Speed. If we are getting throttled, is that a bad thing?

We’ve spoken quite a bit about speeds recently. Our recent news item surrounding possible throttling on T-Mobile data plans caused a lot of controversy. We were told shortly after that the information provided to us by various sources was incorrect. But, it got me thinking…

In my article about GiffGaff some months ago we saw how the operator was dealing with their “data hogs” and I outlined how “unlimited” packages sometimes work. I’ve learnt a lot about connectivity, bandwidth and networks in my previous day job. Working for an ISP it quickly became apparent that, with ADSL, Fibre Optic, Leased Lines and other forms of super-fast connectivity becoming available, the back-end had to be able to support it.

Some ISP’s advertised “unlimited” data packages but they quickly became unstuck when a very small section of their customer base sucked as much data as 100 or 200 customers combined. A tiny percentage of your customers can suddenly screw your whole pricing model. Whilst most customers use 2GB, or 10GB, or even 40GB per month, a few use 200GB or 300GB per month – maybe more. Somehow, somewhere, those connections come together and have to go over a bigger, fatter pipe. That pipe, let me tell you, is bloody expensive. Increasing it costs a lot, and in some cases you might have to add new switches and perhaps add bigger pipes elsewhere too. Should you increase it just because 0.1% of your customer base are pulling that much data?

When I wrote that article I said this..

Yes, “unlimited” should mean “unlimited”, just like “free” should mean “free”, but there’s four options..

1 – Say it’s “unlimited”, but traffic-shape it to heck so that you can’t pull any real bandwidth.
2 – Say it’s “unlimtied”, but block all the ports and only really allow 80 (web), 25 and 110 (mail)
3 – Say it’s “unlimited” and ramp up the charges.
4 – Say it’s “unlimited”, keep the costs low, but boot off those who are pulling completely, utterly, insane amounts of bandwidth.

T-Mobile recently told us that their new unlimited data plan would actually be capped at 1Mb/s. We were told this a few times by various sections of T-Mobile, however, shortly after we were told that there’s no “max speed” and that all customers get “the best possible speeds”. What does this really mean though?

This afternoon I drove out and sat next to a cluster of mobile masts. I had two mobile phones. One uses an Orange SIM, whilst the other has a Three SIM. Both had full HSDPA connections and I used the Speedtest.net app to check the speed on both handsets from the same server. I’ve filmed it and you can see the results below, however I want to make this clear – this is still very unscientific. There’s various factors involved here. One network might be busier than the other, one mast might have a better connection than the other and – although I did check – there may be some background data going on through one of the handsets that I wasn’t aware of.

So, a clear “win” for Three. Yes? Well yes, that’s true. But does this just show two different methods of dealing with bandwidth?

Here’s the results from Orange..

It appears to show a fairly static “cap” on the bandwidth through the Orange network. Either that or that’s all the bandwidth that particular tower has, but I doubt it. I’m going to call it a “cap” purely because I can’t think of another term for it. It never went above 2Mb/s, and I’m pretty sure that the masts must have more than a 2Mb/s backhaul.

I’ve been told that a package called Orange “GPRS Premium” removes this “cap” and there’s a rather substantial thread over on Digital Spy which suggests that iPhones are given this package to make their speed and experience “better”. There’s even one post from a T-Mobile user who sees exactly the same thing we saw above – a constant download “cap” of 2Mb/s. I’ve found similar posts on WhatMobile.net and a post here which seems to suggest that T-Mobile and Orange customers are getting this “no more than 2Mb/s” limit since the cross-network 3G roaming was enabled.

As I mentioned before, an option for ISP’s dealing with “unlimited data” plans would be to traffic-shape it so that you can’t pull an obsene amount of data. Perhaps, based on what we’ve seen and heard, this is what’s happening.

But wait. Before we get ahead of ourselves, I should show you my second video. This was filmed slightly further away from the transmitters but, during the day-time, I was getting similar speed results to the video above. Both phones had 1-2 bars of signal, both phones were again using the same Coventry server on Speedtest.net but it was about 8.30PM – the height of the “data rush hour” when people are wanting to browse.

Same test. Same two networks ….

Now look. Orange, on the left, reports a download of 1.2Mb/s. Three now reports about the same. So… perhaps I’m jumping to conclusions here, but based on what I know, I’d say that we have two methods of traffic management at play here. If I was a betting man, I’d think that Orange (and perhaps T-Mobile) have chosen the “keep the majority of customers at below 2Mb/s so that we have enough bandwidth to go around, but it might dip by a bit at peak times” whilst Three have chosen the “give everyone the absolute best we can, but it might dip by quite a large amount at peak times”. Who knows, perhaps there’s no traffic management on Three at all.

It’s basically like this. If Bob and Sue are on the same transmitter and it only has 4Mb/s available (and this is a very simplistic example), then with both of them downloading with a “2Mb/s limit” there’ll be no issue. With the other method Bob will get 4Mb/s when Sue isn’t downloading, but they’ll both get 2Mb/s when they’re puling data at the same time.

If my guess is correct, the next argument is… who’s right? Which is the better option? What would you choose? Guaranteed bandwidth? Or stacks more speed at quieter times?

I could be wrong with my assumptions here of course, and I could be accused of coming to some incorrect conclusions based on some speed tests and perhaps battery management, a congested cell, a low signal strength or simply bad guess-work, so… I’d like to hear more from you guys.

Let me know if you’ve done speed tests, I’d be interested in knowing what results you’re getting and at what time of the day. We may well start a survey very soon based on your comments.