Terrible traffic shaping on Vodafone

Over the last 3 months, there have been a number of complaints of slowdowns on the Vodafone network,

"Tech Support" (clueless script readers) claim no problems, however there is awful traffic shaping taking place within the Vodafone network. "You sync and 79Mb sir, therefore you are getting the fastest speed possible - there is no problem sir".


Speed tests - are always high (it seems they prioritise that traffic).
Youtube / Netflix / Amazon - always fast. (It seems they prioritise that traffic)
Single thread downloads - run somewhere between 2MB/s and 0.3MB/s between approx 4pm and midnight.
Multi thread downloads - run better between 5MB/s and 1MB/s between approx 4pm and midnight.
Easynews/ NNTP in general is pretty much 0.5MB/s all day.


However - stick all the above through a decent VPN provider and i get full line speed 7.8MB/s So there does appear to be aggressive traffic shaping taking place.

Anyone know the OFCOM policies on Shaping - I understand its acceptable to manage extreme network conditions - but this is everyday implementation to shape as much traffic as possible, now I'm pretty sure this is not allowed under the OFCOM code of practice.


Thoughts?

Edited by Tam (Fri 21-Dec-18 22:45:57)

Hi.
Finally found someone with the exact issue i am having.
Been with Vodafone for past couple of years and took out a new contract two or three months ago and since then (although the timing may be coincidence) my actual download speed has been awful and so unreliable. Updates on Xbox are taking hours rather than mins. Been in touch with Vodafone and get he standard reading from a sheet bull!!!!. Changing channels on the wifi etc. Not really sure what to do next. Probably Leave!
Can anything be done?

Switch providers and sell them your not staying in contract because of your issues not much else to it really.

Vodafone is an extremely congested service at current, with a bunch of problems... They are a recent player to the ISP market in the UK, after deciding to re-enter this space. You will get no traffic shaping with the more established players, BT, Sky, TalkTalk etc.

If you complain and threaten ombudsman, I foresee them allowing you to terminate the contract.

Can you run a single threaded speedtest at www.speedtest.net
At the bottom of the page below GO you will see connections, press it so it changes to single... then run the test.

Edited by ukhardy07 (Sun 23-Dec-18 02:30:31)

In reply to a post by kebabman:

Hi.
Finally found someone with the exact issue i am having.
Been with Vodafone for past couple of years and took out a new contract two or three months ago and since then (although the timing may be coincidence) my actual download speed has been awful and so unreliable. Updates on Xbox are taking hours rather than mins. Been in touch with Vodafone and get he standard reading from a sheet bull!!!!. Changing channels on the wifi etc. Not really sure what to do next. Probably Leave!
Can anything be done?

Nope extremely common with VF and has been for over a year. I've got BT at a few properties now and no issues with the speeds... As far as mass market goes I would stick with Talktalk, sky, BT, Plusnet etc, they are well established and have ironed out these issues. Headend congestion is almost unheard of.

Ones to avoid, Vodafone, Post Office, SSE, Zen (recently had issues - traditionally very good).

Single thread test this very site runs https://www.thinkbroadband.com/speedtest has been showing the problems with Vodafone for some time, but it is a mixed bag i.e. some do get a good service

By running the test here the results do get shared in various monthly summaries, which ISP do seem to worry about (sometimes)

https://www.thinkbroadband.com/speedtest/15455638705...

http://www.speedtest.net/result/7902792572

http://www.speedtest.net/result/7902920355

Edited by kebabman (Sun 23-Dec-18 14:04:36)

Curiously, VF state they don't do any traffic shaping.

It has been suggested in to the VF community they are some messed up routing as well, causing traffic to route to america and back as well.

Looked into someone who was moaning about the routing, but they were looking at a specific Netflix node which may have routed via US

Nothing suggests US routing when we look at the IP of Vodafone customers, what we see is lots of congestion and have told Vodafone this

I don't think Vodafone are traffic shaping, I think this is classic congestion.

All your speedtests both here and Speedtest.net show congestion.

The only real solution is a switch of provider, are you locked in contract?

It feels like congestion to me, I'd not experienced any dodgy routing, but I am not Netflix customer.

Curiously though, VF keep denying it is congestion related and event sent me a new router after they declared the old one didn't like the dark!

The new one has the worst wifi I've ever seem - connection drops about 10 meters away from it!

So are we saying its congested all of the time. Would that not have an effect on overall download speed and not just single thread or am i barking up the wrong tree.

As the Op said it seem be prioritised in someway. I can watch 4k stream but not download a software patch at any kind of consistent speed.
I am in a contract that I am trying to end.

Single thread is a sensitive way of spotting congestion, in time the general speed test result is likely to start dipping

The 4K stream might be being served from a cache inside the Vodafone network, or a better link out of the network

Its *NOT* congestion, if it was congestion then *everything* would be slow -

A speed test shows single thread 10Mb but muti-thread 70Mb
https://www.thinkbroadband.com/speedtest/15455947657...

A download shows at 1-2MB/sec, yet stick it through a VPN and *boom* full line speed 8MB/sec download speeds.

This is a badly managed network and/or over harsh traffic shaping (something they should not be doing) which is limiting traffic speeds in order to keep the network performing correctly for streaming video so that most users don't notice.

I feel an ofcom complaint coming on...

I have to say with the experience I am having I completely agree with this comment. If congestion was the factor it would show different results at different times of day but my results are consistent. Actual download speeds are unreliable at best.

What result does speedtest.net give you for single thread (single connection) speedtest?

This is what the TBB tester churns out for my 330 Mb/s line:



Yet speedtest.net single thread results are totally different:

Result1: http://www.speedtest.net/result/7903582302
Result2: http://www.speedtest.net/result/7903601639

Conclusion: Single Thread speedtests results on TBB may not always be correct, ie use a second tester (such as the one on speedtest.net) to cross check your results.

Edited by baby_frogmella (Sun 23-Dec-18 20:29:20)

Single Thread speedtests results on speediest.net may not always be correct

In reply to a post by MrSaffron:

Single Thread speedtests results on speediest.net may not always be correct

True, but there is a choice of servers on speedtest.net hence why if 2 or more different servers there give you the same single thread result and the one on TBB doesn't, then you know which one is incorrect

Its *NOT* congestion, if it was congestion then *everything* would be slow

This is not true, you do not understand how the internet works if you believe all traffic is routed identically.

When you use the VPN, then yes all your traffic is routed via the VPN, and that route is likely not congested.

Your speedtests show very very clear congestion and your ISP is known for congestion widely.

Read the traffic management document published:

http://www.vodafone.co.uk/cs/groups/public/documents...

NO traffic is managed. They cannot lie about this... It is documented and formally captured.

Edited by ukhardy07 (Sun 23-Dec-18 22:04:13)

Multi threaded hides congestion.

Notice some of my speedtests here, both yellow and green lines are the same, this is showing no congestion:
https://www.thinkbroadband.com/speedtest/14770058691...
https://www.thinkbroadband.com/speedtest/14833184342...
https://www.thinkbroadband.com/speedtest/15456040365...

Compare this with yours
https://www.thinkbroadband.com/speedtest/15455638705...

Notice the green line is way below the multi threaded.
Yours shows clear congestion here.

If you imagine you had 10Mbps shared between 2 people.
2 single threaded downloads would be 5Mbps each.
Now say you download something with 20 threads, and the other person has 1 thread, you will suddenly get the lions share of the connection.

This is why multi threaded "hides" congestion...

A speed test shows single thread 10Mb but muti-thread 70Mb
https://www.thinkbroadband.com/speedtest/15455947657...

That is excessive congestion, please see my other post in relation to this.

In reply to a post by ukhardy07:

Its *NOT* congestion, if it was congestion then *everything* would be slow

This is not true, you do not understand how the internet works if you believe all traffic is routed identically.

When you use the VPN, then yes all your traffic is routed via the VPN, and that route is likely not congested.

Your speedtests show very very clear congestion and your ISP is known for congestion widely.

Read the traffic management document published:

http://www.vodafone.co.uk/cs/groups/public/documents...

NO traffic is managed. They cannot lie about this... It is documented and formally captured.

1. You've linked me to the mobile traffic terms, completly different, even ofcom treat it differently.

2. I have access to several hundred hosted servers within microsoftcloud and google's cloud, a file download from these performs slower than if I VPN to the same server and download the same file, both the route there and the route back is identical as I own both ends and therefore can see the route traffic takes in each direction, I hope to have a spare server that i can do a wireshark capture to full analyse after the xmas break. Equally I can bouce my traffic off over 100 different UK VPN servers using 73 different routes, all demonstrate the same issue. Encrypted VPN traffic always performs at a better speed to the any download source sampled than a direct connection.

My speed test does not show congestion, if there was, then all tests would show a throughput hit, my speed test clearly shows there isn't congestion as I can obtain maximum throughput.


All ISP's employ traffic management, it would be iresponsible for them not too - this allows them to keep the network running if/when they have a major fault affecting their network, in these cases they can apply QOS to ensure that specific traffic can get through the congestion.

The question is - at what point does Vodafone move from deploying their solution (which I am sure they currently are) in order to protect their network to the stage of just leaving it deployed as it means they don't need to invest in the backend infrastructure to support their growth and their real bandwidth needs.


Finally - back to my point - i get less than 1MB/sec for *any* nntp traffic no matter where its hosted, compared with full 8MB/sec when hidden behind a vpn even a VPN to the same nntp server (so highly likely to be indentical router there and back as non VPN traffic) - that is not congestion, its not bad luck, its not the direction of the wind, it is only that speed for 1 reason, and that is that it has been prioritised much lower than all other traffic.
.

Edited by Tam (Sun 23-Dec-18 22:54:29)

1. You've linked me to the mobile traffic terms, completly different, even ofcom treat it differently.

Sorry on mobile: https://www.vodafone.co.uk/cs/groups/public/document...
The statement stands.

Note the document includes "This KFI gives an overview of typical traffic management practices undertaken on this product; it does not cover circumstances where exceptional external events may impact on network congestion level."

2. I have access to several hundred hosted servers within microsoftcloud and google's cloud, a file download from these performs slower than if I VPN to the same server and download the same file, both the route there and the route back is identical as I own both ends and therefore can see the route traffic takes in each direction, I hope to have a spare server that i can do a wireshark capture to full analyse after the xmas break. Equally I can bouce my traffic off over 100 different UK VPN servers using 73 different routes, all demonstrate the same issue. Encrypted VPN traffic always performs at a better speed to the any download source sampled than a direct connection.

That does not prove anything as a VPN is routing traffic differently.

My speed test does not show congestion, if there was, then all tests would show a throughput hit, my speed test clearly shows there isn't congestion as I can obtain maximum throughput.

Your speedtest shows absolute crisp clear congestion, others here will agree when they come online and read this. Your single thread is 10Mbps, your multi is 70, that is congestion.

All ISP's employ traffic management, it would be iresponsible for them not too - this allows them to keep the network running if/when they have a major fault affecting their network, in these cases they can apply QOS to ensure that specific traffic can get through the congestion.

Find me ANY mass market ISP (talktalk, sky, plusnet, BT, EE) where they have traffic management and document this as required. This needs to be in the form required per the voluntary code of conduct. The statement "All ISP's employ traffic management" is false. Almost none do anymore... I have been in this game for 11+ years, and I have worked with almost every big name provider from an infrastructure standpoint.

The question is - at what point does Vodafone move from deploying their solution (which I am sure they currently are) in order to protect their network to the stage of just leaving it deployed as it means they don't need to invest in the backend infrastructure to support their growth and their real bandwidth needs.

They do not deploy a traffic management solution.

Edited by ukhardy07 (Sun 23-Dec-18 23:05:02)

In reply to a post by ukhardy07:

1. You've linked me to the mobile traffic terms, completely different, even ofcom treat it differently.

Sorry on mobile: https://www.vodafone.co.uk/cs/groups/public/document...
The statement stands.

Note the document includes "This KFI gives an overview of typical traffic management practices undertaken on this product; it does not cover circumstances where exceptional external events may impact on network congestion level."

ISP's never lie .... ok then!

That does not prove anything as a VPN is routing traffic differently.

Either you didn't read what I wrote or you don't understand what you are talking about...

Your speedtest shows absolute crisp clear congestion, others here will agree when they come online and read this. Your single thread is 10Mbps, your multi is 70, that is congestion.

Seeing as the speed test servers are probably prioritised any way (It's certainly what i'd do) then its a poor test - if it *was* congestion 6x wouldn't be a neat and smooth as it is. - go look at proper congested ISP results they are still spiky even on 6x test..

Find me ANY mass market ISP (talktalk, sky, plusnet, BT, EE) where they have traffic management and document this as required. This needs to be in the form required per the voluntary code of conduct. The statement "All ISP's employ traffic management" is false. Almost none do any more... I have been in this game for 11+ years, and I have worked with almost every big name provider from an infrastructure standpoint.

All do... fact. None hopefully use it - but all will have it setup to deploy if necessary.

They do not deploy a traffic management solution.

You sound very much like a vodafone employee saying that - here's an example: This is a file download, then a file download over VPN - the start point of the VPN is my machine, the end point of the VPN is the SAME server as the download is from, this is NOT a VPN to some hosted server which then goes to the server with the file on, this is the SAME server, I'm either downloading straight, or encrypting it and downloading - can you see the difference? Pretty easy to spot - So why two different speeds? - That is shaping prioritising traffic differently as one it can see is NNTP and the other it cannot.

8 thread download on NNTP traffic
https://www.dropbox.com/s/2zwicxssc79em0x/donwload%2...

Same 8 thread file download with VPN terminating ON the server providing the file.
https://www.dropbox.com/s/oggyzcx9nwrh1uq/donwload%2...

Edited by Tam (Sun 23-Dec-18 23:34:19)

I won't keep on this same spiral.
ISPs are regulated entities they do not deliberately lie and mislead consumers.

This is not congested with BT:
https://www.thinkbroadband.com/speedtest/14833184342...

It is very different to yours...

EDIT: I see your points and if you had a Speedtest like the one I have linked I would likely agree, and then I'd go away and investigate.

Can you do a traceroute with and without the VPN to the server in question?
Something is different... there is no guarantee VF is handling your traffic the same just because it ends at the same place.

I'd conclude your non VPN traffic is hitting congestion for whatever reason. Most likely VPN a traffic is routed by Vodafone in such a way that the congestion is not manifesting.

Edited by ukhardy07 (Sun 23-Dec-18 23:48:15)

If Vodafone is prioritising speed test servers they are doing an amazingly bad job of it.

https://www.thinkbroadband.com/news/8239-peak-and-of...

Slow single download speeds is generally what is seen before things get worse and multiple downloads start to shift into the spiky area and everyone starts to complain

https://www.thinkbroadband.com/news/5134-new-adverti...

In reply to a post by ukhardy07:

Its *NOT* congestion, if it was congestion then *everything* would be slow

This is not true, you do not understand how the internet works if you believe all traffic is routed identically.

When you use the VPN, then yes all your traffic is routed via the VPN, and that route is likely not congested.

Your speedtests show very very clear congestion and your ISP is known for congestion widely.

Read the traffic management document published:

http://www.vodafone.co.uk/cs/groups/public/documents...

NO traffic is managed. They cannot lie about this... It is documented and formally captured.

Although that document doesn't seem to say anything about what they do with IMAP traffic on port 143. Admitedly it's not throttling but I have proof that they are interfering with the data stream and removing STARTTLS commands. My assumption is that it's an attempt to perform virus scanning.

https://forum.vodafone.co.uk/t5/Mobile-Broadband/Pro...

Yep VF have some issues in the way their filtering works. Poorly implemented and it meant for years sites such as imgur didn't work correctly.

There's still plenty of talk on forums of the "filtering" and oddities such as what you are reporting. They could be different issues and I suspect they are, but overall it doesn't surprise me.

I don't think I've seen concrete proof of what is actually occurring. Explanations have ranged from man in the middle and certificate spoofing, to the IWF filters.

In reply to a post by Andrue:

My assumption is that it's an attempt to perform virus scanning.

How would they handle implicit TLS IMAP traffic over port 993?

I'd suspect they don't mess with that port.

It appears they are "trying" to be helpful on 143 ......

We seem to have gone off track though ....

Edited by Tam (Mon 24-Dec-18 13:36:00)

It just seems unlikely to me they'd go to the trouble of stripping STARTTLS for virus scanning for the very small percentage of IMAP users who use STARTTLS over 143 in preference to the usual 993.

Not really - all mobile traffic is run through a small number of proxy servers, so quite easy to do......

Mobile networks and broadband networks are very very different it design, implentation and managment.

Tam

The "filtering" Vodafone is doing might be causing speed issues if the downloads are from a site with user uploaded files, like Mediafire or a number of other similar hosting services. Compare the router's DNS responses with those from a trusted DNS server to find out.

On the concrete proof of what is occurring with filtering, I added some more info here (sorry for the long post!)
http://forums.thinkbroadband.com/vodafone/t/4610585-...