it doesn’t scale well at all, and as this thread shows, people are quick to come up with ‘theories’ and jump to illogical conclusions.

It's not a contradiction, I was implying the overhead added to each packet caused the SuperHub 3 to fall over when torrenting with that many connections. That may not be the actual cause, looks like a likely UDP issue: https://www.reddit.com/r/VirginMedia/comments/eyd48r/throttling_through_vpn_torrenting/

To clarify here I have no issue with OpenVPN downloading at maximum speed outside of torrenting, it has no negative impact on my line - same with IKEv2 with IPSec. However when using either of these when torrenting I can observe significant impact to internet performance elsewhere on my local network:

400Mb Line

Config 1
PC1: No VPN, torrenting @ 1MB/Sec download and 5KB/Sec upload
PC2: No performance issues.

Config 1
PC1: Wireguard VPN, torrenting @ 1MB/Sec download and 5KB/Sec upload
PC2: No performance issues.

Config 2
PC1: OpenVPN or IKEv2 with IPSec, torrenting @ 1MB/Sec download, 5KB/Sec upload
PC2: Ping timeouts and spikes to 200 when pinging bbc.co.uk, speed massively throttled

There is definitely something up with SH3 here. The host for all VPNs here is a dedicated 4790K with 32GB of RAM in a 1000Mb up/down line in a datacenter.

 

[Avalon]

It's not a contradiction, I was implying the overhead added to each packet caused the SuperHub 3 to fall over when torrenting with that many connections. That may not be the actual cause, looks like a likely UDP issue: https://www.reddit.com/r/VirginMedia/comments/eyd48r/throttling_through_vpn_torrenting/

To clarify here I have no issue with OpenVPN downloading at maximum speed outside of torrenting, it has no negative impact on my line - same with IKEv2 with IPSec. However when using either of these when torrenting I can observe significant impact to internet performance elsewhere on my local network:

400Mb Line

Config 1
PC1: No VPN, torrenting @ 1MB/Sec download and 5KB/Sec upload
PC2: No performance issues.

Config 1
PC1: Wireguard VPN, torrenting @ 1MB/Sec download and 5KB/Sec upload
PC2: No performance issues.

Config 2
PC1: OpenVPN or IKEv2 with IPSec, torrenting @ 1MB/Sec download, 5KB/Sec upload
PC2: Ping timeouts and spikes to 200 when pinging bbc.co.uk, speed massively throttled

There is definitely something up with SH3 here. The host for all VPNs here is a dedicated 4790K with 32GB of RAM in a 1000Mb up/down line in a datacenter.

I’ll assume you either didn’t read the reddit thread you linked to (google will find supporting evidence for all sorts of crazy stuff if you look), or supplied it for comedy value? If not it’s another great example of wild unproven supposition from people who clearly have just enough knowledge to post on reddit so someone can use it as justification to support an equally unlikely theory. Seemingly the logic is if you get slow speeds, it’s not your PC, OpenVPN’s inherent limitations in software, the protocol used, the encryption type/strength, the connection to the router, the routing/peering to the remote end point, the remote end point’s load/WAN/routing/peering, if port forwarding is supported by the end point and if the op is using it, it’s just got to be VM are throttling and/or the SH3 (though if you read the comments it even happened on the SH2 but just ignore that, because facts don’t support predetermined conclusion).

I like facts. I like people who understand variables and the need to eliminate them through testing to prove things. This isn’t that, this is firmly in Volcano + Monkey = Vauxhall Corsa with added tin foil hats all round territory. What you have posted doesn’t doesn’t prove your conclusion, it just proves that when you change multiple variables, you get a different result. Sadly, someone will probably reference it at some point in the future as supporting evidence that other people have the same issue. Still at least this time (unlike the reddit thread) people didn’t claim VM were monitoring downloads in real time and cutting download speeds by 75%, because obviously both of those happen

 

[Kashinoda]

Reddit thread aside I provided my own facts. My testing methodology was the same for each case:

• PC1: In qBittorrent set the speed cap to 1MB/sec download and 5KB/sec upload.
• PC1: Set connection limit to 100
• PC1: Load up qBittorrent with Ubuntu distros to ensure maximum connections and speed reached
• PC2: Ping bbc.co.uk on loop.

Both computers connected via ethernet directly to SuperHub 3, the only variable I change is the VPN type on PC1:

• OpenVPN: PC2 suffered significantly degraded internet performance.
• IKEv2 with IPSec: PC2 suffered significantly degraded internet performance.
  
• Wireguard: PC2 suffered no noticable impact on internet performance.
• No VPN: PC2 suffered no noticable impact on internet performance.

All of this was repeatable. The VPN host on each case was the same dedicated machine running Ubuntu Server 18.04 LTS with a 4790k CPU.

Something is clearly going wrong with the SuperHub 3, PC2 isn't running through a tunnel and is getting severe problems accessing the internet. Did I do something wrong setting up both OpenVPN or IKEv2 with IPSec? Possible for sure. It would be useful to have another ISP and modem on hand to test that, I dont.

What I do have is plenty of threads from people with varying degrees of knowledge who have the same issue. I also have the knowledge of knowing the SuperHub 3 has been plagued with issues and Virgin Media spent the best part of 2 years getting it into a workable state thanks to the class-actioned POS CPU that drives it.

Anectodal? Sure why not. But for anyone reading this thread, you can solve your VPN torrent throttling issues by using a VPN that offers Wireguard as an endpoint. As to why the SuperHub 3 can't handle these other proticols, who knows.

 

[Avalon]

So far we've gone from the SH3 not being able to handle 50 connections to remote hosts (that it only sees as one), to you referencing a reddit thread that you suggested supported your conclusion of a UDP issue, but if you read the thread - and it's genuinely cringeworthy in places - one of the participants clearly states they have a SH2 and therefore you can't blame the Puma chipset, even ignoring them for the moment, if it's a UDP issue, have you looked at what Wireguard uses? Hint: It starts with a U and ends in a P

I'm not disagreeing you have an issue on a technical level, heck it may even be the SH3's UDP issue (have you read mackey's thread when he first flagged the issue?), it could equally be many, many other variables you've chosen to ignore. All you've proven is that Wireguard is faster for you on your PC/host, that's something that was never, ever in dispute. What you haven't proved is why and and the conjecture and reasoning in previous posts isn't helping.