Low single thread speed

[Bill]

My code has appeared for some work tonight... and, depending what's involved, work on other codes could have an effect on mine anyway.

A couple of speed tests this morning lead to some (very) cautious optimism... 

[jaydub]

I too am feeling somewhat happier this morning having migrated from TTB Backhaul to BTW.  It's eraly day's yet, but my TBB speed tests have averaged 70.5Mbps this morning.  Still a bit down on the multithread average of 74.1, but much better than I was seeing on the TTB backhaul, which has averaged 59.6 since I migrated in a month ago.

There is still some interesting variations in some results, but probably not worth worrying about.

[jaydub]

Just one final question as it went unanswered before.

If you do three consecutive TBB speed tests, how consistent are the single thread speeds?

Although life is much better on a BTW backhaul, I still see a significant variation between the fastest and slowest results of the three the vast majority of the time.

[Gary]

Just one final question as it went unanswered before.

If you do three consecutive TBB speed tests, how consistent are the single thread speeds?

Although life is much better on a BTW backhaul, I still see a significant variation between the fastest and slowest results of the three the vast majority of the time.

Depending on your router, the load on the speed test server, congestion, contention, crosstalk and all the various other radio waves that cause background interference at various times of day, plus the laws of physics I think there will always be some variation. As long as single thread is pretty close to the multi thread score you are doing the best you can, considering this is a post war copper (with some aluminium) network with a bit of fibre stuck on in places that was never designed to do any of this in the first place

[Bill]

I still see a significant variation between the fastest and slowest results of the three the vast majority of the time.

Depends what you mean by "significant"... congestion being what it is, I'd expect to see variations of anything up to ~5Mbps, maybe more, between them. And on the odd occasions I run that sort of test, I often do.

I'm coming at it from a different direction- I've been running speed tests at intervals during the day and putting them into a scatter plot to get a "congestion profile" over the day... the 1x tests are virtually always a few Mbps below the 6x ones, don't mind that, but there's a marked shortage of bandwidth somewhere from about 8pm to 11pm... 1x speed falls by about half. It's not causing me any problems at the moment, but doesn't bode well for the future.

Trouble is, data collection is a very slow process 

[jaydub]

Depends what you mean by "significant"... congestion being what it is, I'd expect to see variations of anything up to ~5Mbps, maybe more, between them. And on the odd occasions I run that sort of test, I often do.

I'm coming at it from a different direction- I've been running speed tests at intervals during the day and putting them into a scatter plot to get a "congestion profile" over the day... the 1x tests are virtually always a few Mbps below the 6x ones, don't mind that, but there's a marked shortage of bandwidth somewhere from about 8pm to 11pm... 1x speed falls by about half. It's not causing me any problems at the moment, but doesn't bode well for the future.

Trouble is, data collection is a very slow process 

I too am doing a scatter plot with my speed test data.  I'm probably doing the same as you in terms of testing during the day, it's just that I do three consecutive tests at a time rather than just a single test.

I haven't seen any time of day impact with my connection on either TTB or BTW backhauls and wouldn't have thought it wouldn't be an issue for the Zen backhaul either.

Here's a plot of my IDNet connections.  Quite marked how much better the BTW backhaul has been on my connection.



The mean is over 10Mbps higher and the median about 8Mbps higher and there is significantly less scatter in the results (the SD has more than halved).

Still a few outlying results, but I guess that is to be expected.

[Bill]

You're running a lot more speed tests than I am 

Certainly it's an improvement, but tbh it's not a lot better than I'm seeing now, and I'd still call that a congested connection... but tastes vary 

I use Dropbox which doesn't allow embedded links, but I've put my results in a PDF you can see- https://www.dropbox.com/s/zr9x6pjdw69yq8x/Congestion.pdf?dl=0

(Just kill the "Sign Up" box if you get one)

It's plotted differently to yours, and the worst of the ~2pm and evening peaks seem to have gone since the work Zen did on Monday night, but too early to be sure.

I haven't seen any time of day impact with my connection on either TTB or BTW backhauls and wouldn't have thought it wouldn't be an issue for the Zen backhaul either.

Well something is... IDNet claim no contention on their network so it must be on Zen's. Either Zen are congested at my exchange or IDNet haven't bought enough backhaul bandwidth (or both!), no way of telling from here. But it's not yet bad enough to get excited about afaic.

[jaydub]

Ive split out my results to just display the results following the move to the BTW backhaul.



It looks as though I have a lot more tests that are close to the TBBx6 result, with a smallish proportion of outliers.

At the end of the day, we are not guaranteed a congestion free connection, so I am trying to apply a certain level of pragmatism that there will always be a few results that are out of kilter with the majority.

Remember congestion can be at the cabinet level as well as the exchange.  Zen are certainly able to monitor traffic levels al the way up to the GEA Link in the exchange.

Below is a paraphrase of an explanation I got from Jon Green at Zen, when I was thinking of joining them.

On Zen LLU, we get control of your connection as a PPPoE session at your parent exchange, and then route it through our backhaul network to one of our central PoPs.  Our Network Ops team monitor and manage the capacity on the various parts of our network.

Conversely, a BTWholesale connection would be backhauled from the exchange through BT's network, and then we'd get the connection in one of our central PoPs, so we'd have no direct control of the bit between your exchange and our network.

I assume the latter statement is equally true for IDNet.

I would certainly like to see better, but I don't think I have enough of an issue to push Brian for a further change, particularly since the only other backhaul option is Zen and I'm not sure moving to Zen at the moment would actually improve things.

[Bill]

At the end of the day, we are not guaranteed a congestion free connection, so I am trying to apply a certain level of pragmatism that there will always be a few results that are out of kilter with the majority.

I agree completely, but pragmatism can change with circumstances, and if I hadn't bought a 4K smart TV just before Christmas it's unlikely that I'd taking much notice at all of single-thread speeds, and certainly not plotting them! Multiple thread speed seldom drops more than a meg or so from it's usual 67, and on a decent-sized download it'll normally max out the line.

Even for UHD/HDR streaming it works fine most of the time, but there's not much room for it to get any worse...especially during the evening.

Remember congestion can be at the cabinet level as well as the exchange.

That's why I've been on Zen backhaul for the last few years 

I had a problem that was eventually traced to congestion on my cabinet, BT showed no inclination to fix it so Simon swapped me to a Zen SVLAN. But the problem showed as severe packet loss, which isn't the case this time. Probably doesn't prove it's not the cabinet again, but it's all I've got to go on.

[Bill]

the worst of the ~2pm and evening peaks seem to have gone

Oh no they haven't... 

[andrue]

I'm pretty certain that TBB's testers are hosted in their data centre in London and Zen just make use of them.et

Yes. TBB is owned/operated by NetConnex.

About Us.

"Our staff work for thinkbroadband along side full time jobs at hosting company NetConnex (ncuk.com) providing business cloud servers, co-location and VPS/dedicated servers within the Telehouse Docklands datacentre, including hosting the thinkbroadband infrastructure and speed test platform. "

[andrue]

I had a couple of poor single thread results tonight (I'm on BTW backhaul) but it seems to have recovered now. It's noticeable on my TBBQM though that there's increased latency during peak hours last couple of days. I'm keeping an eye on it.

[Bill]

It's noticeable on my TBBQM though that there's increased latency during peak hours last couple of days.

That's something I'm seeing no sign of whatever...

[Adrian]

I am seeing that too.

[jaydub]

I had a couple of poor single thread results tonight (I'm on BTW backhaul) but it seems to have recovered now. It's noticeable on my TBBQM though that there's increased latency during peak hours last couple of days. I'm keeping an eye on it.

I am also on BTW and had a set of poorer than normal results this afternoon.  I've been out this evening, so not sure whether it has improved or not.

I had a bad day for latency on my migration day to BTW (Tuesday) with the BQM graph showing the maximum ping (yellow responses) to be bad from 2pm to about 8:30pm, but it's been OK since.

[Gary]

Since the vast amount of Zen work being done things are much better, still have more next week including one that is just for my postcode, but I am pretty content with how things are at this time.

[jaydub]

Looking good, Gary.

My connection was very poor yesterday afternoon with one single thread result as low as 13.6Mbps, before jumping back up to normal speeds.

I suspect congestion on BTW equipment at the exchange or on the backhaul, but need to collect more stats to prove the point.

[Bill]

No change here, but my area code is on the list for more work towards the end of next week, so there's hope yet