UDP issues on SuperHub3 - collective thread

[dinth]

Hi. I decided to open a collective thread for UDP issues on SuperHub3, as it seems that many users are pestered by them, the impact is high and Virgin Media is not aware or fails to acknowledge the issue so far (they've been trying to fix my issues for a past year and nobody ever suggested that this has anything to do with UDP packets, until i found whats exactly the issue on my own).

This is the main thread for those issues was this one: https://community.virginmedia.com/t5/Speed/My-350Mb-connection-is-throttled-to-10Mb/td-p/3953746/pag... but it was originally created over a year ago and has a slightly misleading title. 

What we know so far:

1. The issue only occurs on SuperHub3 (and reportedly on SuperHub4) modem. SuperHub2 users are not affected. 

2. The issue occurs if an app/service is using an UDP connection to the internet. Examples of such apps: BBC Iplayer app, most online games, Microsoft Teams, Skype, third party VoIP apps, all kinds of VPN apps, torrent download apps. 

3. All UDP connections are performing very badly on Superhub3 modem, but also whenever there's a significant UDP traffic passing through the modem, other, non-UDP internet connections are slowed down to crawling speeds.

4. Depending on an user, the reported slowdowns are to between 10Mbps and 2kbps (!) speed measured through Speedtest. At the same time ping times go up all the way to 2000+ ms and huge packet losses (sometimes over 40%) occur. In other words, even web browsing is not possible if there's a significant UDP traffic going on through modem. 

5. VM street level fixes do not fix the problem - i had a whole uplink cable between CATV manhole and Virgin Media cabinet replaced by VM and it has not fixed the issue.

6. Replacing a modem does not fix the problem either - replacement SuperHub3 modems also have same issue.

7. Here's how the issue looks like on ThinkBroadband/Speedtest with just 1Mbps of UDP traffic flowing through the modem:

It's really sad, that Virgin Media fails to acknowledge this problem, as hundreds, if not thousands of people may be severly affected by it. The advice VM is giving out to the users is unhelpful and sometimes harmful (I've been told i need to chase the walls in the newly refurbrished house to replace my VM cable and desperate to fix the problem I have done that). Just look how many threads are being open on this forum describing similar issues with apps using UDP connections. 

[jphobbs]

I'm not 100% convinced this is my issue exactly but I've found that unchecking all the options on the firewall section of the admin panel made a big difference in the frequency of outages. Unfortunately they still happen unless I turn off firewall as well. But I can't turn off as that completely breaks port forwarding... 

[M890089]

Getting exactly the same issue here, confirmed only occurs when UDP traffic is above a just a few connections.

SuperHub3 in modem mode

Happens instantly if you start a torrent, like clockwork, on a 350mb line, it drops to just a few kbps when it occurs, stop the torrent, few seconds later it revives, that is if it doesnt halt all traffic completely

When it happens i also lose ability to get onto the superhubs web gui until i reboot it. I have now setup a schedule using a python script to reboot the superhub every night, obviously doesnt get around heavier UDP usage,. but at least im ready to go in the morning.

No feasable option for UDP usage other than using a TCP only VPN Tunnel.

How have VM not commented in this thread, thumbs up to OP for starting it with such detail, hope it gains enough traction to get VM involved, this is embarrasing for them and if they are going to be so awkward as to limit us to use their cheap crappy hubs in modem mode when we want to run our own high end network kit then the least they could do is make sure it works properly. Working as an IT Tech supporting a load of homeworkers its become apparent that this is happening to lot of people.

VIRGIN LISTEN TO YOUR CUSTOMERS! Stop replacing and recabling stuff all over the place and look at the shoddy kit your giving people instead, please for everyone sake refer this to your technicians properly, show them this thread, anyone with any networking knowledge can confirm this is 100% the issue.

[jgtimperley]

I got an update to 9.1.1912.304 overnight, which is a minor bump from 9.1.1912.302, and relatively close to the last update.

I purposefully avoid unnecesarry UDP traffic since discovering the original issue documented here, but wondering if anyone else on here has the new firmware, and is seeing any improvements. I'm not in a rush to test, as we both rely heavily on Teams/Zoom calls, which have been spot on since reducing UDP flows.

[Avalon-One]

Just to echo the same as the above, UDP traffic brings my connection to a crawl, routing traffic via a TCP VPN tunnel works, but obviously at a cost, I would be interested to know if VM are offering any facility for users to downgrade to a SH2 (I can hope can't I?) as frankly the SH3 isn't fit for purpose as it stands and the SH4 is seemingly no better from previous comments.

[dinth]

For now, they have advised me that they can escalate my case to Ofcom, will need to call them after Xmas to proceed with that. They haven't offered me SH2, although I'm not sure if SH2 would be able to handle my 350Mb/s connection.

[Rod-IT]

I'm joining this topic, I've had my own both here and over on Netgate, Netgate say it's the ISP, the ISP refuses to acknowledge technical peoples feedback, instead brushing it under the carpet. I do appreciate that staff who moderate and participate in the community are likely not as technical as some of us posting the issues, but please note this to take it back to those who are - even a statement - yes we are aware and we are working on this - would be a start. Better yet, allow us to buy our own DOCSIS modems.

Let's be clear here, this IS a physical hardware issue, no end of tweaks is going to 'fix it' only mask it further.

If a consumer router used with the ISPs modem in modem mode works, why can't PfSense, what are consumer routers and the modem doing to mitigate this?

I'm sure i am not alone here, but i restrict outbound traffic as well as inbound, hence the need for something more substantial than the routers you can buy off the shelf, even with DD-WRT the options are limited.

I've seen posts where PfSense 2.5 helps or in some cases removes the latencies, but this could be another red herring too, I've read posts about using Unbound in forwarder mode (not the DNS forwarder), that doesn't help for me, I also do not torrent and if i did, it's seldom, so UDP packets from this are minimal.

Like many others, i have no issues with speeds, if i ask for a file, i get it at the speed of the ISP or site, whichever is the slower, likewise, uploads are also fine, i do apply a traffic shaping policy to keep buffer bloat down, but my issue isn't a busy link, i can have latencies and packet loss (like others), every hour or so, even when the connection is almost idle, it is most noticeable when on a call, videocall or streaming - including YouTube.

Also like many people posting here, my network is not a simple plug and play setup, I doubt I'd be here if it was, I invest a lot of time restricting devices, segregating networks and keeping stuff out of my network, as well as stopping stuff talking out to unknown locations, especially IoT devices.

Others have mentioned this too, but i dont get how this affects modem only devices, surely the Puma chipset is disengaged from filtering packets and instead simply redirects them - if it was load, I'd expect to see this only in high saturation, but as above i can see packet loss almost every hour for about 15-20 seconds on an almost idle connection.

I'm paying for more speed but getting less stability, I've thought about going to 1Gig, 500-600MB under the SH4, but that only moves the issue, I've thought about downgrading, but I upgraded because it was cheaper in a bundle, downgrading could end up costing more more as well as reducing what i have in favour of stability.

If VM want to send me a SH4, I'll gladly test it.

FYI, after about 2 days, i can no longer ping the modem at all unless i reset it, then i'll get 1-2 days before it stops replying again.

[gitty]

---

@Rod-IT wrote:

Others have mentioned this too, but i dont get how this affects modem only devices

---

I cannot tell you what the reason to this issue is, but a lot of stuff happens inside a DOCSIS modem. It is aware of the traffic from the upper layers and can even do filtering (protocols, port numbers, etc).

[Rod-IT]

Not going to disagree that it's doing something, but in bridged mode, why does it get involved, isn't the whole purpose of bridged mode, to remove the router from doing anything except forwarding the packets on to the next device?

I wouldn't expect to give any of my staff a task to do, but then pre-do half of it before i give it to them.

It's for these reasons and to help answer questions to people like myself and others here, that technical people at VM, Intel or the router maker should be here posting to at least give us something meaningful in return.

[gitty]

It is a DOCSIS thing. A standalone modem does the same. A router does a lot of other things.

[Rod-IT]

I'm going to see if i can find anything on that.