Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[les-70]

 I am with TalkTalk Business and took a passing interest in the Plusnet congestion/throughput posts.   I had tended not to use the Think Broadband speed test much but decided to do so.  A typical result is below

   I have tried changing TCP settings but they made little difference.  The results are almost always the same with perhaps a bit of drop some evenings when for some reason only around 10pm is there some sign of congestion.

   I am running Windows XP so i don't think TCP settings are very automatic but maybe the difference between signal and multi thread is down to me?

   I have a further puzzle which is that when the line was interleaved the single thread speed was equally consistent but lower. 

    The other puzzle if I look back to old speed tests up until a couple of months ago the single thread was similar to the multi thread.

    I had a chat with TTB support but they did not seem to grasp the issue as they seem to just use there own speed test which I suspect in multi threaded.

    Could anyone offer some help on what is happening?  Do people on TalkTalk home see any issues?

[burakkucat]

I am on a legacy TT residential package and, just for fun, decided to try the flash-based TBB Speed test --



I guess that looks o.k. . . .

[Dray]

can't see it (unless I use quote)

[kitz]

The reason I like the http1 v http6 test from TBB is that it shows up issues often not seen in other speedtests and yes youre correct thats what is used to moan at Plusnet when they had the 'gateway issues' whatever they were.

It usually implies some form of congestion rather than rate limiting, QoS or Traffic Shaping.
However it can also infer an RWIN problem, but if it was RWIN then from what Ive seen both on mine and others is usually around the 30Mbps mark and it doesnt explain why all was ok a couple of months ago unless youve reinstalled the OS or a different PC.

It wont show on b*cats because his line rate isn't high enough - same as with on PN it was only those with the higher speeds that noticed it and adsl2+ was unaffected.

I'd try testing at different times of the day - weekends and evening arent always the best time as thats when congestion is usually most active.   It may also be worthwhile grabbing a copy of Netmeter and seeing what speed it records when downloading some of the test files from TBB.

[Chrysalis]

The easy to know if its RWIN is to do an off peak test. Preferably between 2am and 7am.

RWIN issues would be the same regardless of time of the day.

[Dray]

Unless Windows is auto-tuning the RWIN.
See the command output for:- netsh interface tcp show global

[les-70]

   I have tried at all times of day.  The result are almost all exactly the same and very consistently the same.  The only exception is that on odd evenings about 9-10 pm things are bit worse and ping times a bit erratic. That I guess is congestion.

   The operating system is a consistent XP but I will boot up Linux later to today.  Changing PPOE sessions makes no difference but as noted interleaving did make a consistent difference all the time it was present. It really puzzles me that interleaving made such a consistent difference, I don't have any theory as to why that would be.   Nothing with RWIN has helped, small values make it a bit worse.  Downloading test files from TBB shows the single thread result is real as well. 

  I am at a loss and hence the call to TTB support which does not seem to have produced more than a ticket.  I would guess that with TalkTalk I am subject to parts of Openreach as well some dedicated TT infrastructure.

[burakkucat]

I would guess that with TalkTalk I am subject to parts of Openreach as well some dedicated TT infrastructure.

I would agree. However for your circuit Openreach control a minority of it, whilst TalkTalk control the majority. The only part of your circuit which is under the control of Openreach is that from the DSLAM to the head-end. At the head-end there would be the appropriate (optical fibre) cable-link, "handing over" your circuit (as part of a multiplex) to a TalkTalk MSAN.

Does that make sense? 

[Chrysalis]

Unless Windows is auto-tuning the RWIN.
See the command output for:- netsh interface tcp show global

why would that affect time of day?

[les-70]

  I would also guess most of the network involved is under TT control.  i can't see congestion being involved as things are virtually always the same.

    I remain perplexed over the impact interleaving had.  I have capped my line speed so the sync speed has been totally constant all the time.  When the upstream interleaving was remove the single thread speed jumped from a stable ~15 to a stable ~19 and jumped again to a stable~27 when downstream interleaving was removed. That is nearly  a factor of 2 with no change in the line sync speed of 60.  I can't understand why interleaving would or should impact sustained download speeds when the sync speed is constant and feel that the fact that it has done so may be indicating something?

[burakkucat]

. . . feel that the fact that it has done so may be indicating something?

Again, I agree. But I cannot put my "paw" on exactly what those results indicate. 

[les-70]

  I recall that a calculation of the ideal RWIN can be made using the expected delay and bandwidth.  That suggests something that could change with interleaving due to the varying delay times  -- however none of my changes to RWIN over a wide range of value made any significant difference.  The fact that changing RWIN made no difference may also also of course be indicating something else is wrong.