In reply to a post by chrisparr:

Only if we don't keep adding bandwidth to keep up with demand! There's more than enough focus to ensure this doesn't happen though, along with assurances from Kelly, Bob and myself about this too.

I asked earlier about the "minimum bandwidth allowance" (a phrase used on this page): http://www.plus.net/support/broadband/speed_guide/tr...

Specifically, what is the "minimum bandwidth allowance" for the bronze queue?

In reply to a post by eddie1150:

This is also of interest to me, im currently on the 250gb package with pro addon.
i want to know if ill loose out speed wise if i go to the unlimited!

Question - why do you have the Pro add-on currently?

Hullo.

Really hard to explain. It's not a straight "minimum kbps" per protocol, instead it is a weighting of the traffic. If you are only downloading bronze, it's 100% of your traffic. If you are downloading just Titanium and Gold traffic, it's 4:1, meaning on an 8Mb line, maxed with Titanium and Gold you'd get 2Mb of Gold and 6Mb of Titanium. In reality though, applications that tend to be in Titanium usually have a small kbps so the Gold traffic would run at a much higher rate (i.e 512k Titanium and 7.5Mbps Gold). The weighting for Bronze is much lower because we believe that in the scenario where you had a mix of traffic including Bronze, you'd much prefer that Bronze traffic to be slower because it is non-interactive.

This is per subscriber line, and overall on the platform. So if we were oversubscribed, Bronze generally would be slower. HOWEVER, my team (including me) look at graphing every day to ensure that we aren't seeing latency introduced, and we work to scale the network to keep this latency to a minimum. We've tested this by having Bob and Chris attempt to download the internet each night and what they were finding was that their sources couldn't keep up with the line speed at peak due to their congestion points.

Edited by deleted (Fri 21-Dec-12 11:45:26)

Hi Kelly,

That's interesting information, thanks.

I think the main worry is that at this price point the product will be very popular (it should be) and a few months down the line the bronze queue bears the brunt when the pipes can't cope. Of course you can keep adding capacity, but this can get very expensive and perhaps there's a point where cost considerations means bronze has to take a big hit rather than continuous, expensive network upgrades.

If the "minimum bandwidth allowance" was specified as kbps, this would at least allay any fears that in a few months time, bronze is "de-prioritised" so that it hits 64 kbps (because the product specifications allow it and the upgrades are costing too much).

Wow, so Bob and Chris have downloaded the entire Internet every night, respect!

Yep. This is why my team need to be really sharp on listening to you guys (customers) for speed/performance issues and on top of our own stats and forecasts, and why sites like TBB and our own forum are gold dust.

thanks for this info, I hope this time you guys get it right.

On the view of non interactive, the few times I have used p2p in the past, generally its not something I start, run in the background and think "ok I will leave this going for days/overnight or whatever" its more like I start the download and I am physically sitting waiting for it to finish, this generally applies for 90% of the downloads I do regardless of protocol (most are http/ftp), if I start a download I am waiting for it to finish, I rarely start a download and am happy to leave it for hours in the background, this is why I have always found it off that isp's see downloading as non interactive traffic.

But then again I am not the type of person to download TB's a month, if I download something then its because I need it for something.

So if I was a plusnet customer I probably would get pro to ensure what I download is a healthy speed even when mixed in with what you consider higher priority and the fact I use a lot of non default ports on services which I guess could mean "other" traffic been used a lot. So I dont think pro is useless either.

Edited by Chrysalis (Fri 21-Dec-12 12:24:55)

In reply to a post by Chrysalis:

On the view of non interactive, the few times I have used p2p in the past, generally its not something I start, run in the background and think "ok I will leave this going for days/overnight or whatever" its more like I start the download and I am physically sitting waiting for it to finish, this generally applies for 90% of the downloads I do regardless of protocol (most are http/ftp), if I start a download I am waiting for it to finish, I rarely start a download and am happy to leave it for hours in the background, this is why I have always found it off that isp's see downloading as non interactive traffic.

When I talk about non-interactive I'm meaning that the action you are doing still works without packets arriving in a timely fashion. I.e. If packets go missing when loading a webpage, it stalls. VoIP stutters and video buffers. Those activities are interactive.

The deprioritisation of non-interactive doesn't immediately mean that your p2p drops to 1Kbps. We are meaning that it can live with some latency and packet drops, so that it comes down at 50Mbps instead of 76Mbps (on a really fast fibre line) or 6Mbps instead of 7.5Mbps (on a ADSL1 line). Don't treat those numbers as some sort of gospel, they are just illustrative.

In reply to a post by Chrysalis:

So if I was a plusnet customer I probably would get pro to ensure what I download is a healthy speed even when mixed in with what you consider higher priority and the fact I use a lot of non default ports on services which I guess could mean "other" traffic been used a lot. So I dont think pro is useless either.

Now, that really depends. If you have Pro, your iplayer and your P2P would be fighting probably leading to buffering. Without it, your P2P would slow enough to let the iplayer run happily. Which would you prefer?

Personally I have never seen p2p swamp my connection to the point streaming has a tiny % of useable capacity, however I admit since I am not a heavy user of p2p or even regular user I may not have come across the situation you described although utorrent and I suspect most torrent apps allow throttling controls.

In reply to a post by Oliver341:

I think the main worry is that at this price point the product will be very popular (it should be) and a few months down the line the bronze queue bears the brunt when the pipes can't cope. Of course you can keep adding capacity, but this can get very expensive and perhaps there's a point where cost considerations means bronze has to take a big hit rather than continuous, expensive network upgrades.

If the "minimum bandwidth allowance" was specified as kbps, this would at least allay any fears that in a few months time, bronze is "de-prioritised" so that it hits 64 kbps (because the product specifications allow it and the upgrades are costing too much).

I don't think what you say there follows, Oliver.

As you are talking about the overall effect on the platform, for a second let's ignore the ultimate bit where the system has to be changed, and just take the ongoing monitoring and further provisioning that Kelly describes.

Under that process, if further capacity is added to cope with additional traffic, the minimum capacity for bronze automatically increases in proportion. So if overall the proportions are about right, everything goes smoothly.

At first sight this seems to imply that if all the additional traffic is in bronze, then it will inevitably slow down. But on further thought this is not the case. From the assumption that all the additional traffic is bronze, by definition the higher levels have no increase in usage, therefore all the additional capacity is available to bronze.

Even more thought shows that it isn't quite as simple as that either, as the higher levels may have been be constrained by the minimum proportion allocated to bronze, so take up some of the extra capacity. But almost certainly the bulk goes to bronze even then.

This is why the monitoring is required, to make sure the algorithms and design assumptions are correct, as well as to schedule the next upgrades.

Coming back to specifying the minimum allocation in terms of kbps, this would then have to be reset manually anyway with each addition of capacity, and if the point were reached that the costs of even more capacity were deemd prohibitive the fact that it is specified as kbps rather than a proportion is irrelevant. The rules/algorithms would be changed anyway.