BT Broadband Speed Issues

[mr_chinnery]

My dad recently had to swap his BT business broadband package from one telephone line to another (in the same house). Before it switched, download speeds averaged approximately 300 kb/s. Now they average about 60 kbps, which is clearly unacceptable. This happens at all times, peak and off-peak.

However, http://www.speedtest.net/ gives a download speed of 483 kbps and upload speed of 328 kbps. How come I can't achieve anywhere need these speeds when downloading or browsing? Any help will be much appreciated,

Matt.

[Variable]

Rate limiting common file sharing/torrent ports I would imagine.

[Paul Komski]

Is the swap-over to the other line the only change. Same modem/router, same microfilters, etc. The telephone sockets themselves can be at fault. Only yesterday I had cause to scratch my head with the same hardware being attached to the same BT telephone line but on a different extension. On one there was a great connection - on the other the router could hardly connect to the signal, no or slow downloads and with the green light flashing and refusing to go solid. I would consider asking BT to test the line or help with the troubleshooting.

[Sylvander]

To test the Broadband download/upload speeds...

Install Broadband Download Monitor (http://www.broadbandchoices.co.uk/partner-sw_sp~cu/broadband-download-monitor.html).
It puts itself in the Startup list, then runs every time the PC is started, lists the upload/download speeds by transferring a file, lists all the tests for a chosen/tested period [e.g. a calendar month, or year], and calculates/displays the mean values.
.

[Variable]

He has already tested download speed and it is 400/300... So his speed is fine. Unless of course he should be getting more than that but, that is not the question. Since he said his downloads are 60kbps and his download speed tests are far above that, it is probably a torrent or file sharing application that he is referencing with the 60kbps... that and the fact he mentions it is his Dad's connection... Rate limiting file sharing ports is getting common now.

[mjc]

Also, if it is file sharing, remember that you can't go faster than the machines you connect to...if they are of limited upload speed, then your download will be much lower.

[Paul Komski]

Hmm - I get the point with file-sharing but what about the browsing side of things that was mentioned. Not all distant hosts are going to have slow connections. Or is the suggestion that there is file sharing open and affecting things enough even when not downloading using it. And why should there be a change in speed from just using the other telephone line - that (like the poster) is what I don't really understand. Same software and hardware, same house and ADSL provider and just a change in telephone number and line.

I also didn't think that upload speeds affected download speeds much (unless with large reciprocal uploads I suppose) since there are composite dial-up/satellite connections that are utilised using dial-up for the upload part but download from the satellite.

[Sylvander]

My thinking:

If he tests the speeds that his machine and line are capable of producing...
And then discovers that under certain circumstances he is getting much less than that...
Then he knows it's down to the circumstances other than the machine or the line.

If during the test the download/upload rates are poor...
And it isn't the fault of the machine...
Then it's probably the line.

"He has already tested download speed"
Ah, but...
The nature of the test...method used...might determine the outcome.
If this 2nd test [different in method] confirms the results of the 1st, might it be concluded that the results are true?
My statistics teacher used to insist that two tests [even though using different methods] giving the same result do NOT indicate a tendency to confirm; they could both be making the same error, and giving the same incorrect result. :(

The testing program above is giving me results that vary quite chaotically.
So speeds are not reliable & consistent.
.

[Variable]

Yes, the elusive broadband speed tests. Sylvander those test look pretty consistent. There is a problem with speed tests of any kind that rely on an internet site. First you’re assuming that any site you connect to has the maximum available bandwidth, that you need, to properly test your throughput. This is why the test should be looked at with +/- of say 20-30%. It is a snapshot of the state of the route, latency, and allowed throughput for BOTH sides of the connection. To get a true measure you would need to test speed within your first hop or two, depending on your ISP(i.e. your ISP’s first or second router.) Your ISP cannot control what happens to the packet once it leaves the ISP's border router. After that, speed is at the whim of the internet. Your ISP given speed is only to the ISP router that is it. This is very frustrating to explain to people. I basically lost a client this week because of this issue. Some people hear what they want to hear. TCP/IP can sound complicated but can be explained so that non technical people can understand, if you don’t like what you hear you can simply blame the messenger. But I digress.

A java based speed test will not use the same ports as say bitorrent. And it is very common for ISP's providing limited bandwidth (300kbps internet speed is slow compared to normal US speeds) to impose tighter restrictions, especially on common file sharing ports. Whether the technology is DSL, satellite or cable the same specifics hold true.


BTW happy Thanksgiving to any that celebrate it!

Paul, you are right that consumer level internet connectivity is asymmetric. In general, a small return packet is needed to keep a TCP session going. The packet with the data is larger. However, latency can have a much larger impact on a low upload because of the nature of TCP. When you request a file to be saved locally, the packet containing the first bits of data head to your machine with the appropriate information about: how big the packet is,, what order it is in the stream, how large the total size of the file is and a checksum to make sure it is not corrupt. This packet will be bursting at the seams in the data portion of the packet. The file can be thought of as a train, with each box car a packet containing information about its place in the train. As each packet is received correctly the PC sends a response back saying "I got this packet send some more" If one fails or one is received too far out of order the packet must be resent. This causes a backlog. Whistles start blowing; tiny men leap out and apply breaks to the train and start shifting tracks. Once the packets are back in synch, the whole mess gets back on track. So for your ISP connection, the upload speed throughput can be smaller, but the more you download the more you are stressing this small pipe, and any problem is compounded by the fact the information is still flowing. If you had only ONE download path (i.e. one connection from your machine to a far server, like a single download speed test) there would probably be no problem at all However, if you are file sharing, you are opening multiple streams or connections to far hosts. Now you open a browser while downloads are going off in the background. You just opened another stream or connection. Multiple connections compound the issue I described above. Because instead rearranging the box cars for one train, you now have to contend with multiple parallel train tracks also filled with box cars. Any small problem can cause an inordinate amount of other problems.

Now granted, this all happens extremely fast. You have to imagine it slowed down in speeds we can understand but, if you understand how TCP/IP works basically, not nitty gritty details but, understand the core concepts, you can troubleshoot problems much more affectively and you can dismiss wild goose chases before they get started.

Download speed tests should be taken as an average, because your results are “X” it does not mean someone else will have the same results. A physically closer web site will give you better speed. If you do a download test from your machine in Ireland to a machine in Bangkok, you will get a different speed than testing to a server in Cork. Assuming the routers in Cork don’t go through some Machiavellian route (You see odd trace routes in the states.) Now this doesn’t say anything about your “download” speed does it? Your speed could be the same all the time; a download test is simply a snap shot of both sides of the connection. If it is consistent within 20/30%, great; is it accurate? Well, that depends on a lot of things. The deeper you go down the rabbit hole, the more complicated it gets.

[Paul Komski]

Variable - thanks for the interesting reply. Networking has never been my strong point but I think I have grasped the basics of binding to a port and establishing a handshake with SYN, SYN-ACK and ACK so that two computers can exchange data using TCP. As the number of connections increases it would also stand to reason that the "bandwidth" for any one connection would decrease or at least tend to decrease.

What I don't understand in this case is why simply plugging the same computer into a different telephone line from the same company using the same package and with nothing else changed, should compromise the connection in the way described.

[Variable]

What I don't understand in this case is why simply plugging the same computer into a different telephone line from the same company using the same package and with nothing else changed, should compromise the connection in the way described.

**UPDATE** Hmm wait a second. I just re read what you posted. It is the same company but a different package... Let me think about this but I will leave what I typed as general information.


Rate limiting certain ports... and... I would not assume it is a different telephone line. It is probably not. Let's assume you have a DSL connection. You have the ability to shop who you pay for this. The reality is that the line and physical equipment ("last mile") is owned by one company, there may be a law that says this company has to allow competitors to lease this equipment, for competition. So assuming your scenario is true, even though you buy from a new provider, all the equipment in the last bit of the loop could stay the same. The end point could be different, the last hop routers. This is fairly common in the states; the "last mile" is where this comes in to play the new ISP would probably have their own routers where the connection terminates. You could test this with a trace route before and after you change providers. Buying bandwidth higher up the provider food chain is common. There are just a few core providers, or tier 1's.

So back to the scenario, the equipment could indeed be the same, usually the new provider gives you a new router but, it is not always necessary. In his case, he said he could test download speeds and get 300kpbs. So that at least proves the line is capable of this. So we have a baseline. This is where you need to read between the lines. You cannot have it both ways, how can you test download speed at 300k and then say your downloads are 30k? Both of these scenarios download files. The answer is the application used to do these two scenarios is different. What does an application use to communicate with an OS? Ports... So given the scenario that:

The hardware and physical equipment for the last mile is the same.
The backend owned by the new ISP... your endpoint is new.
You prove your baseline at 300k, but you notice new behavior from certain applications.


What can you deduce? The answer is the new ISP is rate limiting ports that the old ISP was not. If the download speed was not tested at 300k, you could assume the new ISP routers are overloaded, insufficient, mis-configured. But since we know that they can test at 300k this is probably not the case.

[Variable]

The same holds true, instead of a different company insert different package. In my companies case, we use the same routers and tier bandwidth based on rules the client gets when they autheticate. IF it is a large ISP they may shunt one login package to different routers but the end result is the same. Different levels of access rules based on what your package is. The new package rate limits ports that the old package does not. The difference may be a 'business" package and a "consumer" grade package. The most common form of rate limiting is dropping. With TCP, when the packet threshold is reached, the router drops packets. Since it is TCP, the dropped packets are re requested. Read the above explanation.

[mr_chinnery]

Thanks for all your input thus far guys.

A couple of things for you:

It's the same house, same building, but it is a different phone number.

Technically it's the same package, in that it (supposedly) does the same thing, BT just gave it a different name. As far as I know.

It is a different router however. Would it be worth trying the other one? Would the old one work with the new 'broad'band set up?

There could be a slim possibility of spyware or adware causing this problem. Norton's running whenever the PC's on and I have a spyware checker. Plus, this computer has the same usage now as when the connection was good.

The really confusing thing is that http://www.speedtest.net/ says I am getting good speeds, but these just don't translate to any direct/streaming downloads I do. Youtube is comically slow, and Ebaumsworld videos take an age to download.

I'm not sure if this helps, but this is the readout from the router thingy:

DSL Connection Details
DSL Line (Wire Pair): Line 1 (inner pair)
Protocol: G.DMT Annex A
Downstream Rate: 1856 kbps
Upstream Rate: 448 kbps
Channel: Interleaved
Current Noise Margin: 19.0 dB (Downstream) 21.0 dB (Upstream)
Current Attenuation: 53.7 dB (Downstream) 31.5 dB (Upstream)
Current Output Power: 17.2 dBm (Downstream) 12.2 dBm (Upstream)
DSLAM Vendor Information: Country: {0xB5} Vendor: {TSTC} Specific: {0x04}
PVC Info: 0/38

Internet Connection Details
Connection Type: PPPoA
Username: B265343@hg57.btclick.com
Internet Address: 81.132.149.119
Subnet Mask: 255.255.255.255
Default Gateway: 213.123.110.58
Primary Domain Name Server: 194.72.0.114
Secondary Domain Name Server: 62.6.40.162
Domain:
Maximum Transmission Unit (MTU): 1500
Gateway Ping: Successful
DNS Communication: Successful
Configuration Server Post: Successful

Hopefully you geys can help me narrow the problem a bit, then I can take it into my own hands!

Cheers, Matt.