For help, advice and discussion about stuff not related to aviation. Play nice: no religion, no politics and no axe grinding please.
  • 1
  • 9
  • 10
  • 11
  • 12
  • 13
User avatar
By stevelup
FLYER Club Member  FLYER Club Member
#1888452
Forget about diagnosing anything to do with the USG until you've got Starlink working right.

I think you need to open a support ticket with them. They're very responsive and helpful.

Right now, it's not working properly.
Colonel Panic liked this
By Colonel Panic
#1888656
Having turned off UTM, IDS, IDP, DPI and GeoIP filtering within the USG, I have just run a series of 12 tests across 4 different network configurations in quick succession.

It looks, to me at least, that the USG is no longer a bottleneck - which is great news. Thank You for prompting me to narrow down the road blocks. Speeds are more variable than ideal, but still much better than PlusNet via FTTC.

One other thought - might my RPi / Pi-Hole slow the network down? It is plugged in to the Switch, so would only have effected Test 1 and Test 4 (so probably not).

Image
User avatar
By stevelup
FLYER Club Member  FLYER Club Member
#1888674
PiHole can't affect performance.

I do wonder if your starlink unit is on old firmware. Those speeds are what I got for the first 48 hours or so after I took delivery. They then shot up to what they are now (and stayed there) after a firmware update.

Easiest way to check the firmware version is to plug the starlink router back in and then go to 192.168.100.1 (or open the app). You can check for obstructions etc., there as well.

There's something not right still.

You can see the firmware versions on the link below and see if you're up-to-date:-

https://docs.google.com/spreadsheets/d/ ... =383580049
User avatar
By rikur_
FLYER Club Member  FLYER Club Member
#1888675
Some other straws to grasp at......

How are you connecting to the MBP - presumably an ethernet to USB adapter? What speed is it reliably capable of? (albeit that shouldn't affect test 4)

What speed test site/location are you using. I find once you get over 100mbps, you need to choose more carefully, as otherwise peering bottlenecks start to feature.

might my RPi / Pi-Hole slow the network down
unlikely - but you can get a few odd effects by running your own DNS server. Some CDNs and larger sites use ISP DNS servers to route traffic to servers connected directly to the ISP's network. By running your own DNS servers you may find that you are not connecting to the optimal CDN/web server for your ISP. (But this is unlikely to affect your speedtest)
User avatar
By stevelup
FLYER Club Member  FLYER Club Member
#1888677
I forgot to say earlier (well I composed the message but obviously didn't hit submit)...

I suggest you use speed.cloudflare.com instead of speedtest.net.

You'll always hit the same server, and it does a more thorough test (it does multiple transfers of gradually increasing file size).

(Cloudflare URL edited)
Last edited by stevelup on Fri Dec 17, 2021 3:52 pm, edited 1 time in total.
User avatar
By rikur_
FLYER Club Member  FLYER Club Member
#1888679
stevelup wrote:I forgot to say earlier (well I composed the message but obviously didn't hit submit)...

I suggest you use speed.cloudflare.net instead of speedtest.net.

You'll always hit the same server, and it does a more thorough test (it does multiple transfers of gradually increasing file size).

Pedant mode....

speed.cloudflare.com

It doesn't always hit the same server. It always hits a server of the same name / IP address, but uses BGP to identify one located closest to you in one of over 200 different data centres on the internet. (Essentially it better automates the server selection process than Ookla)
User avatar
By stevelup
FLYER Club Member  FLYER Club Member
#1888684
Oops, sorry.

I didn't mean literally the same server, but it does hit servers maintained by them and on their network as opposed to some random jobbie hosted in someones garage (I'm exaggerating, I know, but you know what I mean).
User avatar
By rikur_
FLYER Club Member  FLYER Club Member
#1888686
stevelup wrote:Oops, sorry.

I didn't mean literally the same server, but it does hit servers maintained by them and on their network as opposed to some random jobbie hosted in someones garage (I'm exaggerating, I know, but you know what I mean).
:thumright:
I've never understood Ookla's logic - but it seems to be roughly geographic, so picks some random rural ISP that happens to be physically close to me, but many network hops away, rather than picking their server in my own ISP's data centre which has a shorter path length and performs better. Whereas Cloudflare is basically letting the network decide which is the closest server as it would with real traffic (so even if it gets the answer "wrong", it's probably the same wrong that will affect real traffic).
By Colonel Panic
#1888719
Easiest way to check the firmware version is to plug the starlink router back in and then go to 192.168.100.1 (or open the app). You can check for obstructions etc., there as well. https://docs.google.com/spreadsheets/d/ ... =383580049

The router is on 2021.48.0.mr7520-prod which is only 12 days old, but the dish is on
0a38f6f7-00f1-4187-bc68-8ee12545830.uterm.release
and I can’t find that in the google docs sheet.

How are you connecting to the MBP - presumably an ethernet to USB adapter? What speed is it reliably capable of? (albeit that shouldn't affect test 4)

I am using a 2018 AKALAA USB C Hub, Aluminium USB C Dock, USB Type-C MacBook Pro Adapter, 8 in 1 Thunderbolt 3 Hub, USB C to HDMI, 3 USB 3.0 Ports and SD/TF Card Reader - bought from Amazon but no longer listed.

What speed test site/location are you using

I had been using https://www.speedtest.net , but I have now done some tests through https://speed.cloudflare.com - just the 25MB download results shown here.

Image
User avatar
By stevelup
FLYER Club Member  FLYER Club Member
#1888725
I've just had a proper look (was on my phone earlier)

There isn't a -single- hit on the whole internet for your dish firmware version.

That's just too bizarre to accept. I really think you should give CS a shout - say you're not seeing the performance you expected and see what they say.
  • 1
  • 9
  • 10
  • 11
  • 12
  • 13