50meg now working - but speed poor. Quick question

[Retrovertigo]

Had 50meg installed on Friday and it was just activated a few minutes ago, but am only getting maybe 35meg (only he says!) - but with my old 20meg modem I had lots of speed issues and at one point 7dBv of attentuators were fitted.

When the guy came on Friday he didn't do any checks with the meters he has, he just plugged in the 50meg modem, but left the attentuators on. I'm not entirely sure they will now be needed? And if it is what is affecting the speed.

Is it ok for me to just unscrew them and see if it improves? At the minute the levels are:

Power Level
(dBmV) 3.20 2.64 2.83 2.80

RxMER
(dB) 37.09 37.36 37.64 38.26

Thanks for any input

[Ignitionnet]

Those levels are good, no point in removing the attenuation.

[Retrovertigo]

Ok, thanks for the speedy reply :)

[kamakazi]

Same for me. He left them on mine as well

Are mine OK?

Power Level
(dBmV) -0.50 -0.42 -0.43 -0.16

RxMER
(dB) 35.60 36.17 36.39 36.84

[xocemp]

Those are fine kamakazi.

[kamakazi]

Many thanks for prompt response...

[Kevin]

Try swapping your ethernet cable, I had the same issue but by replacing the cable connecting pc to router with a new one all went up to 50mbit, dunno if it was just a dodgy cable or it was an old cat 5 one... I know it sounds weird and I thought it would never work but it did!

[Retrovertigo]

Well, sat here this evening and no matter what site I try, or grab stuff from newsgroups, I'm struggling to keep above 15meg. 10meg or less is the average off speed tests.

Webpages are being a touch slow to load as well. I can't say I am blown away that I am getting less speed than a 20meg line should be.

I feel bad complaining because VM have done me a very good deal indeed. But as someone pointed out here, if there is over saturation in my area, then I'll still get lag playing online. And on the evidence so far, they will be bang on the money.

I know 50meg may be popular where I am, but is 10-15 meg really acceptable of an evening? I am inclined to think not.

[Sephiroth]

Well, sat here this evening and no matter what site I try, or grab stuff from newsgroups, I'm struggling to keep above 15meg. 10meg or less is the average off speed tests.

Webpages are being a touch slow to load as well. I can't say I am blown away that I am getting less speed than a 20meg line should be.

I feel bad complaining because VM have done me a very good deal indeed. But as someone pointed out here, if there is over saturation in my area, then I'll still get lag playing online. And on the evidence so far, they will be bang on the money.

I know 50meg may be popular where I am, but is 10-15 meg really acceptable of an evening? I am inclined to think not.

Oh yes - 10-15 Mbps of an evening on a 20 Mbps circuit at peak time is pure magic.

In many cases, as Kevin suggests, the difference between Cat5 & Cat5e and similar improvements can bump speed up even further when there is little line contention.

[Retrovertigo]

I'm slightly lost at your bit in red? I am now on 50meg and am getting less than 15meg. I was asking if that was acceptable in the evening - and I don't think it is.

I have several cat5 cables here, all giving the same result. I think the speed is appalling to be honest.

[Sephiroth]

I'm slightly lost at your bit in red? I am now on 50meg and am getting less than 15meg. I was asking if that was acceptable in the evening - and I don't think it is.

I have several cat5 cables here, all giving the same result. I think the speed is appalling to be honest.

OK. My bad - I misread your post and forgot that you were on 50 Mbps.

Neverhteless, 10-15 Mbps is fully useable for most things and contention by others (hoping possibly to do Newsgroups stuff) is taking your speed down.

Many others complaining here on 50 Mbps sometimes get less than 1 Mbps during peak time and that is completely unacceptable. I don't think that 10-15 Mbps at peak time will get a lot of sympathy.

It doesn't stop us trying to help you where we can in case there is anything other than contention involved.

[Retrovertigo]

Many others complaining here on 50 Mbps sometimes get less than 1 Mbps during peak time and that is completely unacceptable. I don't think that 10-15 Mbps at peak time will get a lot of sympathy.



Personally I think there is something else going on. I'm told there is an ongoing error in my area causing dropouts (6 today and counting - and those are just the ones I'm aware of). I'm hoping this is also affecting speeds. It has dropped again to around 5meg and I personally don't count after midnight as a peak time to be fair.

I know you say 10-15meg is usable, and for regular boring stuff it is. But I fear that if contention is such a problem as to cause such dramatic speed drops (and in the big scheme of things, 50 down to 10-15 is a big drop) then it will no doubt affect online gaming, which I really hoped I would improve with the switch to the highest tier.

[Sephiroth]

That's why I like looking at pathping results (to bbc.co.uk). It indicates where a bottleneck might occur and the device name tells us what the capacity of that port is. Then we can speculat5e on contention. For example, if you come into a 1Gb port at a hub and more than 20 other active users are doing the same and they are not browsing but streaming and the like, that port is oversubscribed. UDP packets will be dropped and download spoeeds will degarde hyperbolically with the degree of oversubscription.

As for midnight not being peak tim - the students will stay up until at least 01:00!

A good max speed time is before 09:00 IMO.

Cheers.

[Ignitionnet]

Traffic on broadband networks which are not shaped usually peaks between 8 and 9PM then tails off - see the LINX traffic stats (https://www.linx.net/pubtools/trafficstats.html?stats=day) for more info.

On VM the pattern varies however peak load in most areas remains pre-9PM but is harder to measure because it's so bursty.

As I mentioned previously none of the VM IP network is oversubscribed, indeed none of the links run at 50% even to ensure n+1 bandwidth availability to business customers so issues will only be on peering and transit or at local RF/CMTS port level, any issues seen on the VM core are a fault.

[Sephiroth]

Traffic on broadband networks which are not shaped usually peaks between 8 and 9PM then tails off - see the LINX traffic stats (https://www.linx.net/pubtools/trafficstats.html?stats=day) for more info.

On VM the pattern varies however peak load in most areas remains pre-9PM but is harder to measure because it's so bursty.

As I mentioned previously none of the VM IP network is oversubscribed, indeed none of the links run at 50% even to ensure n+1 bandwidth availability to business customers so issues will only be on peering and transit or at local RF/CMTS port level, any issues seen on the VM core are a fault.

BB - that is worth expanding on if you can be bothered.

There is the coax run to the optical node; how many optical nodes would be fitted into the single housing? And are these segregated by service tier or tier groups?

Is the 500 houses passed per node calculation still in effect?

How is load coming from the optical nodes managed at the local hub (e.g. Winnersh) and the local Host (e.g.) Reading?

I believe that the contention issues centre on more people doing stuff than has been previously calculated by VM. I'd like to know what the bandwidth is between optical node and the local hub. I assume there is a single fibre to the hub, modulated according to the tiers of service. What's this Gigabit? 2Gb? 38 Mb? What?

The maths would be realtively simple to do if we knew that stuff.

Cheers

[Ignitionnet]

BB - that is worth expanding on if you can be bothered.

*Cracks knuckles*

There is the coax run to the optical node; how many optical nodes would be fitted into the single housing? And are these segregated by service tier or tier groups?

No simple answer is it depends, however the 'default' is that the 20 and 50Mbps customers are on the newer equipment while the 10Mbps customers remain on the legacy equipment.

A standard single node can run on up to 4 fibre pairs, each creating a 'virtual' node within the physical one.

Is the 500 houses passed per node calculation still in effect?

It was never in effect, node sizes vary wildly. Some areas were built with 2, 3 or 4,000 homes passed per node.

How is load coming from the optical nodes managed at the local hub (e.g. Winnersh) and the local Host (e.g.) Reading?

Reading is where the CMTS live, a node or group of nodes maps to x downstream ports and y upstream ports. There is no static hard and fast description as it can be all over the place. However the nodes terminate at Reading, from there it's all IP up to Winnersh and onwards onto the Internet. Connectivity from Reading to Winnersh will be on n x Gigabit Ethernet, either CMTS directly to router or aggregated onto switches at the hubsite then on n x GigE or 10GigE to routers. Hubsites may also have their own routers as well.

I believe that the contention issues centre on more people doing stuff than has been previously calculated by VM. I'd like to know what the bandwidth is between optical node and the local hub. I assume there is a single fibre to the hub, modulated according to the tiers of service. What's this Gigabit? 2Gb? 38 Mb? What?

You should only really be concerned by what the node is connected to, and how many other nodes it's sharing that with. I'll give a 'nominal' set up for you however:

Dedicated to each node:

2 x 256QAM DOCSIS / 64QAM EuroDOCSIS downstreams for legacy network

and

4 x 3.2MHz 16QAM upstreams for legacy network

Total 2 x 38Mbps downstream, 4 x 9Mbps upstream.

and for the DOCSIS 3 / overlay network:

4 x 256QAM EuroDOCSIS downstreams on overlay network shared between this node and 3 or more others

and

1 x 3.2MHz 16QAM upstream shared between this node and perhaps 1 other

so 200Mbps downstream shared between 4 or more nodes, and 9Mbps upstream shared between 1 or 2 optical nodes.

The maths would be realtively simple to do if we knew that stuff.

Cheers

See above - no it wouldn't ;)

You should try and think of the cable network in terms of MAC domains, these are abstract areas sharing bandwidth, rather than in terms of physical optical nodes. This said each MAC domain is comprised of x optical nodes so all has to start from there.

500 homes passed by a single node is probably quite optimistic in a lot of areas, I know that in my area the node size is closer to 1000 homes passed, and before resegmentation was 2000+.

Bad speeds usually arise where optical work needs to be done. Where nodes are sharing bandwidth on the legacy network this usually means that they are having the same optical signal split between them downstream, and are having their upstream signals combined, so you may need to install additional downstream lasers in the hubsite to give the nodes their own dedicated optical downstream and install new optical receivers in the hubsite to give them their own dedicated upstream/return paths.

This will make little sense but is best described by Visio diagrams, which I don't have to hand, but there's a fair bit of coaxial patch panel work that goes into feeding the nodes their signals.

[Sephiroth]

Thanks for that, BB.

I am in the (slow) process of constructing scenario based Visio diagrams to incrementally (as info comes on this site via tracert) represent a public view of the VM network. There are also the publicly available network maps from NTL bsuiness although I don't know if that backbone is still valid or if VM sold any of it off and if so whether it is leased back and thus used.

[Ignitionnet]

http://help2.virginmedia.com/help/getContent.jspx?page=h_internet_advanced_dns will help you with that. Virtually every hub has 2 OSRs then connects to its' RHE's 2 OSRs via them. There are exceptions where ex-ntl architecture is present, t3cores (Juniper T320 or T640 I think?) replacing OSRs at RHE.