2gb facts ???

Stand by for the new Gig2 customers complaining that they are connecting at 70 Mbps because they are on their old phones on the 2.4 GHz band. 

Yeah it's the Lambo in the suburban driveway mentality. "Just look at my wad everyone!" as they potter about between red traffic lights every block or so in 20mph zones on the school run. 

---

Client62 wrote:

For folks asking the question of what kit to use on a 2Gb/s connection, perhaps another consideration applies.

A 2 Gb/s service means having 21,600 Gigabytes of file download capacity per day.  

We do not buy a vast excess of any other product or service and allow most of it go to waste.

Selling bandwidth to domestic customers has become taking candy from a baby.

---

For now it's not metered, and that's thanks to a vast excess of it being allowed to go to waste. This would certainly change if customers wanted to ensure they weren't wasting any capacity. It'd move from a burst speed charge to a usage based charge.

Personally I have a very fast service for that burst speed. I don't consider not maxing it out 24x7 as 'waste': I'm not paying to max it 24x7.

Let people buy whatever they want to: the most widely used game distribution system, Steam, will saturate a 2 Gbps connection. Some folks are impatient and pay for the convenience.

VM don't provide DIA (dedicated internet access). If you try to max out a non dedicated line 24/7 you'll be hit with a FUP very quickly as their peering arrangements at IXPs have nowhere near enough bandwidth to sustain all customers maxing out their connections.

VM only has 200Gb of public peering at LINX London and 100Gb of public peering at LINX Manchester.

It's a real shame how everything is about bandwidth and not speed. I wish domestic ISPs would focus more on speeds and peering and not focus on bandwidth to netflix and select peers.

I wish I could have a sub 3ms speed to the nearest DC with excellent peering arrangements with the real internet.

Unfortunately most domestic users just want useless multi-gigabits to netflix, no speed and no good peering arrangements with the actual internet.

At work we have a 1ms speed DIA connection to most London DCs, with only 200Mb bandwidth. Its about 20 times faster than 1Gb Virgin Connection. Browsing the net is an absolute pleasure on a 1ms line.

---

Roger_Gooner wrote:

---

---

This Gig2 is currently only available on nexfibre's network, but I'm sure it will spread to VM's network depending on the rollout of XGS-PON.

---

Nothing to do with XGSPON at all, Gig2 will be available on HFC.

I'm not sure you understand how private peering works. In fact I know you don't. It's cables back to back between the two networks. There's no 'cheap' arrangement there and Cogent have had a few 'issues' recently.

Regarding Aorta being tier 1 https://bgp.he.net/AS6830#_peers is a who's who: Aorta is tier 1: no-one connects to the networks they do without some serious weight. I'm not a fan of 5089 being put largely behind 6830 but that's how it is. 

Not sure why VM don't connect to your ISP directly: lots do - https://bgp.he.net/AS5089#_whois 

Who cares about good international routing anyway? Once you get past 3 ms to the real Internet it all breaks down. Not sure how I'm managing with 6 to Manchester and 10 to London but I'm doing my best

Actually, I do know. You have 4 ms of jitter on your home cable. I have zero, I'm on uncapped XGSPON. It's the jitter that's killing you, not a few ms of delay. Pages are blistering for me thanks to good DNS and low jitter. 10 or 20 ms of latency means nothing for browsing as long as it's smooth: our requests aren't all in the same DC our connections meet the wider Internet.

EDIT: Actually while I'm at it let's discuss this, quoting your good self, emphasis mine: 'VM is fine if you're doing netflix and playing games all day. But if you're working from home all day, using realtime protocols such as VoIP, RDP etc. requiring good international (and national) routing, it's a crap connection.'

So: gaming is far more demanding than either VoIP or RDP. VoIP and video calls have buffers, often 50+ ms, gaming doesn't. Gamers are far less tolerant off issues for this reason. RDP I'll take as remote terminal type applications: these are fine over VM. The additional delay means not a lot: Teams screen sharing, Zoom sharing, etc, 10 or 20 ms of delay doesn't mean much, we aren't talking video games with 120 ticks a second, we're talking scenarios where 150 ms latency is tolerated.

If you're using RDP over WAN and not using a work VPN to get there, RDP is open to the Internet, delete your account. If you're using VPN how much latency is the VPN concentrator adding? 

---

asim18 wrote:

Thanks for that information.

VM's residential fibre service is definitely the same as work's dedicated internet access. And 20ms ping is certainly the same as 1ms. And VM has the best peering and routing in the world.

---

You're welcome for the information! Shame that rather than taking it on board you instead became defensive and wrote a bunch of nonsense I hadn't actually said, potentially because of a bruised ego, but that's life.

You would do well to do a bit of research into how the buffering on voice and video calls works and why they tolerate a range of latencies and jitters.

You aren't using RDP publicly addressed over the Internet, are you? If you are the suggestion to delete your account persists.

---

asim18 wrote:

Yes I'm sorry.

When I found cogent was running 5ms between london and manchester, and VM was running 22 ms between london and manchester, I totally assumed that 5ms was better than 22ms.

I now know that 5ms is the same as 22ms.

---

Or you could've looked at the traceroute you took from the Looking Glass, your words that it came from there, not from VM. The one showing a first hop delay of 17ms. 22ms - 17ms =...? The 17ms will be delay between whatever they're using to generate the traceroute and their routing kit, the rest the actual route. Which came in at ~5ms.

Last post on the previous page as I'm sure you're aware.

---

asim18 wrote:

I used the traceroute to find the route not the ping, I used the ping to find the ping.

 

 

But yes I'm not defending myself at all, I'm totally agreeing with you on all points and future points. 22ms is the same as 5ms. and no one needs to communicate internationally over the internet at all. All great points thanks.

And yes you are right a highly-buffered high latency voice call is the same as a low latency less-buffered voice call and there is no voice delay between a 1ms latency call and a 150ms latency call. It's all within tolerance as you rightly say. Thanks for that info too.

Once again, big apologies for assuming that low latency internationally peered dedicated internet access was better than VM residential connections. I will inform my company director that he is wasting £800 per month and tell him to switch to VM.

---

DIA uses the same transit and peering as the broadband and indeed everything else on pretty much every ISP.

As mentioned earlier the Liberty looking glass isn't working properly and 17 ms is getting added. Consider doing a traceroute and finding out for yourself.

I don't recall saying that DIA had no improvements over VM cable: of course it does. VM cable is notably slower when browsing than alternatives due to the use of DoCSIS: XGSPON fixes this, even over VM's core network with their transit and peering you dislike. Much of the degradation is the DoCSIS jitter and doesn't happen even for those of us 10 ms or more away from London: our browsing is faster than yours at home.

I didn't say anything about 150 ms latency being the same as 1 ms: I said it was tolerated in response to your claiming that VM is only good for gaming but can't handle interactive/real time business apps. My exact comment was:

'So: gaming is far more demanding than either VoIP or RDP. VoIP and video calls have buffers, often 50+ ms, gaming doesn't. Gamers are far less tolerant of issues for this reason. RDP I'll take as remote terminal type applications: these are fine over VM. The additional delay means not a lot: Teams screen sharing, Zoom sharing, etc, 10 or 20 ms of delay doesn't mean much, we aren't talking video games with 120 ticks a second, we're talking scenarios where 150 ms latency is tolerated.'

Gamers don't generally seek out matches with people in California, Singapore or Australia, businesses have voice, video and desktop sharing sessions with folks out there.

You should speak to your company director though: £800 a month for 200 Mbit on a gigabit bearer is insanely expensive for pretty much anywhere and is absurd for London.

Just to reiterate my points...

If you try to max out a non dedicated line 24/7 you'll be hit with a FUP very quickly as their peering arrangements at IXPs have nowhere near enough bandwidth to sustain all customers maxing out their connections.

Congestion would happen on the access network first - 2 Gig1 customers uploading at full pelt on the same segment would saturate it. 3 customers downloading at full pelt would do it. Wouldn't register on the transit and peering.

VM only has 200Gb of public peering at LINX London and 100Gb of public peering at LINX Manchester. -

VM operate like a US cable company in many ways and make extensive use of private peering both directly and via Aorta, which is why their network shifts 20+ Tb/s despite the above. Yes, it's annoying at times but not necessarily awful and makes sense from their point of view.

It's a real shame how everything is about bandwidth and not speed. I wish domestic ISPs would focus more on speeds and peering and not focus on bandwidth to netflix and select peers. -

Most of them don't have their own national networks so aren't in control of most of the end to end latency. Their entire network is some routers, switches and BRAS hosted in London.

I wish I could have a sub 3ms speed to the nearest DC with excellent peering arrangements with the real internet.

If you're in London perfectly doable with FTTP. DoCSIS is the killer there.

Unfortunately most domestic users just want useless multi-gigabits to netflix, no speed and no good peering arrangements with the actual internet. -

They just want their service to do what they use it for well. Up to the ISPs to concern themselves with transit and peering, customers shouldn't have to care. First folks who do notice when it's not up to snuff are usually gamers.

At work we have a 1ms speed DIA connection to most London DCs, with only 200Mb bandwidth. Its about 20 times faster than 1Gb Virgin Connection. Browsing the net is an absolute pleasure on a 1ms line.

You'll be accessing pages involving content coming from all over the place. As long as the flow is smooth and DNS responsive browser rendering becomes the bottleneck. VM cable is not great for browsing due to DoCSIS jitter and TCP proxying. Use VM cable in the North and then full fibre or even FTTC and browsing is noticeably better on the FTTP/C. It's a DoCSIS thing not an every millisecond counts thing.

Using liberty global's aorta.net looking glass from telehouse london to the same webserver it's 22ms and uses cogent's backbone. Not actually tier 1 network in my opinion, if it cant even get from London to Manchester using its own network.

It's actually about 5, there's a 17ms delay not related to the network: see my other traceroutes. See https://en.wikipedia.org/wiki/Hot-potato_routing for why this behaviour is normal regarding getting the traffic to Cogent rather than carrying to Manchester on their own. Nothing to do with being a tier 1 or not.

VM is fine if you're doing netflix and playing games all day. But if you're working from home all day, using realtime protocols such as VoIP, RDP etc. requiring good international (and national) routing, it's a crap connection.

Covered above. Gamers are the most demanding customers, full stop. VM's routing could certainly be better and the move behind Aorta isn't ideal but in the grand scheme it doesn't really matter. Regarding international routing that is out of the hands of ISPs once they've passed the data to their peer. They can control the AS they send to outbound if they've multiple with the route they need, they can influence the return path. Aorta routing should be great here as they've connectivity to everyone while most only have a couple of big upstreams.

I'm not saying VM's transit and peering are incredible but the DoCSIS stuff is far more of a problem for right now. Folks on XGSPON that's working properly seem pretty happy. For their issues fixating on VM's transit and peering is probably not the way to go: browsing is slower because of the access network.

You're welcome, Asim.