My 350Mb connection is throttled to 10Mb

[ProTofik]

Hi,

I have been Virgin Media (and NTL) customer for the past 12 years. Starting with 2Mb/s connection, I went through multiple upgrades, currently at 350Mb/s.

In the past weeks or maybe months, I have noticed that as soon as I turn torrenting on any computer or server in my house on, the WAN download connection drops to 10Mb/s flat on every device in the house (total). On top of that, I experience massive, 25% packet drops (tested by ping -t 1.1.1.1 and 8.8.8.8). Turn torrent off, and bam, problem solved.

My SuperHub 3 is setup in modem mode. My edge firewall is virtualized pfSense 2.4.4 and VLAN routing is done on Cisco 3560CX. Pretty basic setup that works fine.

Has Virgin Media introduced some sort of policy recently that throttles customer's connection down to make it feel like 2007 again for torrenting, or is my SuperHub 3 modem acting up?

Paying over £50 for 10Mb connection does not seem like a good deal for me any more. There are unthrottled 70/20 options out there for half the price.

[DreamOfCheese]

---

@fyah wrote:

I get exactly the same issue with Utorrent during peak times & soon as I pause torrents it speeds up.

Reducing to 100 connections doesnt seem to made a difference.

Virgin secretly throttling torrents.

---

If they're secretly throttling torrents they're doing it by random selection of customers, as I don't seem to be affected by this.

Just tried with the largest legal torrent I could find by means of quickly searching, it maintained >300mbps at all times throughout the download period.

Not to say you didn't have problems downloading, just that it's not some secret scheme to throttle customers.

[chaddock]

---

@DreamOfCheese wrote:

Not to say you didn't have problems downloading, just that it's not some secret scheme to throttle customers.

---

How is he triggering it with opening a torrent connection? For all devices? Just coincidence that he has noise on the line at the exact same time? Sounds like traffic management at the very least. I've been with Virgin in the past, and I've had **bleep**-slow connections in peak-times, reliably, same times every day... that was 3 years ago, and it's still rife today.

Virgin are oversubscribed, and there's no recourse for Virgin giving **bleep** service so why should they stop? 3 days of <50% speed in a row and you can cancel the contract... great, no incentive for Virgin to act right though...just means the customer has to get re-connected with some else.

Your lucky you dont have a duff connection, but these forums are a testament to everyone who has.

[backslash2k4]

I don't mean to necro this post, but after searching for over a week looking for answers into why my SH3 has started choking, I came across this thread; and it looks like I may have some vague idea as to why when a large number of connections are opened simultaneously and data transferred over them the hub chokes, and I believe this has something to do with one of the latest firmware's deployed to the device to resolve the Puma 6 chipset controversy, which has moved the issue elsewhere into the hub.

For some background, I'm a developer and I have a pet project in developing an MMO. I used to run unit tests locally and remotely to profile and improve the networking of my engine, of which used to be no problem. Recently I started to notice massive erroneous values in the data sets of my unit tests when running dev server builds out of my own home, wherein the data being sent would be mysteriously held back for inexplicable reasons, and my connection would choke. I started investigating whether it was my router, a software change I made, etc. It wasn't until I was rebuilding one of our servers and downloading a Debian-based distro using Transmission that I noticed similar behaviour and looked into whether there was something peculiar going on with my network. I swapped out my router out for a far more powerful pfSense switch, and after attempting the same unit tests internal of my network and some external, and I came to the conclusion that it was the SH3 (running in modem-mode in my instance).

First I thought it might be overheating, as it was getting toasty, so strapped a fan to the side of it - no change. I looked into incrementally increasing the packet sizes of the data being sent over the connections I was opening, and the number of connections to ascertain if that was an issue. Packet size seemed to make no discernible difference (sizes varied between 60-90 bytes on average), and the connection number seemed to cause the issue to trigger more often, but didn't seem to relate immediately to the issue. As an example, I opened 500 concurrent connections and it seemed to perform admirably for a period, and then the connection as a whole would choke out, and recover somewhere between 30 to 180 seconds later - these connections were still open, so it didn't make sense as to why it would choke. Once the connection recovered, it would work perfectly for a period of time, after which the entire choke, recover, cycle would start again.

I then started encrypting all of the packets to change the make-up of them, to ascertain perhaps if what I was doing was triggering some sort of throttling or the data I was sending/receiving was causing a false positive in a basic SPI, and that's when the issue went away (previously this was all un-encrypted traffic). Now, I assume the Puma 6 does some form of traffic prioritisation, because you can ping any particular site or endpoint and the SH3 at the same time, and the hub will throw ridiculous numbers at you at times (or not respond at all), and the external site will be unaffected. This can be triggered more often than not by attempting to launch the web-server on the device (192.168.100.1 in modem-mode or 192.168.0.1 in router-mode) when performing a ping to the device (it will stop responding).

Now those experienced in the old issues with the Puma 6 chipset prior to the firmware fix used to see occasional spikes when it would perform some sort of internal maintenance and drive the CPU usage high, I wonder if it's this same process which is still occurring along with some rudimentary packet filtering, that causes the CPU usage to spike, and causes the SH3 to fall behind when the connection is being utilised in a particular way, and of which it struggles to recover from.

Encrypting the packets may cause any rudimentary packet inspection to fail-out of the inspection far quicker, and utilise less CPU. Anyway, the above is just a hypothesis - I don't intend to continue performing any further development tests out of my home connection as this has caused me to waste far more time than I would have liked initially looking into it. Hopefully the above provides some information to those who suffer with similar issues - I imagine VM are fully aware of the shortcomings of the SH3, and indeed the requirement for a better replacement in the future.

[ProTofik]

Thank for the reply, but I gave up on it here. Virgin has completely ignored our suggestions in this thread, engineer sent to my house said they don't support modem mode and I just have to deal with it. This is absolutely a software issue on superhub 3 but virgin won't acknowledge it. Why do it when you can just sit quiet?

Anyway it doesn't bother me anymore because BT's Ultrafast Fibre Plus 2 is rolling out soon in my area. After 13 years with virgin I will be finally making a jump.

[dcapon]

I realise this maybe a dead topic but I just wanted to add I have the exact same issue again in modem mode and I have found that 9 times out of 10 running a speedtest at speedtest.net opens up full speed to the network for the duration of the speed test and then throttles back down to 10 after. The torrent download graph shoots up with almost every click of the go button. Maybe I'm overly suspicious but this seems like intentional throttling with an attempted workaround to save face.

[atissue]

I'm a new customer to VM after many years with BT, with their latest being the 330/30 gfast service. It was ace until I've moved a street away which is just over the 300m limit, forcing me to take up a VM package. 

I'm in modem mode and getting the same thing. As soon as a torrent is started (torrents aren't illegal. illegal torrents are illegal) my connection shortly drops to 10mbps. This is the most replicatable scenario but also happens many times throughout the day. I have literally had a brand new cable ran through the street, new terminations at the property and cab - this ain't a simple 'bad connection'.

See the traffic graphs? Have a guess when i stop/start a simple torrent download of a *nix distro. This is rediculous.

[atissue]

I'm a new customer to VM after many years with BT, with their latest being the 330/30 gfast service. It was ace until I've moved a street away which is just over the 300m limit, forcing me to take up a VM package.

I'm in modem mode and getting the same thing. As soon as a torrent is started (torrents aren't illegal. illegal torrents are illegal) my connection shortly drops to 10mbps. This is the most replicatable scenario but also happens many times throughout the day. I have literally had a brand new cable ran through the street, new terminations at the property and cab - this ain't a simple 'bad connection'.

See the traffic graphs? Have a guess when i stop/start a simple torrent download of a *nix distro. This is rediculous.

[atissue]

So after some investigation this seems to point to the good old issue of the SH3 and it's intel puma chip, or really bad internal QoS

My bets are on state management as it seems to not be able to hold over 100-200 connections reliably - while torrenting or any p2p/high thread traffic for that matter we're saturating the SH3's CPU, it manages this by keeping the connections open but at a significantly lower throughput, shoving a single thread HTTP connection through seems to drop pretty much everything throughput wise, then 5-10s later the single thread HTTP conn speeds up to it's 385mbps link rate and the previous torrent connections drop.

Limiting p2p connections to a max of 20 eliminates this issue for the most part but with a significantly fluctuating ping and jitter. Happens in both modem mode and router mode, and also with my own network gear with a double nat.

Anyone fancy downgrading me to a 200mbps service and shipping me a SH2ac? I think I might shortly be pulling all strings to cancel and going with a couple of VDSL2 lines loadbalanced...

[DavidJWalker]

one they will not ship out a SH2ac they do not have them any more. P2P is managemented on virgin media network even other network openreach. but P2P is low on management on virgin media its a alot Higher on openrech network.

ps torrents is management on every connection and it is monitored thats why i do not use Torrents i used to use it alot on when i was on Openrech connection before it starts to Throttle every company does it now

[atissue]

It may be managed - but it should not effect my other transfers, eg a single thread HTTP connection. It is also effected using a 36/72 thread SFTP transfer, something I do daily.

Something is very funny with the whole modem mode/traffic management. I've settled on router mode with a double NAT, with all firewalls disabled and my own edge router in the DMZ. The hub gets assigned a different public IP, higher line rates and seems generally more stable between throttled limits. Seems like a bit of a kick in the nuts to people who actually know what they're doing, and want to run their own kit which is probably more expensive than the switching gear in the cab...

To the OP, and whoever else gets this and has a decent sense around networking. Do the frowned upon and double nat, it still isn't great, but with a few tweaks to limit max conns to 500-1000 in whatever router you're wanting to use, run the hub in router mode with firewalls/upnp/literally everything else disabled but your router in the DMZ. It somewhat works, and I can download torrents + run speedtests without my network grinding down to unusable speeds every 2 minutes.

Fun.