well this thread was a hit .

In reply to a post by Ignitionnet:

Indeed. Doesn't reflect too well given HFC is far superior in just about every way apart from where the contention point is. Difference between point to point solutions and broadcast media.

... and that's what worries me about the end user optical solutions using PON.

It's all going to be a big hype about having fibre into the home except that you share the optics at the local level with all your neighbours.

Star Ethernet vs Token Ring? Switch vs Hub?

What are your thoughts on advertising / T+Cs like these?

http://www.att.net/speedtiers
https://www.comhem.se/bredband

In reply to a post by Ignitionnet:

In reply to a post by WWWombat:

Ironically, I remember reading some American cable dimensioning data that said *precisely* that: 3 users on the top tier.

Be interesting to see that scaling information. I sincerely hope there weren't 2000 other modems in the service group though! I don't think that would work, let alone work well.

I've found my copies of the 2 documents I read this in; let's hope that they say what I think they say...

Both came from Arris, and seem to describe the same network sizing information. They seem to date to around 2014.

1. "NIELSEN'S LAW VS. NIELSEN TV VIEWERSHIP FOR NETWORK CAPACITY PLANNING"
HTML Link

Quote:

Rule of Thumb Approach for Network Sizing

To date some MSOs have sized their network on a method of multiplying the billboard speed by either doubling (2X) or tripling (3X) the billboard speed to determine the amount of DOCSIS capacity per service group, this is sort of a Rule of Thumb method for DOCSIS Network Sizing.

The document goes on to explain that, as service groups get smaller, that "rule of thumb" might end up allocating too much bandwidth.

This document doesn't give a basis for *why* 2x or 3x is chosen, but does explain a better method for the future.

The next document describes the same future, but has more detail for the past...

2. "IS NIELSEN READY TO RETIRE? LATEST DEVELOPMENTS IN BANDWIDTH CAPACITY PLANNING"
PDF Link

Quote:

Many MSOs transitioned to the (Consumption+Tmax)-­oriented Period roughly around the 2009 time-­frame in conjunction with the MSO transition to DOCSIS 3.0 channel-­bonding. This is where we are today. In this period, The SG Aggregate Average Bandwidth roughly equals the rate of the highest service tier identified by the DOCSIS value: Tmax.

This has allowed traffic engineers to use a very simple rule to estimate the size their High Speed Data (HSD) networks:

Required Service Group Bandwidth Capacity = 2xTmax (1)

The Rule of Thumb can often be used as a quick guideline for estimated capacity needs. However, MSOs have developed tools and optimizations based on volumes of empirical data and traffic analysis that allow them to engineer networks for high quality of experience as key variables change.

This Rule of Thumb has been morphing over time. It was originally 3X and has been shrinking with time (and smaller Service Groups, SG).

I read the first part of this quote to say that Tmax is the speed of the top tier they sell (the "billboard speed" of the previous document). And that, at the current time, it just so happens that average aggregate bandwidth of an SG is about the same value as Tmax. The calculation for the required capacity then comes from Tmax once (as a reflection of the average bandwidth) and from Tmax a second time (as a reflection of one top-tier user running a speedtest).

If the old rule-of-thumb used a 3x factor, it hasn't changed to 2x because usage has lowered. The reduction is because the size of the SG has reduced.

The document then goes on to develop the rule-of-thumb to reflect this by directly including the size of the SG:

Required Service Group Bandwidth Capacity = S*Tavg + Tmax (2)

where S is the number of subscribers within the Service Group, Tavg is Per-Subscriber Average Busy-Hour Bandwidth, and Tmax is the Maximum Sustained Traffic Rate (Billboard Bandwidth) offering to the subscribers.

The document goes into more details of the earlier an later periods, when other calculations were used for network sizing.

In reply to a post by Ignitionnet:

Seems a tad arbitrary, takes no account of the size of the service group or the usage patterns.

You are right. It looks like it depended on appropriate scaling of service groups - which very much depends on the investments being made by individual cable companies.

The papers undoubtedly attempt to point at the fact that this was an old "rule-of-thumb" that, while it worked for maybe 5 years, was no longer appropriate.

It isn't at all obvious how well this relates to VM in the UK...

At the time I read it, VM were selling a top tier of 152Mbps with 8 bonded channels (440Mbps); with my limited knowledge, I found the rule applied well.

But I now understand that the service group (of the time) was likely to have 12 channels for the whole service group, from which each modem would bond a subset of 8. The service group bandwidth would then be 660Mbs.

If we believe that VM themselves worked in the (Consumption+Tmax)-­oriented Period, then they would be leaving 500Mbps for average consumption across the service group.

If average busy-hour consumption at the time was 1Mbps per subscriber, then that would need VM to have been working with service groups of around 500 homes.

Does any of that sound appropriate?

In reply to a post by WWWombat:

In reply to a post by Ignitionnet:

Seems a tad arbitrary, takes no account of the size of the service group or the usage patterns.

You are right. It looks like it depended on appropriate scaling of service groups - which very much depends on the investments being made by individual cable companies.

While reading back, it made me rethink about what a "network sizing rule of thumb" might be used for.

Instead of determining what the maximum tier can be, it might actually be used the other way around ... to determine what the size of the SGs need to be in order to deliver the speeds that marketing want to headline.

It then becomes a driver for splitting the groups, and for the network engineering teams to budget those changes.

Understood.

VM have been using the consumption + Tmax model since they introduced DOCSIS 3.

The Tmax x 2 or Tmax x 3 model was made redundant with DOCSIS 3 as it provided 2 options to MSOs to upgrade networks. They could split nodes to reduce service group size, reducing the m in m x avg or add additional channels to the bonded group, providing more headroom for Tmax.

Your other thoughts are correct. Prior to releasing 20Mb ntl Telewest engaged in a large node splitting programme, especially in ex-Telewest areas.

In reply to a post by Ignitionnet:

The ASA rules that state at least 10% of an ISP's customers must be able to reach the speed advertised.

I know VM and others are very open to average speeds being the advertising criteria, the resistance is coming mostly from BT.

I can only talk from my view and experience (as a customer). I would sign up to VM but I know it's potluck if I recieve a good connection.

I don't know what kind of speed I'll be getting, no matter what package I pay for.

I don't know if my area is congested.

I don't know how long a fault, or congestion will be fixed.. I'm hearing stories its months and only actually fixed when a certain amount of people complain about it, or VM hope enough users leave so the congestion fixes itself.

Upload traffic management limits are imposed.

As a potential customer how can VM fix these concerns I have?

In reply to a post by Chrysalis:

except the problem is you end up with regulation which is only agreed upon when it has no affect on their bottom line, e.g. the very soft 10% requirement to advertise a speed.

I am pretty sure the average joe public does not get the same access to ofcom as a executive of a isp does.

What do you class as inadequate, a isp refusing to accept average speed advertising?

The isps wont agree to something that costs them money. If their agreement is required then any regulation is going to be soft. Presumably BT dont like average speeds because it would require some investment ni the copper loop to rectify which they dont want to do.

Do lawmakers consult with the public before making laws? The whole arrangement seems too cosy.

the problem with VM is you cant use avg speed advertising on cable as your connection is a fixed speed

the problem is areas that have utilisation problems and fail to address it promptly which they fail to do, personally i am Fine with them having a soft speed caps for areas that are under high upload utilisation issues it should affect all customers on the FTTn until they add more capacity to the COXA side of that node as its no good if collectively they are running all peoples internet experience on that street level area as DOCSIS does tend to fall apart when its under high load (above 60%) as its like having wifi on a copper line with lots of channels to much activity it just starts to fall apart and getting packet loss

the last time we had utilisation issues it taken 1 year to fix the last utilisation issue we had in are area with the India CS people was making worse by offering free upgrade on speeds which gave More upload to them and collectively made the problem worse in are area (1-2mb ADSL was the only alternative option but at least you could open a webpage on it between 5pm-11pm, and some on our street left VM erly due to it) most of it likely due to P2P software set to auto start on there PC

the problem is with the Upto thing is for people who don't understand that BT and resellers lines speeds are based on distance (or don't want to know and still complain when changing ISP that there speeds are not faster than last ISP even thought they was told over the phone there expected speed) and VM and FTTP there is no up to speed as its a fixed connection speed you're paying for

VM problem is Local utilisation problems in some streets (and them not correcting it fast enough) , wifi been used and the speed outside there network on the internet,

i am considering to switch back to an FTTC product as latency is an big issue on VM (as my Broadband Quality Monitor is a mess and is noticeable in games, where as the other 2 connections i monitor that are ADSL and VDSL based are Flat apart from spikes from normal activity) its all good having 220 down and 12mb up when the latency jitter is all over the place or even packet loss (waiting to see if VM is going to release a new superhub 3 or release firmware to fix the issue with jitter, my area does not have 3.1 yet)

Thanks.

In reply to a post by Ignitionnet:

The Tmax x 2 or Tmax x 3 model was made redundant with DOCSIS 3

The Arris doc specifically says that these models came in with DOCSIS 3, and seems to say they were a consequence (simplification?) of the "consumption + Tmax" model.

As it says: "This is where we are today" (in 2014).

Those models are a simplification but a combination of the model I mentioned alongside QoE measures are where we are.

Rules of thumb are just that and certainly aren't used to make serious capacity planning decisions.

VM use both - the sustained usage plus TMax measure and, mostly actually, the QoE measures, with the Samknows monitoring platform they have helping out hugely.

https://www.arris.com/globalassets/resources/white-p...