Re: [Linux-users] DSL connection testing
Ok, just to clarify a couple of points, a 3-minute ADSL crash course in some relevant points only: * The subscriber lines used for ADSL are simply the already existing phone lines. This usually is one twisted pair per subscriber (with a 1:1 mapping of a line to a phone number). This means that the subscriber, or their house, has exclusive access to this line, which goes from the house to the local exchange. As ADSL uses the same wire pair, the number of users on this line is simply the number of users connecting via the ADSL router, and it sounds like that is an apple ipad and an oldish PC in this case... not much by the sounds of it. * However, these twisted pair lines are bundled into bundle cables of something like 25 or 50 twisted pairs to physically keep them together on their way to the local exchange, with just the right number of twisted pairs branching out at the right spot to get to the properties where the connections are required (obviously you can get several lines for one house if you want). Now these bundle cables are designed to minimise cross-coupling and interference between the twisted pairs. BUT, these cables were designed with telephony in mind, where the signal bandwidth is limited by filters to 4 kHz. Now ADSL uses a band that starts with its lower end at several kHz above the telephone band to provide a sufficient safety margin, so that voice and DSL signals can easily be separated by inserting a cheap splitter filter (these filters often come with ADSL routers). But the ADSL band goes up to frequencies just over 1 MHz! And this is way more than what these cables were designed for... So you do get a bit of cross-coupling and interference between twisted pairs that are next to each other within these bundle cables. How much of a problem that is depends on various parameters, including cable length, wire gauge, and on which other pairs in the same bundle other people are using ADSL as well. (It is worthwhile noting that, even without interference, a long line will perform noticeably worse for ADSL than a short one. Also, thicker wire gauge helps to minimise problems with long lines. But the lines are as they are, and Telecom are not going to replace several kms of bundle cable just to get a few more kbps of ADSL throughput...) * I would expect that the usual line testing when a customer first complains about ADSL problems would be a pretty quick check taking only a few seconds to run, and this would establish that: - all relevant components of the local exchange (including the Digital Subscriber Line Access Multiplexer = DSLAM) are working, and - there actually is a line from the local exchange to the customer's premises, and - that that line is physically intact all along the way and does not have any significant measurable defects over its entire length, and - possibly that an ADSL router is connected to the customer end of the line. That is all the ISP are responsible for, and once that is ascertained any possible faults would then be the customer's to resolve. But in this case here, it sounds like probably all of the above may be given (maybe check and confirm?), however there is a quality of service issue. And in order to get some traction on this, I can only second Craig's suggestion to request some more thorough testing, with a specific view towards highlighting the potential of cross-coupling or interference from other users in the neighbourhood. There is one further point to check: how are the computers accessing the router? Ethernet cable? Or WiFi? If they are using WiFi, the there obviously is a very real and fairly likely possibility that something is wrong with the WiFi. And this also includes interference from the neighbours. WiFi has been know to suffer from bandwidth degradation due to interference for quite a while, and this is mainly an issue in densely populated areas, but I can certainly tell when the neighbours are active here in South Brighton... To get back on topic slightly, on the WiFi side I am finding wavemon useful - it runs on a text terminal using pretty minimal resources and provides helpful real-time information on WiFi performance that helps a lot to sort out antenna location problems etc. Hope this helps somehow... let us know when you find out what was causing the trouble! Kind regards, Helmut. On 17/06/14 07:42, dave lilley wrote: anyone here able to suggest anyway of testing number of users connected to a line? Mate who lives in Downs Rd South eyreton commented to me about his link going down and he thinks too many people are on it up stream. mate says he spoke to the provider (not sure who it is) but think they claimed the link was ok, I was hoping to nip out using a linux lappie and test the line for some more solid answers, They use an apple ipad and an oldish PC. thanks for tips of many. ___ Linux-users mailing list
Re: [Linux-users] DSL connection testing
Helmut, you did a very thorough overview of ADSL. Good job. me thinks it took more than 3 minutes to write. The original poster did note that the problems are present when it is ethernet only. Which means that the wifi explanation fades. Not completely - wifi is always a problem and is never fast enough. Given the information in your email, it says that at non peak times the throughput should be better (as it is likely that no oneelse who is in the 25/50 twisted pair bundle is active on ADSL). Since you have carefully explained the mindset of the ISP, it is up to the original poster to find evidence of what failed. This will terminate a mutual blame game where everyone blames someone else. My view is that you need to run a 24 hour test of connectivity and plot the result. Email them the graph, and say this is not good enough. If it is load and crosstalk, there will be a correlation between day and night. (peak times/not peak) Cheers, Derek. On 17/06/14 19:58, Helmut Walle wrote: Ok, just to clarify a couple of points, a 3-minute ADSL crash course in some relevant points only: * The subscriber lines used for ADSL are simply the already existing phone lines. This usually is one twisted pair per subscriber (with a 1:1 mapping of a line to a phone number). This means that the subscriber, or their house, has exclusive access to this line, which goes from the house to the local exchange. As ADSL uses the same wire pair, the number of users on this line is simply the number of users connecting via the ADSL router, and it sounds like that is an apple ipad and an oldish PC in this case... not much by the sounds of it. * However, these twisted pair lines are bundled into bundle cables of something like 25 or 50 twisted pairs to physically keep them together on their way to the local exchange, with just the right number of twisted pairs branching out at the right spot to get to the properties where the connections are required (obviously you can get several lines for one house if you want). Now these bundle cables are designed to minimise cross-coupling and interference between the twisted pairs. BUT, these cables were designed with telephony in mind, where the signal bandwidth is limited by filters to 4 kHz. Now ADSL uses a band that starts with its lower end at several kHz above the telephone band to provide a sufficient safety margin, so that voice and DSL signals can easily be separated by inserting a cheap splitter filter (these filters often come with ADSL routers). But the ADSL band goes up to frequencies just over 1 MHz! And this is way more than what these cables were designed for... So you do get a bit of cross-coupling and interference between twisted pairs that are next to each other within these bundle cables. How much of a problem that is depends on various parameters, including cable length, wire gauge, and on which other pairs in the same bundle other people are using ADSL as well. (It is worthwhile noting that, even without interference, a long line will perform noticeably worse for ADSL than a short one. Also, thicker wire gauge helps to minimise problems with long lines. But the lines are as they are, and Telecom are not going to replace several kms of bundle cable just to get a few more kbps of ADSL throughput...) * I would expect that the usual line testing when a customer first complains about ADSL problems would be a pretty quick check taking only a few seconds to run, and this would establish that: - all relevant components of the local exchange (including the Digital Subscriber Line Access Multiplexer = DSLAM) are working, and - there actually is a line from the local exchange to the customer's premises, and - that that line is physically intact all along the way and does not have any significant measurable defects over its entire length, and - possibly that an ADSL router is connected to the customer end of the line. That is all the ISP are responsible for, and once that is ascertained any possible faults would then be the customer's to resolve. But in this case here, it sounds like probably all of the above may be given (maybe check and confirm?), however there is a quality of service issue. And in order to get some traction on this, I can only second Craig's suggestion to request some more thorough testing, with a specific view towards highlighting the potential of cross-coupling or interference from other users in the neighbourhood. There is one further point to check: how are the computers accessing the router? Ethernet cable? Or WiFi? If they are using WiFi, the there obviously is a very real and fairly likely possibility that something is wrong with the WiFi. And this also includes interference from the neighbours. WiFi has been know to suffer from bandwidth degradation due to interference for quite a while, and this is mainly an issue in densely populated areas, but I can certainly tell when the
Re: [Linux-users] DSL connection testing
In the past I've managed to get ISP's to run more in-depth checks of connections for clients by showing evidence of something wrong on their side. I would have client run a script that did continuous pings to a problem site for a few days, if it replied with destination unreachable it immediately runs a traceroute. Often times this would show an interruption getting through one of the ISP's servers and was enough to get them to investigate. Sometimes it showed fault at clients end which was good too. -Bryce Stenberg. -Original Message- From: linux-users-boun...@lists.canterbury.ac.nz [mailto:linux-users- boun...@lists.canterbury.ac.nz] On Behalf Of Derek Smithies Sent: Tuesday, June 17, 2014 9:26 AM To: linux-users@lists.canterbury.ac.nz Subject: Re: [Linux-users] DSL connection testing Dave, There is quite a bit of information you can glean from looking at the arrival time of packets, and round trip times. A just sit down with a laptop and test experience that lasts 30 seconds is almost completely worthless - as it does not tell you how the link is doing at any other time. I will quote a story from some years ago that happened to me - I may have written it here before, but it is too good not to use again. A customer rang up and complained that his voice over the internet implementation (that used my software) had substandard audio quality. Connected to his network, measured link quality with ping and scp and so on. All ping times were good, file copies site to site were good. Of course, I was connected at nighttime so that my heavy testing did not interfere with his daytime usage. It was the usual finger pointing game. Then I added link testing (record ping times and drop rate) for all links my software ran on. This data was graphed using javascript. Beautiful. Turned out the guy had switched from cable link (which are high quality) to a wireless internet service provider. The wireless links were being flooded with data from all over and packets were going missing. With the graphs, we showed him that our software was fine - he was experiencing high packet loss on the network. The basic first step is to record different performance metrics of the link over a 24 hour period. 1)Test dns lookup of different sites. Does that always work? Every 30 seconds a test should run 2)ping to remote host - how reliable is it? 3)Telecom are offering unlimited home data plan (There was a comment about someone in the states with an unlimited plan and doing 35T a month. Sheesh) The interesting part of telecom's plan is that they may shape your traffic (limit it) during the evening. Which tells me that the peak time for network usage is in the evening. Thus, measurements night time and day time should be different, if it is a shaping/network load thing. Are the night time and day time measurements different? 4)Automated tests to download data from a remote site - what is the measured speed? and how does it vary over a 24 hour period? 5)It is almost certain that your ISP (internet service provider) is traffic shaping your data. You can verify this with a number of open source projects to determne if shaping is happening, and how extreme it is. One could use such tools to plot the level of shaping they apply, and how brutal it is over a 24 hour period. 6)Finally, does your router have a web access page with logs on it? If so, do they tell you anything? Cheers, Derek. On 17/06/14 07:42, dave lilley wrote: anyone here able to suggest anyway of testing number of users connected to a line? Mate who lives in Downs Rd South eyreton commented to me about his link going down and he thinks too many people are on it up stream. mate says he spoke to the provider (not sure who it is) but think they claimed the link was ok, I was hoping to nip out using a linux lappie and test the line for some more solid answers, They use an apple ipad and an oldish PC. thanks for tips of many. ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users -- Sent from my Ubuntu computer ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
Re: [Linux-users] DSL connection testing
dave lilley wrote, On 18/06/14 11:36: Today I'very got more info on this and setup is adsl to modem modem to wireless router. Connection is faster before 7.30pm and after 10.30PM. I'm gonna try and access a terminal and run a ping test for 24 hrs to a log file. Dave - can you email me your IP address directly? I'll set up some ping monitoring from outside. -- CF ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
Re: [Linux-users] DSL connection testing
Hi, me too. I can run some tests from here - and then I will log it and graph it. Derek. On 18/06/14 11:53, C. Falconer wrote: dave lilley wrote, On 18/06/14 11:36: Today I'very got more info on this and setup is adsl to modem modem to wireless router. Connection is faster before 7.30pm and after 10.30PM. I'm gonna try and access a terminal and run a ping test for 24 hrs to a log file. Dave - can you email me your IP address directly? I'll set up some ping monitoring from outside. -- Sent from my Ubuntu computer ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
Re: [Linux-users] DSL connection testing
thanks for the offer. I've forwarded onto steve the offers as it his not mine that's on DSL (i've got a vodafone cable here. I've asked him to contact you directly (off list) if he want the help. I've also sent him some url's on how to do a ping test via ipad and suggested he do 3 days worth of testing ( off peak, on peak and 24hr period) this so he can capture what the results are over a period of time for comparison. On Wed, 18 Jun 2014 14:14:17 Derek Smithies wrote: Hi, me too. I can run some tests from here - and then I will log it and graph it. Derek. On 18/06/14 11:53, C. Falconer wrote: dave lilley wrote, On 18/06/14 11:36: Today I'very got more info on this and setup is adsl to modem modem to wireless router. Connection is faster before 7.30pm and after 10.30PM. I'm gonna try and access a terminal and run a ping test for 24 hrs to a log file. Dave - can you email me your IP address directly? I'll set up some ping monitoring from outside. ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
Re: [Linux-users] DSL connection testing
So how you go about proving that there is a problem where once there wasn't?, and you suspect over subscription/connections to the phone line you're using? - Original Message Follows - On Tue 17 Jun 2014 07:42:03 NZST +1200, dave lilley wrote: Your whole concept is flawed. Drop that idea. Volker -- Volker Kuhlmann http://volker.top.geek.nz/Please do not CC list postings to me. ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
Re: [Linux-users] DSL connection testing
Dave, There is quite a bit of information you can glean from looking at the arrival time of packets, and round trip times. A just sit down with a laptop and test experience that lasts 30 seconds is almost completely worthless - as it does not tell you how the link is doing at any other time. I will quote a story from some years ago that happened to me - I may have written it here before, but it is too good not to use again. A customer rang up and complained that his voice over the internet implementation (that used my software) had substandard audio quality. Connected to his network, measured link quality with ping and scp and so on. All ping times were good, file copies site to site were good. Of course, I was connected at nighttime so that my heavy testing did not interfere with his daytime usage. It was the usual finger pointing game. Then I added link testing (record ping times and drop rate) for all links my software ran on. This data was graphed using javascript. Beautiful. Turned out the guy had switched from cable link (which are high quality) to a wireless internet service provider. The wireless links were being flooded with data from all over and packets were going missing. With the graphs, we showed him that our software was fine - he was experiencing high packet loss on the network. The basic first step is to record different performance metrics of the link over a 24 hour period. 1)Test dns lookup of different sites. Does that always work? Every 30 seconds a test should run 2)ping to remote host - how reliable is it? 3)Telecom are offering unlimited home data plan (There was a comment about someone in the states with an unlimited plan and doing 35T a month. Sheesh) The interesting part of telecom's plan is that they may shape your traffic (limit it) during the evening. Which tells me that the peak time for network usage is in the evening. Thus, measurements night time and day time should be different, if it is a shaping/network load thing. Are the night time and day time measurements different? 4)Automated tests to download data from a remote site - what is the measured speed? and how does it vary over a 24 hour period? 5)It is almost certain that your ISP (internet service provider) is traffic shaping your data. You can verify this with a number of open source projects to determne if shaping is happening, and how extreme it is. One could use such tools to plot the level of shaping they apply, and how brutal it is over a 24 hour period. 6)Finally, does your router have a web access page with logs on it? If so, do they tell you anything? Cheers, Derek. On 17/06/14 07:42, dave lilley wrote: anyone here able to suggest anyway of testing number of users connected to a line? Mate who lives in Downs Rd South eyreton commented to me about his link going down and he thinks too many people are on it up stream. mate says he spoke to the provider (not sure who it is) but think they claimed the link was ok, I was hoping to nip out using a linux lappie and test the line for some more solid answers, They use an apple ipad and an oldish PC. thanks for tips of many. ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users -- Sent from my Ubuntu computer ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
Re: [Linux-users] DSL connection testing
dave lilley wrote, On 17/06/14 07:42: anyone here able to suggest anyway of testing number of users connected to a line? Mate who lives in Downs Rd South eyreton commented to me about his link going down and he thinks too many people are on it up stream. mate says he spoke to the provider (not sure who it is) but think they claimed the link was ok, I was hoping to nip out using a linux lappie and test the line for some more solid answers, They use an apple ipad and an oldish PC. Speak the lingo - customer needs to request a 24 or 48 hour line test in wireline. Only the customer's ISP can do this. This testing will not impact their service. -- CF ___ Linux-users mailing list Linux-users@lists.canterbury.ac.nz http://lists.canterbury.ac.nz/mailman/listinfo/linux-users
