Kitz ADSL Broadband Information
adsl spacer  
Support this site
Home Broadband ISPs Tech Routers Wiki Forum
 
     
   Compare ISP   Rate your ISP
 
Please login or register.

Login with username, password and session length
Advanced search  

News:

Pages: 1 [2]

Author Topic: Ongoing issues with Single Thread Speeds  (Read 1716 times)

jaydub

  • Reg Member
  • ***
  • Posts: 213
Re: Ongoing issues with Single Thread Speeds
« Reply #15 on: January 06, 2018, 01:31:51 PM »

Hi

Just to add I've not had maximum throughput via Uno (which use TalkTalk) for many, many months on the single thread speed test, regardless of the time day or night, I usually get around 60-64Mbps on the single thread test, with the multi-thread test always ~74Mbps. I know when you first posted about this and I replied I was getting maximum on both types of test, but at some point that changed, although nothing my end has changed.

However playing about with TCP Optimizer https://www.speedguide.net/downloads.php I can improve the single thread speeds back up to almost match the the multi-thread speed, so I don't think it is something being deliberately throttled by TalkTalk, just some interaction between Windows (10 in my case) automatic network settings just not hitting an optimal match against TalkTalk and the site it's connecting to.  This may explain why some speed tests are okay and others are not.

Edit:  if I test here https://testmy.net/results and use the single thread speed test option I get consistently > 73Mbps, it's only ThinkBroadband's single threaded part of the test that is impacted, which I think was your original issue?  I think the problem is with Thinkbroadband and how the test is routed or network parameters and there isn't really any problem at all with the speeds, just the test.
 

Regards

Phil

Hi Phil,

Interesting that you are seeing a reduction in single thread speeds with Uno.  There's another IDNet customer on TBB who has been moved from a BTW backhaul with evening congestion to a TTB backhaul and seeing something akin to what I am seeing.  He's had his fill of IDNet and has just placed a migration order with Uno, so may give him a heads up of what you are seeing at Uno.

I've not done any TestMy.Net testing but did run their auto tests for a couple of days after joining IDNet.  As you can see from the attached graphs, the speeds were generally better, but there were some definite blips as well.  I might just try running a few manual tests back to back to see how consistent they are one after another.

Out of interest, have you tried running TestMy.Net in multithread mode to just the London server.  It just won't run smoothly for me.

Uno used to have a single thread tester, but they seem to have taken that off line.  I found that a useful reference point in the past, so it's a shame it has gone.

I'm running off a Mac, so speedguide.net tweaking is not an option for me.  I have validated my TBB results with my son's Win10 laptop, so fairly certain it's not the Mac at fault here.

Thanks,

Jon
Logged

jaydub

  • Reg Member
  • ***
  • Posts: 213
Re: Ongoing issues with Single Thread Speeds
« Reply #16 on: January 06, 2018, 01:40:58 PM »

Didn't AA have a similar experience with TT at some point? I can't remember.

Would AA suit you? They are very keen on performance monitoring.

They did.  At the beginning of last year IIRC.  Pulse8 had something similar when I was with them, which affected certain gateways and was unrelated to the AA issue.  That was fixed after about a month of investigation.

I'm not mad keen on going to AA for a couple of reasons:
1)  I've just paid for 12 months up front with IDNet and not willing to write that money off, unless I can negotiate a refund.
2)  I've still got an adult child at home and the other one threatening to come back at some stage, so not sure I really want to go on a limited bandwidth product offering.

They do remain an option for the future though.
Logged

Chrysalis

  • Content Team
  • Addicted Kitizen
  • *
  • Posts: 5371
Re: Ongoing issues with Single Thread Speeds
« Reply #17 on: January 06, 2018, 03:04:15 PM »

For me the small niche providers arent what they used to be and I now consider risky, they dont have the benefit of large swarms of customers able to absorb heavy users like the large isps do and as such I actually now thing visible contention is more likely on smaller provider, AAISP I have as an exception tho as their managing director has always made a point of AAISP never been the bottleneck.
Logged
Sky Fiber Pro - Billion 8800NL bridge & PFSense BOX running PFSense 2.4 - ECI Cab - LINE STATISTICS CLICK HERE

jaydub

  • Reg Member
  • ***
  • Posts: 213
Re: Ongoing issues with Single Thread Speeds
« Reply #18 on: January 06, 2018, 03:14:25 PM »

I might just try running a few manual tests back to back to see how consistent they are one after another.

Done.  Results attached.  I then ran 5 TBB tests and got single thread results of: 71, 50.2, 67.5, 68.8 & 67.7 Mbps.  Better than I have been seeing off TBB, but not as good or consistent as TestMy.Net.
Logged

PhilipD

  • Reg Member
  • ***
  • Posts: 349
Re: Ongoing issues with Single Thread Speeds
« Reply #19 on: January 07, 2018, 10:57:01 AM »

Hi

The question is, are we just testing the tester now?

There is something going on with TalkTalk I think and as Uno get handed off to Daisy this I would think points more towards their back haul. 

This is a test just now, not too bad for the single thread test but it the speed used to always been on parity with the multi-threaded speed, so it isn't' constantly around 10% less that I've often seen recently.  So probably congestion rather than a deliberate throttling, it's hard to complain as it is sold as a contended service, and it never gets any worse than a 10% reduction, and of course if a second person in the household is doing something online or we are streaming and downloading, we will be maxing out the line no problem. 

I wonder if this is some QOS setting on their network and is by design to reserve bandwidth for TalkTalks TV service?



Regards

Phil


Logged

bilbokitz

  • Reg Member
  • ***
  • Posts: 195
Re: Ongoing issues with Single Thread Speeds
« Reply #20 on: January 18, 2018, 11:35:38 PM »


Would the single thread be the limiting factor for an openvpn client connecting to a vpn provider from within my network?
« Last Edit: January 18, 2018, 11:38:16 PM by bilbokitz »
Logged
Alcatel Cab 100/10 Vectored VDSL - Technicolor DGA4130 Bridged + Asus RT-AC86U Asuswrt-Merlin

jelv

  • Helpful
  • Reg Member
  • *
  • Posts: 847
Re: Ongoing issues with Single Thread Speeds
« Reply #21 on: January 19, 2018, 08:44:45 AM »

You are questioning why you are not getting the speed you were getting from Pulse8. The question not asked (and not possible to answer without trying it) is what speed would you be getting now with Pulse8 if you were still with them? Would it have also degraded?
Logged
Programming today is a race between software engineers striving to build bigger and better idiot-proof programs, and the Universe trying to produce bigger and better idiots. So far, the Universe is winning. Rick Cook, The Wizardry Compiled

Weaver

  • Addicted Kitizen
  • *****
  • Posts: 5146
  • Retd sw dev; A&A; 3 7km ADSL2; IPv6; Firebrick
Re: Ongoing issues with Single Thread Speeds
« Reply #22 on: January 19, 2018, 09:50:38 AM »

Btw, this has been said before. AA publish their own congestion figures. I don't know if anyone else does. There used to be a detailed table of packets dropped - don't know where that went. Perhaps it is because they are using some Cisco core switches and routers now instead of all Firebrick kit, until the forthcoming high speed Bricks become available.
« Last Edit: January 19, 2018, 09:54:16 AM by Weaver »
Logged

jaydub

  • Reg Member
  • ***
  • Posts: 213
Re: Ongoing issues with Single Thread Speeds
« Reply #23 on: January 19, 2018, 10:26:21 AM »

You are questioning why you are not getting the speed you were getting from Pulse8. The question not asked (and not possible to answer without trying it) is what speed would you be getting now with Pulse8 if you were still with them? Would it have also degraded?

That in itself is an interesting question, but one that is likely to remain unanswered.

I guess the more fundamental question is whether there was actually any value in changing, especially as I am now paying more per month than with Pulse8.

By way of an update, this Tuesday IDNet transferred my backhaul from TTB to BTW, and although the results aren't absolutely amazing they have resulted in a significant improvement (see attached IDNet scatter plot).

The first couple of days were pretty good, but am seeing some degradation in single thread speeds since yesterday (see attached IDNet BTW scatter plot).

In terms of a straight forward figures analysis, the answer at the moment is that IDNet on a BTW backhaul are still doing better than Pulse8 did in my last few weeks with them.


          IDNet (BTW)      IDNet (TTB)      Aquiss       Pulse8
Mean        69.4                   59.6              60.8          64.2
Median     71.1                   63.3              69.5          66.9
Std Dev     5.5                   10.0              17.5            9.4

What I'm really struggling to understand is the variation in TBB single thread speeds on consecutive tests.  I always do them in packets of 3 tests and the most recent gave me 72.5, 59.7 & 42.3 Mbps.

Does anybody else see that level of variation?  Any guesses why back to back tests should be so different?

Logged

Ixel

  • Reg Member
  • ***
  • Posts: 956
Re: Ongoing issues with Single Thread Speeds
« Reply #24 on: January 19, 2018, 10:49:01 AM »

Btw, this has been said before. AA publish their own congestion figures. I don't know if anyone else does. There used to be a detailed table of packets dropped - don't know where that went. Perhaps it is because they are using some Cisco core switches and routers now instead of all Firebrick kit, until the forthcoming high speed Bricks become available.

I don't think that table shows up unless there's actually a problem. I'm pretty sure I recall seeing such a table at some point in the last few months.

That in itself is an interesting question, but one that is likely to remain unanswered.

I guess the more fundamental question is whether there was actually any value in changing, especially as I am now paying more per month than with Pulse8.

By way of an update, this Tuesday IDNet transferred my backhaul from TTB to BTW, and although the results aren't absolutely amazing they have resulted in a significant improvement (see attached IDNet scatter plot).

The first couple of days were pretty good, but am seeing some degradation in single thread speeds since yesterday (see attached IDNet BTW scatter plot).

In terms of a straight forward figures analysis, the answer at the moment is that IDNet on a BTW backhaul are still doing better than Pulse8 did in my last few weeks with them.


          IDNet (BTW)      IDNet (TTB)      Aquiss       Pulse8
Mean        69.4                   59.6              60.8          64.2
Median     71.1                   63.3              69.5          66.9
Std Dev     5.5                   10.0              17.5            9.4

What I'm really struggling to understand is the variation in TBB single thread speeds on consecutive tests.  I always do them in packets of 3 tests and the most recent gave me 72.5, 59.7 & 42.3 Mbps.

Does anybody else see that level of variation?  Any guesses why back to back tests should be so different?



That level of variation does seem a bit large compared to the variation I see on my connection. I'm seeing a variation of no larger than 10 megabits typically (download speed maxing out at around 60 megabits).

This is a recent test I did (via the TTB backhaul):


Bear in mind I'm using QoS (fq_codel) on my EdgeRouter.

There's some occasions on other tests where it can be a larger variation though, but when I monitor my own network it's usually when another device is also using the internet for that moment. My ping would normally be much lower but DLM kicked in recently and I don't know what exactly happened but I guess downstream MTBE exceeded the threshold. I've noticed I'm getting large bursts of FEC errors on my DrayTek again, so I'm guessing my original fault is returning (or has become a bit more noticeable now?). I might put the HG612 or Zyxel back on, capped though, to mainly check if the Hlog and QLN have changed... but that's another story.
« Last Edit: January 19, 2018, 11:58:30 AM by Ixel »
Logged

jaydub

  • Reg Member
  • ***
  • Posts: 213
Re: Ongoing issues with Single Thread Speeds
« Reply #25 on: January 19, 2018, 01:03:49 PM »

I don't think that table shows up unless there's actually a problem. I'm pretty sure I recall seeing such a table at some point in the last few months.

That level of variation does seem a bit large compared to the variation I see on my connection. I'm seeing a variation of no larger than 10 megabits typically (download speed maxing out at around 60 megabits).

This is a recent test I did (via the TTB backhaul):


Bear in mind I'm using QoS (fq_codel) on my EdgeRouter.

There's some occasions on other tests where it can be a larger variation though, but when I monitor my own network it's usually when another device is also using the internet for that moment. My ping would normally be much lower but DLM kicked in recently and I don't know what exactly happened but I guess downstream MTBE exceeded the threshold. I've noticed I'm getting large bursts of FEC errors on my DrayTek again, so I'm guessing my original fault is returning (or has become a bit more noticeable now?). I might put the HG612 or Zyxel back on, capped though, to mainly check if the Hlog and QLN have changed... but that's another story.

The variation is getting worse and the poorer results slower and more frequent.  The last couple of tests have been 70.4, 31.7 & 17.5 Mbps and then a couple of hours later 70.7, 28 & 34.6.  Time to speak to IDNet support again.
Logged
Pages: 1 [2]
 

anything