Congestion returning with a vengeance?

Forums › One NZ (including Farmside) › Congestion returning with a vengeance?

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | ... | 34

12188 posts

Uber Geek
+1 received by user: 8480

Trusted

Lifetime subscriber

#1796985 9-Jun-2017 00:39

Kodiack: Not congestion, but the internet is dead here tonight, both for us and for the neighbours (who also have FibreX).

Middle of the night maybe planned work?

Linux

MikeB4

MikeB4
18776 posts

Uber Geek
+1 received by user: 12767

ID Verified

Trusted

Subscriber

#1797064 9-Jun-2017 09:05

During last evening I gave a read books and discussed world affairs. My download and upload speeds were both in the 0.* range, so hopeless.

Here is a crazy notion, lets give peace a chance.

Sideface

9651 posts

Uber Geek
+1 received by user: 15602

Trusted

Lifetime subscriber

#1797363 9-Jun-2017 15:53

Background:

I am in Island Bay, Wellington, on "FibreX" 200/20Mbps cable, on node WKH3.
I typically use 600GB of data per month.
95% of this traffic is from overseas.
As a TrueNet volunteer, my connection has multiple tests performed on it hourly, 24/7, by an automated "probe" - a modified MikroTik router.

I have three sets of statistics for every day of "FibreX"s existence (since November 2016), all tested via ethernet under ideal conditions:

1. TrueNet automated stats recorded hourly
2. Manual OOKLA testing on multiple servers (local, including VF, & overseas) twice daily - off-peak and on-peak.
3. Records of actual download speeds, mostly international

Facts:

The "fastest" result is always the OOKLA speed test - typical peak-hour (8pm to 10pm) speed is 195/20Mbps on NZ, Australian (East coast), and US (West coast) servers.
The "slowest" result is always real download speed - typically 25-50 Mbps - maximum 80Mbps - nowhere near 200Mbps.

Alternative Facts:

The TrueNet results lie somewhere between OOKLA and "real-world" speeds:

ToD stats for Snap Christchurch:

ToD stats for Vodafone Wellington:

ToD stats for CityLink Auckland:

ToD stats for Telstra Sydney:

(click to view)

Note that all of these results are for my own connection only.

All of these servers showed almost full "200/20" speeds on OOKLA.

We have recently been assured in this blog that ISPs do not give priority to speed testing sites.

But there is a wide gap between "good" speed tests and real-world performance, both within and beyond New Zealand.

Sideface

kharris

1209 posts

Uber Geek
+1 received by user: 327

ID Verified

Trusted

#1797366 9-Jun-2017 16:03

@Sideface am I reading this correctly? Is Sydney test significantly better than Auckland or Wellington?

Still well below 200Mbps though.

Kirk

Talkiet

4819 posts

Uber Geek
+1 received by user: 3934

Trusted

#1797380 9-Jun-2017 16:14

I've lost count of the number of times I have suggested people go and have a look at the documented methodology that speedtest.net use for their tests. When people finally do go and read (and understand) it, they usually understand why the speedtest results are so much better than so called "real world" results.

Cheers - N

Please note all comments are from my own brain and don't necessarily represent the position or opinions of my employer, previous employers, colleagues, friends or pets.

Sideface

9651 posts

Uber Geek
+1 received by user: 15602

Trusted

Lifetime subscriber

#1797382 9-Jun-2017 16:14

kharris:

@Sideface am I reading this correctly? Is Sydney test significantly better than Auckland or Wellington?

Still well below 200Mbps though.

The Sydney TrueNet results have consistently been "better" than Auckland (ACSData, Spark, and sometimes Vodafone).

Since this Wellington-to-Australia traffic (at least until recently) went via Auckland, this does not make sense.

There has been minimal change in test results since the new Trans-Tasman cable opened.

Traffic to Christchurch tends to be "faster" than to Auckland.

EDIT:

For the sake of completeness, here are my ToD results for CityLink, Auckland:

(click to view)

Ugly

Sideface

AliExpress

Shop now on AliExpress (affiliate link).

Kodiack

712 posts

Ultimate Geek
+1 received by user: 351

Trusted

#1797387 9-Jun-2017 16:22

Linux:
Kodiack: Not congestion, but the internet is dead here tonight, both for us and for the neighbours (who also have FibreX).

Middle of the night maybe planned work?

Linux

It's possible. I remember when they did the D-CMTS cutover and we were without internet for a couple of hours. Things were much better when we got back, so I certainly won't complain if this is the case. :P

gaddman

224 posts

Master Geek
+1 received by user: 98

Trusted

#1797389 9-Jun-2017 16:32

Sideface:

The "fastest" result is always the OOKLA speed test - typical peak-hour (8pm to 10pm) speed is 195/20Mbps on NZ, Australian (East coast), and US (West coast) servers.
The "slowest" result is always real download speed - typically 25-50 Mbps - maximum 80Mbps - nowhere near 200Mbps.

You've got some great data there, and seeing 195Mbps on Ookla is good news. But can you elaborate a little more - when you say 'real' download speeds, can you give some specific examples?

This is the crux of the speed issue here - it absolutely depends what and how you measure. When you buy a 200Mbps plan from an ISP, you're buying a speed from the edge of your home to the edge of the ISP network. What happens outside that is largely out of the ISP's control. To give a few key examples:

wifi is the biggest influence - don't expect 200Mbps speeds from your 10 year old iPad.
the endpoint you're testing to - even if you've got a 1Gbps connection, I wouldn't count on your local Steam server allocating that much just to you
protocol, distance and filesize - TCP takes a while to ramp up, longer when higher latency is at play. Test with a 1MB file from Christchurch to Sydney and you'll get a different figure than with a 2MB file.

For comparison, Ookla tests try to swamp your connection as much as possible - using decent sized files and multiple parallel streams.

Sideface

9651 posts

Uber Geek
+1 received by user: 15602

Trusted

Lifetime subscriber

#1797390 9-Jun-2017 16:35

Talkiet:

I've lost count of the number of times I have suggested people go and have a look at the documented methodology that speedtest.net use for their tests. When people finally do go and read (and understand) it, they usually understand why the speedtest results are so much better than so called "real world" results.

Cheers - N

Perhaps I should have listed the OOKLA results under "Alternative Facts"?

Sideface

Talkiet

4819 posts

Uber Geek
+1 received by user: 3934

Trusted

#1797392 9-Jun-2017 16:41

Sideface:

Talkiet:

I've lost count of the number of times I have suggested people go and have a look at the documented methodology that speedtest.net use for their tests. When people finally do go and read (and understand) it, they usually understand why the speedtest results are so much better than so called "real world" results.

Cheers - N

Perhaps I should have listed the OOKLA results under "Alternative Facts"?

I would :-)

Please note all comments are from my own brain and don't necessarily represent the position or opinions of my employer, previous employers, colleagues, friends or pets.

Sideface

9651 posts

Uber Geek
+1 received by user: 15602

Trusted

Lifetime subscriber

#1797396 9-Jun-2017 16:49

gaddman:

Sideface:

The "fastest" result is always the OOKLA speed test - typical peak-hour (8pm to 10pm) speed is 195/20Mbps on NZ, Australian (East coast), and US (West coast) servers.
The "slowest" result is always real download speed - typically 25-50 Mbps - maximum 80Mbps - nowhere near 200Mbps.

You've got some great data there, and seeing 195Mbps on Ookla is good news. But can you elaborate a little more - when you say 'real' download speeds, can you give some specific examples?

This is the crux of the speed issue here - it absolutely depends what and how you measure.<snip>

Agreed.

The "real world" speeds that I have quoted are from "fast" overseas servers (Los Angeles or London or Amsterdam), downloading multiple large (>500MB file size) files using a download manager (JDownloader) via gigabit ethernet and a Ubiquiti Lite router.

I got faster speeds using the same servers, download manager, and hardware when I was on the old DOCSIS3 100/10Mbps cable.

"FibreX" has proved to be a downgrade.

I won't have a "real fibre" option until late 2019.

Sideface

Geekzone

Want to support Geekzone and browse the site without the ads? Subscribe to Geekzone now (monthly, annual and lifetime options).

kharris

1209 posts

Uber Geek
+1 received by user: 327

ID Verified

Trusted

#1797434 9-Jun-2017 18:17

Talkiet:
I've lost count of the number of times I have suggested people go and have a look at the documented methodology that speedtest.net use for their tests. When people finally do go and read (and understand) it, they usually understand why the speedtest results are so much better than so called "real world" results.

Cheers - N

So are you saying that Spark's connection real world results will also be significantly advertised and Speedtest results.

Kirk

Talkiet

4819 posts

Uber Geek
+1 received by user: 3934

Trusted

#1797441 9-Jun-2017 18:34

kharris:
Talkiet:

I've lost count of the number of times I have suggested people go and have a look at the documented methodology that speedtest.net use for their tests. When people finally do go and read (and understand) it, they usually understand why the speedtest results are so much better than so called "real world" results.

Cheers - N

So are you saying that Spark's connection real world results will also be significantly advertised and Speedtest results.

I'm sorry... I can't decipher your question/statement.

Please note all comments are from my own brain and don't necessarily represent the position or opinions of my employer, previous employers, colleagues, friends or pets.

kharris

1209 posts

Uber Geek
+1 received by user: 327

ID Verified

Trusted

#1797446 9-Jun-2017 19:03

Talkiet:
kharris:
Talkiet:

I've lost count of the number of times I have suggested people go and have a look at the documented methodology that speedtest.net use for their tests. When people finally do go and read (and understand) it, they usually understand why the speedtest results are so much better than so called "real world" results.

Cheers - N

So are you saying that Spark's connection real world results will also be significantly advertised and Speedtest results.

I'm sorry... I can't decipher your question/statement.

N

Not sure what happened there.
So are you saying that Spark's real world results will also be significantly slower than advertised (and Speedtest) speeds?

Kirk

Talkiet

4819 posts

Uber Geek
+1 received by user: 3934

Trusted

#1797449 9-Jun-2017 19:17

kharris:
Talkiet:

kharris:
Talkiet:

I've lost count of the number of times I have suggested people go and have a look at the documented methodology that speedtest.net use for their tests. When people finally do go and read (and understand) it, they usually understand why the speedtest results are so much better than so called "real world" results.

Cheers - N

So are you saying that Spark's connection real world results will also be significantly advertised and Speedtest results.

I'm sorry... I can't decipher your question/statement.

N

Not sure what happened there.
So are you saying that Spark's real world results will also be significantly slower than advertised (and Speedtest) speeds?

Totally depends what you mean by "real world results". With the staggeringly good quality of BB in NZ for most subscribers, there are MANY MANY cases where the resulting real world speed is going to be dictated by physics and characteristics of TCP rather than the quality of the ISPs network.

As an example, I have 2 Fibre Max lines here and on either I can get 940Mbps+ on Speedtest to our Chch Speedtest server, 920ish to Wellington, and 890ish to Auckland. I can also download a single thread file from a Chch server at a "real world" (Size of file/time taken to download) speed of 112Megabytes/sec (896mbit).

However, on the same lines I might get only a few 10s of Mbit from a server in Tajikistan because of TCP, latency, other ISP networks and the originating server.

And for other sources in between, the speeds will vary. I know if I download something off the local Netflix or Google servers (or parts of Akamai) then I get amazing speeds... Parts of Australia give great speeds too (several hundred Mbps) but other tests or downloads from parts of Aussie will be a lot worse. There are speedtest servers in the US that give me results of several hundred Mbps, and some that give well under 100. There are download sites that give several hundred mbps if I can configure a multithread download, and some that are clearly artificially throttling the download. SO what is "real world"?

Loading web pages is not a good measure of "real world" speed either, because it relies on your machine, CPU, RAM, browser, number of threads, possibly 3rd party ad servers, requires dozens (sometimes hundreds) of elements to be downloaded, might involves lots of DNS lookups - and that's even before we try and decipher if the web page is a static page or generated on the fly so might involve remote database lookups etc.

At least with the very high speed access available now in NZ it's usually not possible to accurately state an expected "real world" result without adding too many qualifiers for it to be a palatable sound bite - and anyone telling you otherwise is in marketing and/or doesn't understand. (Or have their own agenda to push and aren't interested in accuracy)

So what is a "real world" speed? I don't know. I don't think anyone does, and what qualifies for one person might seem incredibly contrived or artificial for someone else.

Cheers - N

Please note all comments are from my own brain and don't necessarily represent the position or opinions of my employer, previous employers, colleagues, friends or pets.

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | ... | 34