Kitz Forum
Broadband Related => Broadband Hardware => Topic started by: SE on November 07, 2013, 05:15:03 PM
-
Hi guys
I've found a problem with the Technicolor TG582n FTTC supplied by PlusNet
When using Address Based Filtering it slows down fibre from 47Mbps to as slow as 8Mbps and 0.5 Mbps up (max 20Mbps)
The settings were set to block
As soon as I turn Address Based Filtering off it's back to full speed
To me it looks like the router cpu is underpowered
Has anyone else come across this at all?
-
Hi SE,
I'd be interested to know what throughput results you obtain from e.g. a BT speed test.
Kind regards,
Walter
-
Hi SE,
I'd be interested to know what throughput results you obtain from e.g. a BT speed test.
Kind regards,
Walter
Hi, m8ty
I made the tests (pics link) last night before I read your post, hope speedtest.net is okay
www.tiikoni.com/tis/view/?id=31173c8
I tested with bt and got the same results but never snapped them :blush:
also since getting fibre, the bt test never goes to part two, don't know why ???
The max I saw was 20Mbps, low as in the snaps was 8
The 47Mbps was with the routers Address Based Filtering turned off
turn it on, back to slow, off, back to full speed
I only had 2 sites blocked, facebook.com and facebook.co.uk
I must get a faster router :'(
-
Does seem like it's CPU limited, if you can telnet in and run TOP you can monitor the CPU usage while running tests. It does take some processing power to run filters and such like at FTTC speeds, my asus rt-n16 can't run an fttc line at fulll speed with qos enabled and that's meant to be quite beefy.
-
Does seem like it's CPU limited, if you can telnet in and run TOP you can monitor the CPU usage while running tests. It does take some processing power to run filters and such like at FTTC speeds, my asus rt-n16 can't run an fttc line at fulll speed with qos enabled and that's meant to be quite beefy.
Hi
I haven't installed the open firmware atm so cant get much info
I was going too, then thought id get a all in one router/modem a some point :-\
I think the Technicolor TG582n FTTC is a modded ADSL router, so I guess it was never made for that sort of speed
The thing is, wifi is off and I only have one cat6 from it to a switch, then tp the PC's
I guess it would be even slower if wifi was on, unless it has a dedicated chip to do that work
-
Im not so sure its CPU.
Something else is going on.
This is my base test - 72.7Mbps /17.19Mbps
http://www.speedtest.net/my-result/3090197134
Enabled Address based filtering and I got 14.82/0.42 Mbps
http://www.speedtest.net/my-result/3090201269
I download a file and I get a practically rock-steady 7.01Mbps on Netmeter.
hmmmmm
So speedtest.net uses parallel threads 2 x 7Mbps = 14Mbps
To test this theory.. TBB uses 4 threads... and what I get is 27.28 Mbps with a burst speed of 30Mbps
http://www.thinkbroadband.com/speedtest/results.html?id=138402437698510383121
4x7Mbps = 28Mbps.
So it would appear something is restricting each thread to circa 7Mbps. Since its not the overall speed affected..
Im thinking more like the Ellacoyas? Perhaps it doesnt recognise the traffic type.. and is restricting it.. but Im supposed to be on unlimited which isnt meant to be shaped??
----
ETA
On 2nd thoughts as I was walking away, it just occurred to me that the Ellacoyas are supposed to be intelligent enough to not allow multi-threading to bypass any restrictions. Theres definitely some form of traffic shaping going on. The fact that 7.1Mbps is suspiciously similar to an IPprofile is making me point the finger at either BTs network if not Plusnets.
-
Im not so sure its CPU.
Something else is going on.
This is my base test - 72.7Mbps /17.19Mbps
http://www.speedtest.net/my-result/3090197134
Enabled Address based filtering and I got 14.82/0.42 Mbps
http://www.speedtest.net/my-result/3090201269
I download a file and I get a practically rocksteady 7.01Mbps on Netmeter.
hmmmmm
So speedtest.net uses parallel threads 2 x 7Mbps = 14Mbps
To test this theory.. TBB uses 4 threads... and what I get is 27.28 Mbps with a burst speed of 30Mbps
http://www.thinkbroadband.com/speedtest/results.html?id=138402437698510383121
4x7Mbps = 28Mbps.
So it would appear something is restricting each thread to circa 7Mbps. Since its not the overall speed affected..
Im thinking more like the Ellacoyas? Perhaps it doesnt recognise the traffic type.. and is restricting it.. but Im supposed to be on unlimited which isnt meant to be shaped??
Hi
That's interesting
I'll test some more :)
The thing is, I found out about it because of a steam download
my steam downloads at the max speed of 47Mbps, but was only downloading at 20Mbps so that's what made me look in to this
As soon as I set filtering to off, steam was full speed, set in on boom, steam was slow
I tested steam download servers in Manchester and London as it was still the same
Kitz can you try steam see what you get? ;)
-
btw.. I edited my above post to add something else.
Kitz can you try steam see what you get?
I dont use steam.. are you able to give me a url which I can test something with?
-
btw.. I edited my above post to add something else.
Kitz can you try steam see what you get?
I dont use steam.. are you able to give me a url which I can test something with?
Do you run a vpc so you down have to install steam on your main pc?
You could try it by installing steam on a vpc, then downloading a free steam game like War Thunder (Aircraft Sim Type Game)
http://store.steampowered.com
http://warthunder.com
I'm going to test P2P with filtering set to on as soon as I have a free moment :graduate:
Though i'm not sure if steam uses p2p
-
A visualware analysis also gives me 28Mbps, although my upload is 16.7Mbps.
I cant find out how many threads they use.
This is the results (http://www.connectionanalyzer.com/index.php?testtype=speed&speed_dspeed=28362096&speed_uspeed=16791336&speed_qos=84&speed_bandwidth=38219928&speed_avgpause=1&speed_routespeed=262140000&speed_forcedidle=0&speed_routeconc=1.3475707&speed_rtt=2&speed_maxpause=47) of my analysis report
The download speed is too low for the latency/bandwidth of the connection
This means that download throughput has been recorded lower than the attainable speed for the connection being tested and this degradation is not accounted for by the latency of the connection. This indicates that the connection has data flow interference through congestion or regulation problems.
A high TCP Max Delay and/or a low Data Flow QoS will also confirm any occurrence of data flow quality problems.
-
No sorry dont have a sandbox or vpc installed atm.. and Im outa here soon :/ :drink:
-----------
PS could do with someone on an IPStream connection testing using the TG582n
Zen also use this router.
Is HP still on Be using a Bebox that may also be useful.
-
My graph is also interesting.. looks like there's artificial shaping going on there.. each of the results mentions QoS problems.
-
My graph is also interesting.. looks like there's artificial shaping going on there.. each of the results mentions QoS problems.
Hi
Hope you had a good night :drink: :angel: :police: lol
I will try a few other download services like origin, p2p, and steam again. see what I get
Do you think it down to the router being in realty a ADSL family speed router?
EDIT:
If I use the usb port as network storage, will this slow the router down at all
-
Sorry for the tardy reply, not getting home from Manchester hospital till gone 11oclock last night and a bad nights sleep - read none - meant I wasnt up to much today. :(
Anyhow getting to the PC this evening and Im having what appear to be DNS problems in that pages wont load for ages, so I looked on the PN forums to see if there was any known issues... and I happened across this (http://community.plus.net/forum/index.php/topic,120071.new.html#new).
Following a link to Release R8.4.3 (http://www.technicolorbroadbandpartner.com/getfile.php?id=7446) Section 5, page 50 - Known Issues, we come across this:-
Services - Parental Control / Web Site Filtering
Using Parental Control, i.e. URL-based, or content-based filtering, negatively impacts the Technicolor Gateway
download performance.
I fear I may have been a bit distracted with the 7.01Mbps limitation thinking it was possibly a BT IPprofile. But that does give us a clue....
7.01 Mbps will also be the max TCP/IP throughput for adsl1.
-
Has anyone tested the 10.X firmware to see if the problem still occurs when using that? It doesn't say anything about parental controls in the 10.2.1 release notes.
-
I havent jelv. My nettage seems real screwy tonight.
Ive not really used the internet since I tested the settings on Saturday, and although Ive put everything back how it was, its being a right PITA and I cant get pages to load properly. I can get to google fine, but if I click on a link it just sits there waiting forever for the page to load, anything cached is ok. It feels like a DNS issue (which is why I visited the PN forums & saw the parental control post there), but Ive changed to googles DNS servers and Im still having the same problem.
Its totally doing my head in. >:(
The fact that I havent used my PC over the past few days, means I cant for certain point the blame at changing the parental settings being the cause... because I dont know when the DNS-like slowness exactly started.
---
Added
Damn i cant get stuff all on the ipad.. Ive 'broken' my internets :(
nor if I amend this post can I see any amendments.. I guess Im just seeing the cached page.
...and on that note Im off to watch TV or something, because I don't have the inclination tonight to mess any more to try and get my internet sorted, if its something that may resolve itself tomorrow.
Good night from a grumpy kitz.
-
Sorted - I also had to reboot my BTmodem & resync before everything went right again.
Whilst a reboot of the TG582n seemed to resolve the speed issues, for some reason the DNS type symptoms didnt clear until I did a full resync of the connection.
Use of address based filtering in the parental settings for certain cause 2 separate issues:-
- DNS like symptoms where web pages are very slow to load or wont load at all unless the page has already been cached.
- Speed throughput limited to 7.01 Mbps downstream & 0.5 Mbps upstream. This equates to throughput for adsl1 after allowing for tcp/ip overheads
A further observation is that opening multiple streams can increase throughput but also implying that something is also wrong with QoS.
Analysis of the test performed on 09/11/13 (http://forum.kitz.co.uk/index.php?topic=13224.msg249436#msg249436) indicated issues with QoS data flow. The graph provided here (http://forum.kitz.co.uk/index.php?topic=13224.msg249439#msg249439) shows strong signs of artificial shaping and QoS restrictions.
This is possibly related to the already known issue with QoS.
Quality of Service Queue-based rate limiting has no impact.
------------
It is highly likely that adsl2+ lines would also suffer the same symptoms, but not quite as obvious as when used on FTTC.
Im therefore moving this thread to the router hardware section as its a problem specific to the TG582n rather than just fibre.
-
I still have concerns over this router.
Something I happen to have noticed since carrying out the above tests and have put everything back 'as was', is that although I can achieve 70Mbps+ on speedtesters, this is because they use multi-threading.
If I try to use single threads then my speed is more like 24 Mbps.
You can see on here, what single thread speed is.
(http://www.thinkbroadband.com/speedtest/button/138462244756365893145-mini.png)
Ive tried downloading http files and ftp.. and unless Im using multi-threading then my speeds are vastly reduced. This is not right!
-
Is there anyone else on VDSL2 please that can do some testing on this and make sure its not just me thats seeing this, because if this is the case then this router is not fit for vdsl/vdsl2.
Tools that may be useful are
Net Meter (http://www.metal-machine.de/readerror/).
- Set Options to Graph -> Kilobit for UL/DL speeds.
JDs Speedtester (http://www.gmwsoftware.co.uk/).
- Make sure you turn off multi-threading in Menu -> Configuration -> Download -> Tick box at the bottom of window.
ThinkBroadband Test files (http://www.thinkbroadband.com/download/).
-
I'm not sure if this is helpful or may it just raise more questions
This is the TG582n on PlusNet with address based filtering
(http://www.thinkbroadband.com/speedtest/button/138442008996499670728-mini.png) (http://http:http://www.thinkbroadband.com/speedtest/results.html?id=138442008996499670728)
Using a Netgear DGND3700v2 as a router with same filtering
(http://www.thinkbroadband.com/speedtest/button/138428179164379481651-mini.png) (http://http:http://www.thinkbroadband.com/speedtest/results.html?id=138428179164379481651)
Results from speedtest.net are 20Mbps/0.7Mbps with TG582n and 48Mbps/15Mbps with Netgear
-
Oh my :no:
Thank you for that digbey. The address based filtering results are 'backwards' to what you'd expect ??? ???
It certainly does raise more questions about how the TG582n is handling each thread.
I think I shall give azakka a poke and see if Zen are seeing the same thing.
-
I'm happy to try out any other tests if you can suggest anything.
How about this other weirdness between the TG582n and the Netgear DGND3700v2 both with fairly light single user.
The Netgear produced similar results on BE ADSL2+
TG582n
(http://www.thinkbroadband.com/ping/share-thumb/c0c888225351238d6b76ff2b30620050-13-11-2013.png) (http://www.thinkbroadband.com/ping/share/c0c888225351238d6b76ff2b30620050-13-11-2013.html)
DGND3700v2
(http://www.thinkbroadband.com/ping/share-thumb/1b3a290c5bc564521d4e8b6b1a198736-17-11-2013.png) (http://www.thinkbroadband.com/ping/share/1b3a290c5bc564521d4e8b6b1a198736-17-11-2013.html)
-
Quick question: I take it these tests are being run via Ethernet and not via Wireless, also are all these tests being run via the same providers? or different providers?
Kitz, would you (or anyone else for that matter) be willing to use the Zen F/W for the 582N?
Whilst i think of it, the 582N's being used, do these have the Gig Ethe Ports or normal 100mb ports.
-
@ digbey wonder wth is causing that on the netgear :/ :-\
It would be interesting is to see what download speeds you get whilst downloading a file from here (http://www.thinkbroadband.com/download.html).
Youd need something like netmeter (http://www.metal-machine.de/readerror/) though to see the throughput.
No matter what I do I cant seem to get much more than 30Mbps (see attached)
-
@ azzaka
Yes Ethernet. The 582n has the normal 100mb ports and not GigE
-
Im currently doing some testing. This may not be the TG582n afterall.
I'll post full results when I have them all.
-
Using Ethernet here also
Using Netmeter
TG582n without address filtering
Down 48Mbps Up 1.75Mbps
TG582n with address filtering
Down 6Mbps Up 225Kbps
-
Thank you very much digbey.
I think its pretty certain that the address based filtering on the TG582n definitely limits a single connection to adsl1 speeds and multi-threading is needed to achieve anything more :/
This afternoon Ive been doing some further testing with regard to the limitation of circa 28Mbps that Im seeing on vdsl2.
The 30Mbps limitation would not appear to be a restriction of the TG582n nor the Huawei Modem.
Ive opened up a new thread here (http://forum.kitz.co.uk/index.php?topic=13279.0) to discuss this and Im still gathering results.
So far the 28-30Mb is looking like it may be Plusnet specific :/
-
Hey Digby,
I take it you are on PlusNet too? Kitz/Digby is the 582N locked down at all? If not would you be willing to test a different F/W for the Address based filtering?
-
Hi Azzaka
Yes I'm on PlusNet. I don't normally use the 582n and have no idea if it's locked down but I am happy to have a go to test your F/W if you give me some details.
-
ok guys do this test.
disable the filtering.
goto speedguide.net
on left goto broadband hover over it for menu.
goto broadband tools on the menu
click on sg tcp/ip analyzer on the menu
note the results.
enable the filtering.
repeat same test making sure not cached result.
I suspect the router is overiding tcp buffer sizes, disabling rfc1323 etc. on http traffic and as such throughput is restricted, this theory is working on the same principle that most people think cloud servies are throttled as they get low upload speeds but thats simply down to restricted tcp send buffers.
Bear in mind tho if using windows vista or newer with automatic tcp tuning speedguide.net will only show a buffer below 65k as autotuning only goes above that when downloading large files. To analyse it more will probably need to sniff the traffic.
Its very unlikely to be BTw shaping/throttling, could be plusnet as we know those guys still use the shaping devices but I think most likely its the router.
Incidently if I enable QoS on my router it maxes out at about 45mbit/sec probably due to cpu, and possibly also due to manipulation of tcp headers.
-
Before we get engulfed in the festive season, I wonder if there are any conclusions on this subject ?