NZ Broadband testing from Samknows

Forums › New Zealand Broadband › NZ Broadband testing from Samknows

1 | ... | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17

1011 posts

Uber Geek

Trusted

# 2106471 11-Oct-2018 19:47

hio77: ... Sure they appear to be having a slight teething issue.

They've also been doing this long enough in multiple locations. From their About page: 'A long time ago, at a computer not far away, our hero Sam Crawford was hard at work ...' The MLab site linked above states 'Founded in 2008, SamKnows works alongside ...'

10 years later, I would have hoped they had ironed out a lot of the wrinkles in their go-live process.

Please keep this GZ community vibrant by contributing in a constructive & respectful manner.

SiliconAudio

96 posts

Master Geek

# 2106513 11-Oct-2018 21:12

Talkiet:

The data collection is in a test phase at the moment. There will be no RSP comparison reports generated from current data and I suspect they are still making sure the servers are dimensioned appropriately.

Throwing around comments like I see here is irresponsible, especially at this stage.

All of the issues mentioned here have been hashed out between industry experts and Samknows/Comcom. I think it's fair to say that the technical RSP representatives were very impressed with the Samknows expertise and methodology - both for data collection and for the analysis constraints. It's in a different league to the previous Comcom testing.

If you're still seeing really inconsistent results after they launch properly (after you see their PUBLIC reports) then it's reasonable to ask questions about why the data seems inconsistent - until then you don't know if the data is going to be used, or even if it's using final test servers or traffic paths.

Cheers - N

OK, that completely changes the situation. I had no idea they were in a testing phase. Problem here is that this hasn't been communicated very well at all - I thought the system was live. From an end-user point of view, the situation is they sent me a box, gave me access to the portal and told me I was all set up. No info from any official channel that this was not live. All I've seen is unbelievably inaccurate results, which concerned me greatly. I had absolutely no reason to believe that the data should not have been accurate.

I have an open case with SamKnows - it will be very interesting to see if they make mention of the fact that they are still tuning. The first level engineer made absolutely no mention of them being in testing.

BarTender

2852 posts

Uber Geek

Trusted

Lifetime subscriber

# 2106547 11-Oct-2018 23:36

One person supports this post

The other interesting wrinkle I suspect will be is the number of folks on GB plans and if the Whitebox can really keep up with that.

We know it's hitting the MLab servers at REANNZ. I assume they are 10GB connected. The weathermap doesn't indicate if they are 10GB and what frequency the graphs take their stats... The year graph is pretty interesting for Auckland though:

Auckland: https://weathermap.reannz.co.nz/node.php?src=and01&int=xe_0_0_9

Wellington: https://weathermap.reannz.co.nz/node.php?src=and02&int=xe_0_2_7

dfnt

defiant
1004 posts

Uber Geek

Lifetime subscriber

# 2106559 11-Oct-2018 23:46

I was reading the archive thread @hio77 posted earlier, apparently their nodes at WIX and CHIX were only 1gbit and they are going to/have upgraded to 10gbit because of the Samknows testing

hio77

'That VDSL Cat'
11043 posts

Uber Geek

Trusted

Spark

Subscriber

# 2106561 11-Oct-2018 23:49

dfnt:

I was reading the archive thread @hio77 posted earlier, apparently their nodes at WIX and CHIX were only 1gbit and they are going to/have upgraded to 10gbit because of the Samknows testing

based on the graphs i'd suspect that work has not been finalized yet, or mlabs still has a 1G link to them.

#include <std_disclaimer>

Any comments made are personal opinion and do not reflect directly on the position my current or past employers may have.

SiliconAudio

96 posts

Master Geek

# 2106564 11-Oct-2018 23:53

hio77:

dfnt:

I was reading the archive thread @hio77 posted earlier, apparently their nodes at WIX and CHIX were only 1gbit and they are going to/have upgraded to 10gbit because of the Samknows testing

based on the graphs i'd suspect that work has not been finalized yet, or mlabs still has a 1G link to them.

According to this post: https://list.waikato.ac.nz/pipermail/nznog/2018-September/022888.html

"The nodes each consist of 4 x Dell servers front ended by a Juniper QFX switch. They're connected at 10Gbps"

So, they are already 10 Gbps.

hio77

'That VDSL Cat'
11043 posts

Uber Geek

Trusted

Spark

Subscriber

# 2106569 11-Oct-2018 23:56

SiliconAudio:

hio77:

dfnt:

I was reading the archive thread @hio77 posted earlier, apparently their nodes at WIX and CHIX were only 1gbit and they are going to/have upgraded to 10gbit because of the Samknows testing

based on the graphs i'd suspect that work has not been finalized yet, or mlabs still has a 1G link to them.

According to this post: https://list.waikato.ac.nz/pipermail/nznog/2018-September/022888.html

"The nodes each consist of 4 x Dell servers front ended by a Juniper QFX switch. They're connected at 10Gbps"

So, they are already 10 Gbps.

I'm aware of what was said, as i said 'based on the graphs'.

#include <std_disclaimer>

Any comments made are personal opinion and do not reflect directly on the position my current or past employers may have.

SiliconAudio

96 posts

Master Geek

# 2106573 12-Oct-2018 00:12

In case anyone missed it, the Technical FAQ PDF linked in that post is an excellent read.

A whole bunch of info in there that also covers off the story behind the NZ-based servers. Wish I'd found this earlier :) Tell me it wasn't only dumb me who missed this.

SiliconAudio

96 posts

Master Geek

# 2114590 26-Oct-2018 17:58

Still seeing inaccurate results here:

SamKnows - most sample points off peak:

Speedtest results taken between 5:20 and 5:45 PM on a Friday (peek usage time)

NZ:

Sydney:

London:

San Francisco:

My query to SamKnows support was initially escalated a couple of weeks ago, but have heard nothing since. Is this a SamKnows issue or an M-Lab problem? I don't know, but I do know that the data being collected on my white-box is about as useful as an ashtray on a motorbike.

Talkiet

4201 posts

Uber Geek

Trusted

# 2114593 26-Oct-2018 18:03

One person supports this post

And to re-iterate, there will be no final reports generated from this data... Reporting results you think are incorrect or out of step with reality is a good thing, but don't get too worried yet.

Cheers - N

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

SiliconAudio

96 posts

Master Geek

# 2114614 26-Oct-2018 18:28

Talkiet:

And to re-iterate, there will be no final reports generated from this data... Reporting results you think are incorrect or out of step with reality is a good thing, but don't get too worried yet.

Cheers - N

According to the Commerce Commission technical whitepaper:

Our initial findings report is planned to be released in late October. The first full quarterly report will likely be released early in 2019.

I guess this means that the initial report is due very shortly, and that the first quarterly report will be based on data collected the last two or three months of this year. Given we are nearly in November, if they plan to stick to their schedule, the data we are collecting today will be in the first quarterly report, won't it?

If not, how do we actually know when the "real" testing starts? Will there be some official communication from SamKnows or the Comcom? I just want to know that when the "real" data is collected, there is some kind of assurance that the data is of reasonable quality. It seems to me that there is no official process for disseminating this information - have I missed something?

Talkiet

4201 posts

Uber Geek

Trusted

# 2114673 26-Oct-2018 20:01

In reality what you have missed is that you're seeing the data before it's processed fully - you're also seeing a single datapoint. The whole reason for having thousands of volunteers is so that there's a significant number of datapoints for each set of variables.

I sat in the room and asked the Comcom rep and Sam explicitly about their plans for maintaining decent statistical validity when they started to break down the numbers and they gave very reasonable answers - at least for me (and I have done postgrad stats at University)...

The initial report won't have any RSP comparisons so with fewer variables the impact of some unexpected results will be less important.

Also remember that volunteers are only seeing the data as a courtesy, it's dangerous to try and infer performance from single lines. If there are any widespread issues with the data, I guarantee it will be pointed out by the RSPs (if it's not pointed out by Samknows themselves or the Comcom). I fought for the raw data to be made available for 3rd party analysis and validation so I think the chance of it being arbitrarily modified without someone noticing is very low.

There's no obligation to tell the volunteers when the 'real' testing will start.

Cheers - N

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