FTTC Survey (new)

After the other survey started to become troublesome (mixing those using FTTC and those who aren't), I've had a go at programming my own on a dedicated web page.

Simple form asks which FTTC ISP you use, your usual downstream and usual upstream speeds. Just select your details from the drop downs. You submit it and it gets logged in a database. You only need to click submit once and the page will tell you if everything entered correctly (or not, if i've broken something).

The results from this database are then read out on the results page. The advantage is it's my own domain so I can adjust it easily or add extra functionality if I'm capable. Please only vote if you're actually using FTTC. If your ISP is missing then reply and i'll add it to the list so you can vote for it. Please only vote if you use FTTC and only vote once.

You can access the survey here

Just filled the details in. My typical speeds are actually slower than they can be, due to using wireless 'g' rather than 'n'. With 'n' USB adapter fitted to my netbook, I can hit 34Mbps, but I don't need to at this moment in time.

Done.

My PHP and MySQL programming efforts seem to be holding up so far. Even something as basic as this takes a bit of work. Doesn't help that it's been a few years since I did anything similar.

Added a new page to the site that shows each entry as it was entered. Gives a little bit more information as you can see how each down/up response relates to each ISP rather than the results page which just gives totals. You can order the table columns per your specification.

Edited by orly (Sat 02-Apr-11 19:37:51)

done

Might be an idea to link a speedtest in to the results for some accuracy

Mat

http://www.speedtest.net/result/1233473894.png

I meant, link the survey in to a speedtest automagically

Matt

That looks like a deliberate and brilliant typo .

The trouble is, who would then rack up the speed test usage? Not all connections are unlimited.

You'd only need to run it once, surely? and if the results are anything to go by, the majority of testers won't need to worry because of their "unlimited" FTTC connection

Matt

In reply to a post by uno:

Might be an idea to link a speedtest in to the results for some accuracy

Mat

Don't really see the point- most FTTC users are limited by where they are rather than who they're with.

That's what I was getting at. If orly sets off an automatic speed test, how does that use the user's connection rather than his own, and if it does use the user's connection goodness knows what may be running on it simultaneously.

Thinks - I'm very confused by this idea.

I never said it would start the test automatically.. I said it could tally any results to a speedtest result that the survey participant could take at the time.

I think you've over complicated this in your mind when its actually quite simple on a testing basis.

Matt

In reply to a post by uno:

Might be an idea to link a speedtest in to the results for some accuracy

Mat

http://www.speedtest.net/result/1233545824.png
http://speedtest.net/result/1208090695.png
http://speedtest.net/result/1194556394.png

A couple of random tests + 1 for tonight

Survey duly filled in .

how do you mean

Early days in this version of the survey but seems to be finding some interesting results.

The ISP front is as you were...BT out in the lead.

Only a few dozen responses so far but less than half receive/get 35Mbit or more from their connection. About a quarter say they get under 20Mbit - less than half the headline speeds. Whether people are synced at 40Mbit doesn't really matter if they don't receive it.

For uploads, if we assume anyone voting 2Mbit or less is actually on a 2Mbit product, then just over half of those left get 8Mbit or more - ie top whack. A slightly better performance.

If you have FTTC and haven't voted yet, please do. It literally takes 5 seconds.

Edited by orly (Sun 03-Apr-11 18:26:54)

In reply to a post by orly:

Whether people are synced at 40Mbit doesn't really matter if they don't receive it.

But your survey won't be able to work this out - it only asks for one downlink speed, rather than both the download speed *and* the IP Profile (in the absense of knowing the sync speed).

Actually, the page isn't clear about whether you are asking for the actual download/upload speed, or the value from the IP profile. I know you ask people to perform a speedtest, but that action is the only way to find out either bit of information.

Would it be useful to collect both?

It doesn't try to work out the sync speed.

Simply it asks you to do a speedtest and/or enter the info if you know it. Most users this early, and reading this forum, are knowledgeable enough to find out and understand their speed.

Getting IP Profiles and sync rates is more cumbersome or nearly impossible so it would complicate matters too much and simplicity is what I was going for.

My point was that while FTTC often results in big speed increases, it seems that a minority are reaching the top speeds at this stage. If you have a speedtest over 35Mbit, it's a safe bet you're on a high sync speed too. Obviously you could be synced at the full 40mbit (or 38.7 or whatever is max) and still get a slower speedtest but the point still stands. If you had a gigabit connection and still only got 25Mbit what's the point of calling it a gigabit connection?

As I've mentioned it's all subjective and interpretation rather than 100% accurate. I find it interesting that thus far less than half are reporting they get towards the top end. Given the charts we've all seen showing VDSL2 supposedly capable of handing out 40Mbit up to about 1KM do we infer that ISPs on the whole aren't up to it or that most people live quite a bit further from their cabinets than they think due to roundabout cable routes?

Consider a news post on this site that suggests that only 12% of lines are further than this 1KM distance from the cabinet. I'd have hoped the data coming back to me would be closer to this if it was true. So are speedtests duff, equipment not good enough, ISPs slower most of the time or people running other stuff in the background when testing?

Hopefully as more people take a minute to complete the survey we can all get more data to work with. Subjective of course, but not totally invalid.

BTW, Ofcom's survey from Nov/Dec 2010 includes FTTC data.

Link: Ofcom Summary
Media Slides: PDF

PDF Pages 10,11 and 16,17 show their average speeds (slides are numbered 9, 10 and 15,16).

Even there, Ofcom had nothing to try to correlate actual download speed vs IP Profile (or Sync Speed).

However, Page 8 (slide 7) has a *great* graph of actual download speeds vs exchange distance. Even allowing for the fact that the distances are estimates (so some are a crock of xxxx), the results how that a lot of people are *way* off the curve you'd expect.

In reply to a post by orly:

Getting IP Profiles and sync rates is more cumbersome or nearly impossible

Sync speeds yes, IP profile no... if they're on FTTC they can run a BT speedtest as readily as any other.

(Though I'll concede that the BT speedtester isn't always the most co-operative of beasts )

I know IP profile can be got reasonably easily if the speedtester is cooperative and it would be nice info to have. Might add it to the form as an optional field so that anyone who has the info can add it on if they want.

In reply to a post by orly:

Getting IP Profiles and sync rates is more cumbersome or nearly impossible

But that was my point... Getting the IP profile is NOT any harder - it is displayed on the same screen at the end of the BT speedtest.

So running the BT speedtest means that a customer has both bits of data to hand.

Writing that makes me realise that you don't specify *which* speedtester you expect people to use. Perhaps this is an issue.

If you had a gigabit connection and still only got 25Mbit what's the point of calling it a gigabit connection?

I agree, but unfortunately, the UK broadband industry has decided to sell & price services based on the best capability of the technology, and not on the results achieved. It helps BT recoup money invested in that new technology, but it must smell rotten to those in the slow-spots, who only want it to get the same service everyone else has had for years.

Given the charts we've all seen showing VDSL2 supposedly capable of handing out 40Mbit up to about 1KM do we infer that ISPs on the whole aren't up to it or that most people live quite a bit further from their cabinets than they think due to roundabout cable routes?

That graph might represent the theoretical best-attempt by VDSL2. How about the graph in This Ericsson document (page 8, or 43 on-page). It seems to be more realistic, taking into account crosstalk from other lines. That suggests 22Mb at 1km, and 40Mbps within 200m.

Also remember that BT has deployed profile 8c, which is pretty much the lowest power, lowest speed variant of VDSL2; seemingly to reduce the effect of crosstalk.

I agree, though, that this early in deployment we're probably not too heavily subscribed as yet, so we should be nearer the theoretical best.

But to try to decide whether the blame lies with BT or an ISP simply needs more information than you are collecting.

Consider a news post on this site that suggests that only 12% of lines are further than this 1KM distance from the cabinet.

Fascinating article; not one I'd seen before. The graphs really show something other than the words - they really highlight how big a gap there is to jump from "current best implementation" (VDSL) up to "theory".

But it makes a great point that there's a good proportion of the country who could be looking at good things from VDSL - and quite a proportion who would benefit from plain ADSL2+ hosted there too.

Interestingly, isn't it about 12% of lines on market 1 exchanges? I wonder how much of an overlap there is between those 2 sets of 12%.

I'd have hoped the data coming back to me would be closer to this if it was true. So are speedtests duff, equipment not good enough, ISPs slower most of the time or people running other stuff in the background when testing?

Any & all of those. Add in exchange or backhaul congestion as well as ISP slowdowns.

Did you see the Ofcom figures? By their definition, they're seeing "typical" FTTC speeds as 30-36Mbps (ie 30Mbps for 25%ile and 36Mbps for 75%ile). Of course, as the sample is for BT Infinity customers, the subscribers are already guaranteed to have had an estimate of > 15Mbps - and we know how conservative the estimates were.

My personal feeling is that, with FTTC, we're also at the point where the "last-mile" is no longer the significant bottleneck that it used to be. And that the old datacomms practices of "best-effort" and dropping packets are becoming more visible to us. We used to believe that all the equipment from your exchange to the speedtest server and back again was "perfect", and that the speedtester's result gave an accurate reflection of the "last-mile". Now it merely reflects the accumulation of all the limitations in the end-to-end connection - and now the blemishes in the rest of the infrastructure become a little more obvious.

Hopefully as more people take a minute to complete the survey we can all get more data to work with. Subjective of course, but not totally invalid.

And a very good aim, too. I'll be trying to collect my own statistics, when I'm connected (and I'll fill in your survey too). I do for my current 8Mbps service - but it'll be annoying to lose the modem stats. But it makes it many times better if lots of people are collecting & reporting the stats.

In fact... thinking about the Ofcom slides I posted earlier, I was most impressed by that scatter graph that showed Speed vs Exchange-distance with "dots" for each person.

I wonder if we can create a similar speed vs cabinet-distance one.

Cleared it up a bit to default to the BT speedtest. Also included a link to Kitz' site which shows you how to get the IP profiles too.

The "view each entry" page show show each person from now on who adds their IP profile info. This of course assumes that any of the code i've written to sanitise the information put in the text boxes works right and doesn't wreck it all. Don't want some sneaky [censored] to come along and compromise the whole thing with a crafty SQL attack.

Log all input to a transaction database? Then not too difficult to trace the culprit if this happens, and reconstitute the results database?

"Simply it asks you to do a speedtest and/or enter the info if you know it."

If I use the TBB Java speedtest, my XP netbook upload speed is capped at approx 4Mbps. If using the alternative Flash test, I get 8-9Mbps upload. Speedtest.net usually shows approx 6Mbps.

Seems that last night I had in fact broken it when trying to secure it a bit...would seemingly accept the data provided but wouldn't enter anything into the database except blanks.

Had to delete a few responses so if you voted earlier on monday morning then you might need to look through all the entries and see if yours hasn't shown up right.

The form *should* work now. Don't expect shenanigans from users here but no point leaving an HTML form with text boxes linked to a database open to the world to play with. Hopefully my attempts this evening have made it better.

still not sure what you mean

I think he means that participating in a survey would automatically trigger a speed test using your (orly's) choice of test, and that somehow you would know the results even if the respondent didn't post them.

Perhaps, as you know the IP address of the respondent, you could trigger this. I must admit I'm confused as well. Hence my reply to uno at the time.

Maybe as a speedtest.net host he could do that, but .... Quite apart from the questionability of Ookla.

Which reminds me, as I'm no longer on Avira, which destroys Ookla just like Kaspersky, connected wirelessly, speedtest.net,
TBB - 08/04/11 22:01 Fri 14136 Kbps 8249 Kbps,
BT - Download speed achieved during the test was - 35130 Kbps, Upload speed achieved during the test was - 8069 Kbps.

Now that's interesting, to me at least!

Don't see a simple way to do it.

Even if I could "force" a browser to automatically do a speedtest on another site, I'd have no way of automatically getting the results from what I can see. Don't have the bandwidth to host one myself obviously

Yes, that's what I thought.

Are there actual performance questions?

I have just signed up , so should I wait until actually connected?

Ta.

Which ISP

Usual downstream

Usual upstream

IP Profiles (optional...you don't need to fill them in if you don't know)

Thats all it asks.

I'm curious about the download speeds reported in the survey and, for that matter, in posted/linked-to speedtest.net results.

Taking into account the ATM, PPPOE and TCP overheads then over 35Mbps should essentially be impossible ie

40000x(48/53)x(1492/1500)x(1452/1492) = 35067kbps

Presumably there's some more lost in the AAL5 encapsulation and in the FEC but my brain always hurt when thinking about AAL5 and I don't actually know what the FEC overhead is.

So how come people are reporting 37Mbps+? Do the speed testers attempt to "correct" for overheads in their calculations.

FWIW I'm syncing at 39879, IP profile 38717. BT speedtester 30Mbps, speedtest.net 34-35Mbps and TBB speedtest all over the place but 32Mbps this morning.

I forget the detail right now, but TC/IP overheads on BT Wholesale systems are accounted for within the IP Profile. There is also something different from normal about the Openreach-reported sync speeds - the phrase "data rate" springs to mind in relation to the 40Mbps maximum.

If you look at your upstream IP Profile as well, it equals the reported sync speed. At least it does for the 10Mbps upstream service.

Speedtest.net is known to give high readings, even when not affected by the user's IS, as it selects samples from its testing rather than the whole lot.

Too hot and too much to do to search for all aboiut the Openreach VDSL at the moment.

VDSL 2 doesn't have to use ATM it can, and usually does, use Ethernet which wipes out some of the overheads.

That was it I think .

Interesting, that would reduce the overheads quite a lot - doesn't do anything to make me feel better about only getting 30Mbps out of the BT speedtester though!

I forget the detail right now

I think the detail's in SIN 498. It's clear that this is ethernet over dsl - in fact the SIN even gives a figure of 39,282kb/s ethernet throughput based on a 40,000kbps sync and on the wire frame size of 1532bytes.

In reply to a post by Anonymous:

I forget the detail right now

I think the detail's in SIN 498. It's clear that this is ethernet over dsl - in fact the SIN even gives a figure of 39,282kb/s ethernet throughput based on a 40,000kbps sync and on the wire frame size of 1532bytes.

Yes, I'd found that but was "tired/in a rush/too much vino" and misread it because of the handover to CP description just before the relevant stuff you have quoted from.

Thanks .

Trend seems to be less than half getting top speeds in either download or upload.

I'm quite skeptical of the "15mbit or less" numbers however. Seems unlikely that around a quarter of people would be seeing this given the checker usually denying anyone under this threshold.

In reply to a post by orly:

I'm quite skeptical of the "15mbit or less" numbers however. Seems unlikely that around a quarter of people would be seeing this given the checker usually denying anyone under this threshold.

Why?

1) There are ISPs, such as ourselves, who will take orders below 15Mbps
2) This could show (as we are already seeing) that there are places within the UK where slowdowns are already starting to appear on FTTC
3) The speedtest checker does not deny results below 15Mbps and neither does the Availability checker.

Indeed...but as no one who voted for Aquiss stated they get 15mbit or less, that rules you out so far.

Most of them are from BT users...and BT won't let you order Infinity if it suggests you'll get less than this speed without a fight. So either people are just leaving it on the default option in the poll or are getting slow speeds from any speedtest they run.

Edited by orly (Sun 17-Apr-11 20:07:46)

Updated the survey to include UTV who now provide FTTC products. Thanks to Nelix01 for letting me know. If there are any other ISPs I'm unaware of, please let me know.

(Robertos, you might want to add UTV to your site too)

Edited by orly (Tue 19-Apr-11 18:48:04)

Done.

Never been so happy to fill in a survey!

Now have your IP address muhahaha. Nah, won't be divulging it or doing anything naughty. It's just used to verify the IP is from the UK or to spot duplicate voters

Thanks for voting. Everyone else do the same!

Edited by orly (Tue 19-Apr-11 18:58:38)

Did a bit of spring cleaning and removed some duplicates that people were putting in.

If you want your entry updated/removed just send an email as the line at the bottom of the website asks

Shouldn't be a problem. The form as it's currently designed shouldn't accept duff input.

(Famous last words)

also, bump.

Added "Merula" to the list of ISPs.

(robertos, you might want to include them on your list too)

Thanks, I spotted it when I checked to see you hadn't already got Goscomb. (Earlier PM has the link).

Added them. Not sure if they do unmetered uploads or have a charge for a router. I'm guessing they just don't supply one at all

Added TalkTalk

added NewNet. Seems i'd added it to the spreadsheet but not the survey.

Also removed some duplicates. Seem to be a few virgin media clowns around voting for a number of ISPs