Print

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

[Alex Atkin UK]

It would make sense for Zen to be able to use CityFibre backhaul with an Openreach FTTP service, although this would be the first I've heard of it.

That does make me wonder why I'm not on it though given CityFibre are well under way the north of my exchange.

[bogof]

@bogoff has already confirmed that there is more than 1G connection from AWS to Zen speedtest server. I want to see the results from AWS to @bogoff, I happy to pay the 25p 

I don't have any useful way to do an apples-to-apples comparison as I can't set up an Ookla server on AWS without it costing me an arm and a leg, and Zen don't have any public iperf3 server.
I can run iperf3 to AWS from home.  I get about 600Mbps in a multithreaded test.  Single threads can be up to 250Mbps.

That's about the same as I get to Zen's Ookla server ID 40788 (https://www.speedtest.net/result/c/9354578d-ea29-45ae-a8a7-c0d1f2e5a6df). 
A local ISP has an Ookla server on just a 1G link (Voicehost - ID 9060).  I get almost line rate to that meagre box... https://www.speedtest.net/result/c/05ea7e76-bb18-4857-9730-4c6524f12e3d

Yet from AWS to Zen's server I can get 9200Mbps. https://www.speedtest.net/result/c/75aaeabd-72fe-404d-ba49-5e5155bf8dd7
And AWS to the Voicehost server as expected tops out at 940Mbps (if no one else is using it). https://www.speedtest.net/result/c/daf365f6-cde8-4318-a483-dafd45565044

The above tests were all done within a short while of each other, and done a few times to check consistency.

So you can see the issue here; I can get traffic at speed from here to some servers, and even to some ones that are not particularly well specified I approach line rate.  Yet to Zen's Speedtest server, and my Amazon server, performance is merely passable.  I don't know how anyone can look at such results and not come to the conclusion that there is something odd and unexpected going on in Zen's network for me, at least.  It appears there is a bottleneck or behaviour which means traffic going certain places is not travelling as fast as you might hope.  This behaviour is independent of peak times, but can and does vary with gateways and PPPoE clients.
(edit:fixed some broken links)

[craigski]

Forget Ookla/speedtest for the moment, you have proven its unreliable/inconsistent on your connection.

Did you try my 2 minute experiment above? Make sure you turn off any additional services on your firewall.

[Chrysalis]

EDIT- Zen came back to me again confirming I am on BTW..

I was interested to find out if I was a gea or BTW for my recent new service.. zen report I am neither as I am City fibre apparently which I was not aware of... have they got this right as my upgrade was direct with zen...somewhat confused.

I thought there was no CF FTTP in your area?  You on Openreach FTTP?

[skyeci]

Hi Chris

Yes, no CF networks here so I think they were looking at the wrong info. I am on BTW. Have done some speedtest cli tests.
Currently running at 913Mbps down
110 Mbps up
latency 7ms

Cheers

[bogof]

Noticed a couple of disconnections early morning today, but hadn't heard anything from Zen for a week or so.  They'd asked Openreach to swap out a bunch of SFPs in the exchange here in Norwich as apparently they were from a defective batch.  In any case, they didn't seem to have any effect at all on the performance here, so it has gone back to the Zen network team.

[Alex Atkin UK]

Just wanted to confirm that despite single-threaded speedtest.net results off Zen server still maxing out at 300Mbit I was able to download in Firefox at 100MB/s earlier which would rather confirm everything is fine under real-world scenarios.

[bogof]

Not much more to report.  Apparently something was actioned over the weekend, no indication what as yet, but it kicked off my PPPoE session, and when I picked up another one it had a fair bit of background ping packet loss visible at the top of BQM.  I dropped that session last night and picked up another on a different gateway, which fixed the packet loss, and had marginally higher throughput, but still not right.

On the upside Zen have assigned a dedicated case manager now to it instead of it just getting picked up from the pool, which may help at least with the continuity there.   

[bogof]

Small update; Zen are going to be sending out a Samknows speed testing device to install on the network.  It will be interesting to see how it behaves.  I'm not sure how the results will influence the process - we already know the line is capable of line rate to some servers and services, but not others.  So it might report full speed, or not.  But it will be an interesting exercise nonetheless.

For what it is worth, the web-based Samknows speed test seems to consistently test well below line rate here (https://speedtest.samknows.com/)

[Ixel]

For curiosity I thought I'd see what the Samknows speed test did here, different ISP though of course.

Code: [Select]

Latency Jitter Download Upload
4.90 ms 0.90 ms 946 Mbps 939 Mbps

It will be interesting to see what result you get with the Samknows device. Hopefully it will result in a similar slow speed that the Samknows speed test website does for you so Zen can't potentially try to say something like "well the Samknows device say it's working as expected so that's that".

[bogof]

Hopefully it will result in a similar slow speed that the Samknows speed test website does for you so Zen can't potentially try to say something like "well the Samknows device say it's working as expected so that's that".

Thanks for running the test, always useful to have other data points.  I am hoping for a result to be as you say as it does give the clearest path forward. 

To give them the best chance of success though I've moved back to the Zen supplied router doing PPPoE, into which I will plug the Samknows box.  The home Unifi UDM Pro SE network for now is plugged into the Zen router over DHCP (so double NAT, which annoyingly has broken my site to site VPN at the moment, and also no WAN ping) . 

Offloading PPPoE duties to the Zen router hasn't improved matters from the UDM network.  I will drag the laptop downstairs later and check the Samknows test site directly from the Zen router, which hopefully will give some idea of what to expect.

[craigski]

For what it is worth, the web-based Samknows speed test seems to consistently test well below line rate here (https://speedtest.samknows.com/)

Another day, another speedtest. I looked, this one starts 8 downloads from your browser, to same server IP 80.71.21.229, measures the total amount of data downloaded over 12 seconds, and ESTIMATES the speed. I know I keep banging on about this, but take any speedtest website with a pinch of salt, its just an estimate guide of your speed. *IF* they could accurately measure the speed they would all give the same result.

A far more accurate way to confirm what you speed is is to look at the interface on your firewall router when you are loading the network, using multiple devices on your LAN. I honestly believe you are chasing your tail trying to get the results you want from a speed test website with a fast FTTP connection.

I would also suspect that speedtest traffic from websites is given a low priority on the various networks. It makes sense ISPs would not want speedtest traffic run by a very few people to impact the many majority that are streaming music/video and making VoIP calls.

I think @Alex as sort of confirmed this:

Just wanted to confirm that despite single-threaded speedtest.net results off Zen server still maxing out at 300Mbit I was able to download in Firefox at 100MB/s earlier which would rather confirm everything is fine under real-world scenarios.

I have personally found some gateways on Zen perform better than others, find one that works for you, and stick with it. If PPPoE drops, redial until you get back on the one that performs best for you.

[bogof]

Another day, another speedtest. I looked, this one starts 8 downloads from your browser, to same server IP 80.71.21.229, measures the total amount of data downloaded over 12 seconds, and ESTIMATES the speed. I know I keep banging on about this, but take any speedtest website with a pinch of salt, its just an estimate guide of your speed. *IF* they could accurately measure the speed they would all give the same result.

A far more accurate way to confirm what you speed is is to look at the interface on your firewall router when you are loading the network, using multiple devices on your LAN. I honestly believe you are chasing your tail trying to get the results you want from a speed test website with a fast FTTP connection.

I was only interested to see what the Samknows web based test looked like as potential indication of how the servers they use for the Samknows box might behave on my connection.  Most of my testing is done using native binaries under Linux and not browser based (you seem to be the one referring back to browser stuff all the time).

I'm interested to see what the Samknows box ends up reporting.  You can see here a report on how Samknows helped Zen achieve VCOP compliance with their monitoring of FTTP packages, which will give you some indication I hope of what rates should be achievable and were achieved across a broad cross section of users (see the chart at the top of the article): https://samknows.com/blog/zen-vcop.  From reading around the Samknows box methodology it seems like they have a pretty comprehensive test methodology.

And just to be perfectly clear as I can't help but feel you keep missing this point (or are perhaps just deliberately being obtuse) - the connection prior to GEA migration was testing via any server I cared to choose at the line rate or very close; it's not like I've just got a FTTP service and am just having unreasonable expectations, or that I don't know how to have a network set up so that I could test at maximum rate, yet your messages continuously imply that this is likely an issue of PEBKAC... A change clearly has happened and since then the service benchmarks notably much worse.  Zen seem interested enough to send a box out...  And if you follow the big thread over at TBB, you'll see there is precedent for performance taking a tumble after migration.

I would also suspect that speedtest traffic from websites is given a low priority on the various networks. It makes sense ISPs would not want speedtest traffic run by a very few people to impact the many majority that are streaming music/video and making VoIP calls.

Traffic management of speed test traffic is an interesting one; not least because the Zen server is one of the worst behaving on Zen's own network; according to traceroute the traffic never leaves the Zen network (or they all at least have Zen IP range 51.148.0.0/16 and Zen reverse DNS), and Zen make a point of saying they do not do any traffic shaping on their network.  So I'm sorry but I can't square that as being a cause.  And remember, I know their server is capable as it tests at 10G from a 3rd party network, and I know their server previously tested at line rate any time of the day pretty much for me.  It's not possible to look at that result with a logical mind and come to any other conclusion than in recent time something is getting in the way of throughput to that Zen server; and when that is a new situation it raises a valid question of what that is.
See here for Zen's stated stance on traffic management:
https://www.zen.co.uk/blog/posts/zen-blog/2018/06/04/zen---the-network-for-gamers

Traffic management

Many ISPs advertise a certain feature, such as a particular speed or download limit, but tend to provide varying degrees of traffic management.

For example, an ISP might slow down your connection speed at certain times of the day, or when you’ve downloaded a certain amount of data in a month.

As well as throttling speeds, an ISP might put restrictions on certain types of traffic passing through their network.

At Zen, you won’t be victim to this type of management – we believe in providing you with what you’ve paid for. This means that, if you’re downloading lots of games, or you like to play even during the busiest periods of the day, you shouldn’t experience any degradation in performance.

[Alex Atkin UK]

I honestly believe you are chasing your tail trying to get the results you want from a speed test website with a fast FTTP connection.

I would also suspect that speedtest traffic from websites is given a low priority on the various networks. It makes sense ISPs would not want speedtest traffic run by a very few people to impact the many majority that are streaming music/video and making VoIP calls.

I think @Alex as sort of confirmed this:

I have personally found some gateways on Zen perform better than others, find one that works for you, and stick with it. If PPPoE drops, redial until you get back on the one that performs best for you.

This is true, but on the flip side I get full speed on SamKnows.  So if I can get full speed and someone else on the same ISP even during off peak hours does not, I'd call that an issue.  The particular background behind the change being all the more concerning.

This is a case of Zen backhaul performing worse than BTW, all parties want that fixed as it would cost Zen a fortune to revert everyone back, plus we have no idea if the problem might get worse as more users come online at the exchange.

[bogof]

This is true, but on the flip side I get full speed on SamKnows.  So if I can get full speed and someone else on the same ISP even during off peak hours does not, I'd call that an issue.  The particular background behind the change being all the more concerning.

This is a case of Zen backhaul performing worse than BTW, all parties want that fixed as it would cost Zen a fortune to revert everyone back, plus we have no idea if the problem might get worse as more users come online at the exchange.

I did drag the gigabit LAN laptop downstairs yesterday and briefly check - samknows webchecker appears to be middle of the ground via the Zen supplied router, testing at between 700-750Mbps (while at the same time Zen's test server comes in around 600mbps and Voicehost still close to 900Mbps, when pre-migration nearly everywhere was >900Mbps).  So I'll be interested to see what the Samknows dedicated testbox clocks in at.  If the range holds when moving to the dedicated box then that range would put this connection below the average of the worst 20% performing 900mbps FTTP lines Zen had in their test pool it would seem.  Though I note from the chart in that article that they have rounded all their numbers to the nearest 25Mbps - as 950Mbps for FTTP is a little bit of a stretch...