Upstream Traffic Management Trial 1st of February

[qasdfdsaq]

Originally Posted by Milambar

Skype, one of the most used applications on the internet, I do believe, relys on a P2P protocol to make it work. The average user most certainly WILL notice it.

Skype is NOT being targetted by VM's traffic shaping systems. It also uses minimal amounts of bandwidth, and quickly adapts to various traffic levels by itself.

We moderate the total volume of file sharing traffic on our network between 5pm and midnight on weekdays and midday and midnight on weekends. This policy, which applies to all broadband packages, is restricted to Peer to Peer ("P2P") applications and Newsgroups (which are commonly used to distribute large amounts of data)
This policy does not impact any applications other than Peer to Peer and Newsgroups, so things like watching iPlayer, online gaming, making calls via Skype, downloading music tracks from iTunes or streaming them from Spotify and sending an email or normal browsing are unaffected

I've already said VM are referring to P2P specifically with relation to file sharing, nothing else.

---------- Post added at 15:27 ---------- Previous post was at 15:21 ----------

Originally Posted by Chrysalis

every implementation of shaping I have witnessed is always noticeable as it often shapes things its not supposed to. However the average user would probably not even know about shaping so wouldnt blame it.

The average user still can't tell if they're getting 8 meg or 20 meg from their "Up to 20mb broadband".

[whizzard]

Originally Posted by Chrysalis

I agree with Masque on this tho, if customer rings up and support have no idea, then customer as a result will get wrong support response.

---------- Post added at 13:40 ---------- Previous post was at 13:20 ----------

My question is what are VM going to do about the inevetible overloading of off peak after this comes into force. This will just move the high use to a different time of day.

I would imagine they would have call routing in place to cover this, though obviously it is dependent on the end user dialling from the correct number.

[Peter_]

Originally Posted by BenMcr

They will know before the trial starts

Well nothing today on the systems about it apart from a post on the Community Forums.

[jtaylor06]

They've let their April Fools joke out too early!

[Ignitionnet]

Originally Posted by Masque

Well nothing today on the systems about it apart from a post on the Community Forums.

I doubt the sky will fall in from a technical trial for a couple of hours in the middle of the night.

1st line technical support haven't been updated because if you were every fault even remotely likely to be in any way, shape, or form related would be in with a pretty good chance of immediately being blamed on the trial instead of being properly worked.

That would cause far more inconvenience to customers than the alternative - to not mention it until it is getting towards going properly live.

[Peter_]

Originally Posted by Ignitionnet

I doubt the sky will fall in from a technical trial for a couple of hours in the middle of the night.

1st line technical support haven't been updated because if you were every fault even remotely likely to be in any way, shape, or form related would be in with a pretty good chance of immediately being blamed on the trial instead of being properly worked.

That would cause far more inconvenience to customers than the alternative - to not mention it until it is getting towards going properly live.

Its my day off tomorrow so not really overly interested to be honest.

[General Maximus]

you just knew it was going to happen, they increase upload speeds only to find an excuse to bang them back down again. It is all marketing, they have might as well advertise symetrical 200mbit knowing that they arent going to give it to everyone.

[qasdfdsaq]

"Up to x mb broadband"

What was it they said a couple years ago... Something like 40% of bandwidth goes to the top 0.5% of users, 80% of total bandwidth is used by 10% of users and the bottom 80% of users use only 10% of the bandwidth. Note: Independant figures, but not UK figures.

And apparantly P2P was down to 20% or less in '08 with youtube using over 50%... No wonder traffic shaping hasn't made a huge difference.

In that context shaping P2P traffic down to 25% of port capacity isn't actually that draconian if it was only using 20% to begin with

Mind you, I use a lot lower percentage of my line's capacity than I did a while ago mainly because my hard drive space hasn't increased in line with internet connection speeds.

[Chrysalis]

Originally Posted by Ignitionnet

I doubt the sky will fall in from a technical trial for a couple of hours in the middle of the night.

1st line technical support haven't been updated because if you were every fault even remotely likely to be in any way, shape, or form related would be in with a pretty good chance of immediately being blamed on the trial instead of being properly worked.

That would cause far more inconvenience to customers than the alternative - to not mention it until it is getting towards going properly live.

does the properly worked mean blaming the user's pc instead?

[Peter_]

Originally Posted by Chrysalis

does the properly worked mean blaming the user's pc instead?

Now why would we do that if we are using our online tools correctly we can see any issues or management on our side and only then will we look at your equipment.

It is amazing how many 3rd party routers actually cause slow speeds and once a direct connection is made the connection is running at full speed.

You have a known issue which will be in your notes so you cannot use yourself as an example.

Originally Posted by qasdfdsaq

"Up to x mb broadband"

What was it they said a couple years ago... Something like 40% of bandwidth goes to the top 0.5% of users, 80% of total bandwidth is used by 10% of users and the bottom 80% of users use only 10% of the bandwidth. Note: Independant figures, but not UK figures.

And apparantly P2P was down to 20% or less in '08 with youtube using over 50%... No wonder traffic shaping hasn't made a huge difference.

In that context shaping P2P traffic down to 25% of port capacity isn't actually that draconian if it was only using 20% to begin with

Mind you, I use a lot lower percentage of my line's capacity than I did a while ago mainly because my hard drive space hasn't increased in line with internet connection speeds.

Not according to Cisco

By 2014, global online video will approach 57 percent of consumer Internet traffic (up from 40 percent in 2010).

Interesting stat (imho) from Sandvine

BitTorrent remains the most used file-sharing protocol in North America, and the total amount of P2P traffic is still very significant. Sandvine’s research reveals that on an average day, 53.3% of all upstream traffic can be attributed to P2P applications. P2P is less dominant on the downstream side. It is currently at 13.2%, following real time entertainment (45.7%) and web browsing (24.3%).

The bandwidth usage patterns during peak hours are slightly different, but still a massive 34.31% of all upstream traffic can be attributed to BitTorrent at these times. The BitTorrent percentage of downstream traffic lies at 8.39% during the busiest time of the day.....

.......In common with North America, BitTorrent also remains the most used file-sharing protocol in Europe. The report doesn’t give any exact stats, but roughly 40% of all upstream traffic and 10% of all downstream traffic can be attributed to P2P applications on an average day.

Bandwidth usage patterns during peak hours show that of 29.97% of the upstream traffic can be attributed to BitTorrent during these times, versus 8.29% of downstream traffic. PPLive, the popular peer-to-peer streaming video network, also has a significant share with 11.76% of all upstream traffic and 4.41% of downstream traffic during peak hours

[qasdfdsaq]

Alright, I was generalizing a bit, and the data I quoted was from one particular manufacturer of traffic shaping/DPI equipment. In any case, the general consensus seems that "real time entertainment", i.e. streaming audio/video make up 40-60% according to all parties, fairly close to half of all internet traffic with some variation by region.

And P2P traffic, in the downstream, varies from the high single digits to low double digit range, which is fairly close to what I quoted anyway. Again patterns will vary by ISP and region, but shaping what takes up 10% of your bandwidth down to a maximum of 25% of bandwidth (well, with NNTP included) was never going to have much effect - that is of course if VM are actually doing it as they say they are.

Even in the upstream, if P2P is using ~35% overall and ~30% at peak times, capping it to 25% really won't reduce overall traffic by more than 10% if you look at it logically. We can pretty much ignore NNTP here since, well, there is negligible amounts of upstream NNTP on a home user's connection.

[Chrysalis]

Originally Posted by Masque

Now why would we do that if we are using our online tools correctly we can see any issues or management on our side and only then will we look at your equipment.

It is amazing how many 3rd party routers actually cause slow speeds and once a direct connection is made the connection is running at full speed.

You have a known issue which will be in your notes so you cannot use yourself as an example.

To be fair only the one time I used normal tech support they blamed my pc.

The CEO office and tier 2 on the VM forums didnt do that.

---------- Post added at 01:00 ---------- Previous post was at 00:57 ----------

Originally Posted by Hugh

Not according to Cisco

Interesting stat (imho) from Sandvine

Hugh I have to say those stats are a brilliant match with what I would expect. In that p2p doesnt dominate downloading (as some claim) but does use a large chunk of uploading. This falls in line with what I have been thinking.

---------- Post added at 01:05 ---------- Previous post was at 01:00 ----------

Originally Posted by qasdfdsaq

Alright, I was generalizing a bit, and the data I quoted was from one particular manufacturer of traffic shaping/DPI equipment. In any case, the general consensus seems that "real time entertainment", i.e. streaming audio/video make up 40-60% according to all parties, fairly close to half of all internet traffic with some variation by region.

And P2P traffic, in the downstream, varies from the high single digits to low double digit range, which is fairly close to what I quoted anyway. Again patterns will vary by ISP and region, but shaping what takes up 10% of your bandwidth down to a maximum of 25% of bandwidth (well, with NNTP included) was never going to have much effect - that is of course if VM are actually doing it as they say they are.

Even in the upstream, if P2P is using ~35% overall and ~30% at peak times, capping it to 25% really won't reduce overall traffic by more than 10% if you look at it logically. We can pretty much ignore NNTP here since, well, there is negligible amounts of upstream NNTP on a home user's connection.

In VM's case tho there is a issue not accounted for and that is regional congregation of users. So eg. VM might have a 100 UBR's with not a single p2p user on, whilst another 100 UBR's may have 20 p2p users on each going on 24/7. The first 100 UBR's would have pointless throttling and the 2nd 100UBR's would have not enough throttling. It does seem currently VM's downstream traffic shaping overall is a pointless waste, in a highly congested area I have seen pretty much no improvement from it. If I am seeing nothing from it I fail to see what good areas gain from it. I dont mind posting my tbb graph's for the days this upstream trial is running so we can see the affects from that. As even tho the trial is dusk hours, I do still have congestion then on the upstream.

[pip08456]

Originally Posted by Chrysalis

In VM's case tho there is a issue not accounted for and that is regional congregation of users. So eg. VM might have a 100 UBR's with not a single p2p user on, whilst another 100 UBR's may have 20 p2p users on each going on 24/7. The first 100 UBR's would have pointless throttling and the 2nd 100UBR's would have not enough throttling. It does seem currently VM's downstream traffic shaping overall is a pointless waste, in a highly congested area I have seen pretty much no improvement from it. If I am seeing nothing from it I fail to see what good areas gain from it. I dont mind posting my tbb graph's for the days this upstream trial is running so we can see the affects from that. As even tho the trial is dusk hours, I do still have congestion then on the upstream.

But those 100 UBR's with no-one using P2P would not be affected by the throttling so it would be as if it didn't exist. It's not pointless throttling in that case, it is no throttling.

[Chrysalis]

it would still be a throttle to 25% of capacity on the false positives

my example was extreme tho, obviously we can go to less extremes and have eg. 1 or 2 p2p downloaders. vs 18 or 19.