Sync speed, IP Profile and actual measured speed discrepancy

Hi all, (this is my first post so please be gentle with me)

I've got an ADSL2+ connection and live only 300m from the exchange (as the crow flies) so I expect to get a decent speed.

My modem syncs at 20898kbps and my IP profile is 19.5Mbps (according to BTW diagnostics) but the actual measured speed (on BTW speed test) is only 11Mbps. I have a Netgear DG834 modem which indicates line attenuation = 13db and Noise margin = 3db.

I've had ADSL2+ for a couple of months now so expect it to have settled down, but the achieved speed has never exceeded 11Mbps.

My understanding is that the achieved speed should be somewhere close to the IP profile speed as the IP profile is the throttling mechanism to prevent the local-loop bottleneck from causing buffer overflows and packet loss. Am I right or can anyone correct my understanding here?

I've talked to my ISP, Freeola (Entanet), and but they say the achieved speed is within acceptable limits and can't be investigated. One suspicion I have is that my ISP is capping my speed at 11Mbps, if so should I consider changing ISP? Obviously I'd like to get the highest speed possible.

Thanks for any feedback on this.

Edited by deleted (Wed 19-Sep-12 11:16:40)

If the BT speed tester indicates 11Mbps for throughput then it is very unlikely that the ISP is capping your speed.

I assume that you are using a wired ethernet connection between the router and the computer and that you have perhaps eliminated any possible negative effects of telephone extension wiring by plugging the router into the master NTE5 (?) test socket. Also have you tried different filters and adsl rj11 leads at the test socket?

While unlikely, it can depend on the ISP and how they interconnect to the WBC network, as with some the speedtest is totally on BT Wholesale network, for some the ISP does end up influencing.

The router stats suggest that probably not a wiring issue, and more to do with a throughput issue on the PC or the backhaul network.

I presume the speedtest was done when using Ethernet, as wireless can often be the bottleneck in the home.

Thanks for your reply 4M2.

I have a 16-port Netgear Pro Gigabit switch connected between the modem and the PC that's measuring the speed all wired up with CAT5e.

The modem is connected directly to the BT master socket with a 1m cable. I've not disconnected the ring cable though, I've read that this can help SNR.

Is it normal for there to be such a large difference between the IP profile and the actual measured speed? My understanding is that the IP Profile is based on the sync speed and the sync speed is based on the measured parameters such as SNR and error rates over time. So ultimately the IP profile is a measure of your local-loop's potential capacity. If I had a noise problem on the line, surely the sync speed and hence IP profile would be lower.

As such, I don't see why the measured throughput is only just over half of the IP profile speed.

you could find a large file and download it, looking at the throughput on your connection while it happens. Do this at 7am and see if it caps out at a particular value. http://www.thinkbroadband.com/download/

What's your OS ? The MTU / RWIN settings can limit throughput, esp with Win XP or earlier - check on a tweak testing site.

Even better download two files at once, to help saturate the link then monitor the stats on the switch

In reply to a post by Decepticon:

Is it normal for there to be such a large difference between the IP profile and the actual measured speed? My understanding is that the IP Profile is based on the sync speed and the sync speed is based on the measured parameters such as SNR and error rates over time. So ultimately the IP profile is a measure of your local-loop's potential capacity. If I had a noise problem on the line, surely the sync speed and hence IP profile would be lower.

As such, I don't see why the measured throughput is only just over half of the IP profile speed.

Your line is "very good" (see: http://www.coolwebhome.co.uk/calc/calculator.php ) and there are certainly no problems with SNRM, sync, attenuation and IP profile so there's absolutely no need to be concerned about those

But something is affecting your throughput and I think we have to explore all the possible causes...

In reply to a post by MrSaffron:

Even better download two files at once, to help saturate the link ...

That's interesting because if I do a TBB speed test at the same time that my AV app (KIS 2012) is downloading an update I only loose ~200Kbps, i.e. ~11000Kbps down to ~10800Kbps on the result.

an AV app updating isn't always the same as a download - a lot of updates trickle along rather than downloading full blast.

In reply to a post by yarwell:

an AV app updating isn't always the same as a download - a lot of updates trickle along rather than downloading full blast.

Especially Kaspersky updates!!!

Thanks for the info yarwell.

In reply to a post by Decepticon:

My understanding is that the achieved speed should be somewhere close to the IP profile speed as the IP profile is the throttling mechanism to prevent the local-loop bottleneck from causing buffer overflows and packet loss. Am I right or can anyone correct my understanding here?

It's nowt to with throttling etc as you say. The diff between throughput and Sync speed is accounted for by the protocol overheads (ADSL & TCP/IP) of the transmission, of which the IP Profile accounts just for the ADSL overheads.

As a rule of thumb:
IP Profile = 88.2% of Sync
Throughput = about 83.5 % of Sync

So you should be downloading at about 17.5 Meg.

That you are not is likely to be down to congestion somewhere in the network.

Would the OP have suffered with the effects of congestion continuously for a couple of months though?

In reply to a post by Decepticon:

Hi all, (this is my first post so please be gentle with me)

I've got an ADSL2+ connection and live only 300m from the exchange (as the crow flies) so I expect to get a decent speed.

My modem syncs at 20898kbps and my IP profile is 19.5Mbps (according to BTW diagnostics) but the actual measured speed (on BTW speed test) is only 11Mbps. I have a Netgear DG834 modem which indicates line attenuation = 13db and Noise margin = 3db.

I've had ADSL2+ for a couple of months now so expect it to have settled down, but the achieved speed has never exceeded 11Mbps.

My understanding is that the achieved speed should be somewhere close to the IP profile speed as the IP profile is the throttling mechanism to prevent the local-loop bottleneck from causing buffer overflows and packet loss. Am I right or can anyone correct my understanding here?

I've talked to my ISP, Freeola (Entanet), and but they say the achieved speed is within acceptable limits and can't be investigated. One suspicion I have is that my ISP is capping my speed at 11Mbps, if so should I consider changing ISP? Obviously I'd like to get the highest speed possible.

Thanks for any feedback on this.

Your noise margin is bare minimum (3db) and I have seen dropping connections and low throughput at these levels. This may be causing your throughput problem as its possible that there are bursts of noise or other forms of REIN/SHINE on occassion that might interfere since there's very little margin for error with a 3db noise margin.

Generally 3db is not even selectable to put someone on to, the minimum is 6db.

EDIT: It could even just be an over-subscribed street cabinet.

Edited by deleted (Wed 19-Sep-12 15:37:43)

On lines that BT Wholesale determine to be stable they lower their standard 6dB target to 3dB now. So it is not a total indicator of a fault/problem

Also an over subscribed street cabinet would not affect ADSL throughput, as the ADSL signal is 1:1 contention until it reaches the MSAN at the exchange.

Generally 3db is not even selectable to put someone on to

3 dB profiles have been available for years, but only recently widely used. Quick gain to meet the 2M USC perhaps.

In reply to a post by MrSaffron:

On lines that BT Wholesale determine to be stable they lower their standard 6dB target to 3dB now. So it is not a total indicator of a fault/problem

Also an over subscribed street cabinet would not affect ADSL throughput, as the ADSL signal is 1:1 contention until it reaches the MSAN at the exchange.

Strange considering I've seen this affect the noise margins unpredictably in the past when sending out engineers, maybe I didn't get the full side of the story.

Is there any reason one might not be able to select a 3db noise margin on upload / download when performing SNR resets from a CP point of view?

Cannot comment on what your provider provided to you in your support interface.

A full cabinet can increase cross-talk, but we are looking at a line with almost full sync and no complaints of poor noise margin, rather than appears to be a backhaul throughput issue.

Remember the DSL segment is perhaps 1% of the network between home and internet. Yes it does account for a lot of issues, but in this case I do not believe it is the issue. The trick to forum support is jumping outside the standard script.

In reply to a post by MrSaffron:

Cannot comment on what your provider provided to you in your support interface.

A full cabinet can increase cross-talk, but we are looking at a line with almost full sync and no complaints of poor noise margin, rather than appears to be a backhaul throughput issue.

Remember the DSL segment is perhaps 1% of the network between home and internet. Yes it does account for a lot of issues, but in this case I do not believe it is the issue. The trick to forum support is jumping outside the standard script.

Ok, well the support software we had only allowed a range between 6db and 15db to be selectable so my information may now be out of date.

There is not a great deal of investigatory evidence so far:

The only thing I picked up is that his ISP said there can be no investigation, I would question whether he is using a popular speed test site as opposed to the BT performance tester, as those are recorded and can be seen by support staff, typically a test on that resulting in more than a 30% loss would result in a fail which can be investigated.

What I do know is that if its a congestion issue thats going to be beyond our scope of support here on a forum and rests firmly in the hands of the ISP and BT.

In reply to a post by Kaytfoh:

I would question whether he is using a popular speed test site as opposed to the BT performance tester

Well, OP did state "but the actual measured speed (on BTW speed test) is only 11Mbps" so we can only go by what he says and not speculate he is telling porkies.

You do seem very last century on your info, viz, 3 dB NMs, NTE5s, MTUs ...

In reply to a post by XRaySpeX:

In reply to a post by Kaytfoh:

I would question whether he is using a popular speed test site as opposed to the BT performance tester

Well, OP did state "but the actual measured speed (on BTW speed test) is only 11Mbps" so we can only go by what he says and not speculate he is telling porkies.

You do seem very last century on your info, viz, 3 dB NMs, NTE5s, MTUs ...

Then he basically needs to push the support staff more because that would show as a fail on any BT recorded speed test, they are not doing all that they can.

Noise margins still play a part in speed and in particular those with high amounts of errors and HEC errors on lines at 3db. Some people still don't even have NTE5s, and the information for MTU was correct as with IPStream, maybe that has changed with WBC but not really had the need to re-do that test on someone with WBC.

So rather than just say my information is wrong provide me with evidence or reasoning on how my answers are incorrect and what they should actually be rather than just say its out of date. Then we can all learn something.

In reply to a post by Kaytfoh:

Then he basically needs to push the support staff more because that would show as a fail on any BT recorded speed test, they are not doing all that they can.

OP said: "I've talked to my ISP, Freeola (Entanet), and but they say the achieved speed is within acceptable limits and can't be investigated." - might need a very strong push then!

1. user is using a DG834 so no error counts
2. Sync >20 Meg on 13dB attenuation is fine, suggesting no major noise issue
3. Backed up 3dB target that line is running
4. ISP is one of those that would have walked through the usual steps

Most of us who answer questions spend a fair bit of time using the wonderous Google and read ultra interesting documents on SINET. Also a lot of help is based around pattern matching in our heads over time, and observing the symptons and seeing how right we were or not when poster comes back to us.

MTU issues usually highlight with https sites not working too, rather than a classical throughput issue. Also if was noise based then I'd hope ISP would have noted the error counts that are visible to them over the BT Wholesale interfaces, a good ISP can also see a lines history.

As your experience sounds pretty old, i.e. not having seen MTU issues with WBC which has been around for ages now, I suggest a day or two reading at www.sinet.bt.com

In reply to a post by 4M2:

In reply to a post by Kaytfoh:

Then he basically needs to push the support staff more because that would show as a fail on any BT recorded speed test, they are not doing all that they can.

OP said: "I've talked to my ISP, Freeola (Entanet), and but they say the achieved speed is within acceptable limits and can't be investigated." - might need a very strong push then!

Unless I'm again sadly out of date in terms of information, 11mb out of 17mb would basically be a breach of the lower threshold which by ISP and Ofcom standards I believe is 30% of the rated speed for the line.

Has that suddenly changed too? lol

In reply to a post by Kaytfoh:

So rather than just say my information is wrong provide me with evidence or reasoning on how my answers are incorrect and what they should actually be rather than just say its out of date.

2 out of 3 ain't bad!

And on the other, MTUs, yes they certainly did change with WBC but you were only thinking IPStream as you now admit. So I wasn't far out by saying you were still in the 20th Century .

In reply to a post by Kaytfoh:

In reply to a post by 4M2:

In reply to a post by Kaytfoh:

Then he basically needs to push the support staff more because that would show as a fail on any BT recorded speed test, they are not doing all that they can.

OP said: "I've talked to my ISP, Freeola (Entanet), and but they say the achieved speed is within acceptable limits and can't be investigated." - might need a very strong push then!

Unless I'm again sadly out of date in terms of information, 11mb out of 17mb would basically be a breach of the lower threshold which by ISP and Ofcom standards I believe is 30% of the rated speed for the line.

Has that suddenly changed too? lol

What do you suggest then: perhaps the OP could send the results of a few BT speed tests to the ISP to prove the point, if they are ignoring or not seeing them?

We are still in the 20th century for a bit, upgrading the exchanges to be 21CN doesn't fix the 20CN cabling infrastructure. Its like saying we'll upgrade to a gigabit ethernet card when your running the link on CAT1 cabling. Only as good as the worst component in my eyes.

But despite that, this is a throughput issue, and yes I have seen people on 3db margins having issues but then after a forced reset to 6db have a better time with their connection until the equivalent of Rambo kicks in and knocks it back down to 3db, so I guess I was limited in the information I could gather with old eco plus and KBD

In reply to a post by 4M2:

In reply to a post by Kaytfoh:

In reply to a post by 4M2:

... nested quotes trimmed ...

OP said: "I've talked to my ISP, Freeola (Entanet), and but they say the achieved speed is within acceptable limits and can't be investigated." - might need a very strong push then!

Unless I'm again sadly out of date in terms of information, 11mb out of 17mb would basically be a breach of the lower threshold which by ISP and Ofcom standards I believe is 30% of the rated speed for the line.

Has that suddenly changed too? lol

What do you suggest then: perhaps the OP could send the results of a few BT speed tests to the ISP to prove the point, if they are ignoring or not seeing them?

This would have to be handled by a more thorough BTW investigation because I believe an SFI will not be able to see the issue. Just speculation here but maybe they are on a 2nd gen line card with the bit swapping issues (Fujitsu Siemens mark 1 cards).

Either way out of our hands

In reply to a post by Kaytfoh:

yes I have seen people on 3db margins having issues

Yes, and we've seen many people, incl. myself, on WBC 3 dB NM not having any issues.

And with the ones that do have issues, it's usually discons rather than low throughput, tho' it is true that sometimes high error rates w/out discons forcing retransmissions will damp the throughput.

High amounts of initialisations and retrains can be seen by ISPs, well ... If they are any good that is

Wow! Thanks for all the discussion and feedback from everyone, I'm amazed by the extensive knowledge out there in the ether.

So, from reading all the posts here, it seems that I need to push hard on my ISP to investigate the situation. One point I perhaps didn't make 100% clear, is that my ISP said that the current speed was such that *BT* would not investigate it - but it sounds from what people have said here that that is bobbins and I should be expecting a minimum throughput speed of 30% less than the line speed.

The PC that I ran the BTW tests from mostly is running Windows 7, but I have also tried from another machine running Linux (CentOS 6) and got the same result. My ISP asked me to run the BTW tests when I first queried the speed with them, they told me they do have access to the results.

The master socket and router are in an under-stairs cupboard, co-located with a 16-port Gigabit switch (Netgear Prosafe - unmanaged, so I can't read throughput from it) and an RJ45 patch panel. The mains electricity fusebox is also in close proximity - I wonder if any of these factors affect the noise levels on the DSL line?

Easy enough to see if the electrics is having an effect monitor the SNR for a few hours and see if it straws from the 3dB figure at all.

The 30% less than line speed, is NOT referring to throughput, but rather line sync if memory serves me right. The throughput is allowed to go a lot lower, if memory is serving me right.

If the problem is contention, which is what would cause the speed to drop in a normal situation you would usually find it performed well between about midnight and 9am. If it is consistently 11Mbps and have ruled out your PC by checking on another connection of similar speeds, then soe mis-configuration in the BT Wholesale network seems most likely.

Unlikely but it could be lots of errors on the line as already mentioned. Use RouterStats, for example, to monitor SNRM over a period. See link: http://www.vwlowen.co.uk/internet/files.htm

In reply to a post by Decepticon:

I have a Netgear DG834 modem

Which DG834?

You may be able to use Telnet on it, see here Netgear DG834/DG834G, to get its full router stats incl. error counts which people here have expressed interest in.

Just to confirm, are you performing the tests over a wired ethernet connection to the router with any wireless on the PC switched off?

Just in case your node is busy when you happen to be testing (unlikely to be happening consistently though) you can view the Freeola/Entanet WBC node usage here. All nice and quiet as I write this.

Shame if Freeola did say they couldn't help - the Freeola rep views this forum so hopefully he will be along to comment

Faraday WBC node has been having a few issues of late, but don't think this is linked

Agreed - that was only for the one afternoon.

The 30% less than line speed, is NOT referring to throughput, but rather line sync if memory serves me right. The throughput is allowed to go a lot lower, if memory is serving me right.

I'm with you on that, the throughput is the lower limit shown on BT speedtests like 600k on a 7150 IP profile.

Unless I'm again sadly out of date in terms of information, 11mb out of 17mb would basically be a breach of the lower threshold which by ISP and Ofcom standards I believe is 30% of the rated speed for the line.

I think you're talking bout the FTR which is a 30% drop in sync speed not in throughput.

Have seen it appear once or twice more, but shorter durations it seems

In reply to a post by yarwell:

Unless I'm again sadly out of date in terms of information, 11mb out of 17mb would basically be a breach of the lower threshold which by ISP and Ofcom standards I believe is 30% of the rated speed for the line.

I think you're talking bout the FTR which is a 30% drop in sync speed not in throughput.

In this argument would the "rated speed for the line" be the "Attainable Rate" or "Max Rate" - the theoretical sync rate that could be attained if any line card or DSLAM restrictions were removed?

So actual sync rate should not see a drop of more than 30% from the theoretical "Attainable Rate"? And hence the 30% figure has nothing directly to do with throughput speed since that may be quoted within an acceptable range and the lower limit may be considerably less than 70% of sync speed.

Edited by 4M2 (Thu 20-Sep-12 12:44:53)

We are still in the 20th century for a bit, upgrading the exchanges to be 21CN doesn't fix the 20CN cabling infrastructure.

OK, if DLM sees a circuit as super stable, it will wind the snr down to 3db to produce it's highest sync. That the OP sees this, is an indication that the pair is in fine fettle, and not the cause of the issue.

and yes I have seen people on 3db margins having issues but then after a forced reset to 6db have a better time with their connection until the equivalent of Rambo kicks in and knocks it back down to 3db

I'm guessing you work for ISP of which many punters use a Cisco as modem/router of choice. If correct, this would explain the basis of your theory. Cisco's don't like a 3db profile. They will error, and quite often won't hold sync.
No need to worry about Rambo, get BT Wholesale to hard pin the snr to 6db.

In this argument would the "rated speed for the line" be the "Attainable Rate" or "Max Rate" - the theoretical sync rate that could be attained if any line card or DSLAM restrictions were removed?

None of the above. It would be a 30% reduction from the minimum stable line rate recorded in the initial training period.

In other words if a line originally synced at a lowest speed (MSR) of 16000 during its initial training it has to drop to a sync speed of 11200 for that in itself to be reportable as a fault.

The theoretical speeds reported by modems don't really enter into it - they are quite variable and commercial services can elect to run at SNR margins of their choosing and take the drop as experienced under their regime as the trigger for a fault.

In reply to a post by yarwell:

In this argument would the "rated speed for the line" be the "Attainable Rate" or "Max Rate" - the theoretical sync rate that could be attained if any line card or DSLAM restrictions were removed?

None of the above. It would be a 30% reduction from the minimum stable line rate recorded in the initial training period.

In other words if a line originally synced at a lowest speed (MSR) of 16000 during its initial training it has to drop to a sync speed of 11200 for that in itself to be reportable as a fault.

Yes I was aware of that with ADSL MAX connections - guess it applies to 21CN WBC also...

So in the context of this thread MSR would perhaps be the "rated speed for the line" and if that speed was to drop by 30% or more then a fault could be reported?

Therefore an ISP should be aware of the MSR and if consistently slow throughput speeds are reported by an average user then an investigation should be conducted into the current sync speed and whether it has dropped by 30% or more from the MSR.

If an ISP's support personnel imply that throughput should not be less than 30% of the "rated speed for the line" aka MSR then perhaps they are misinforming the customer...

If an ISP's support personnel imply that throughput should not be less than 30% of the "rated speed for the line" aka MSR then perhaps they are misinforming the customer...

They would be making a rod for their own back, yes.

The reliance on speed test data to indicate line condition depresses me.