I got 99 problems but my WiFi ain't one

Too many variables; version of iperf, commands issued (packet size etc), which direction did you test (client/server), network topology etc will all affect the numbers it spits out.

If anyone’s in the market for a 10Gb ethernet switch/router (if you have a NAS with 10Gb ports or similar):

I’ve been playing around with the new L3 hardware offload feature of the Mikrotik CRS317 and it’s shit hot. I have L3 routing between VLANs working at 10Gbit/s line-rate, with over 2 gigabytes/second aggregate speed to my FreeNAS box using LACP bonding two 10Gb interfaces and copying files from several clients.

To have 16 SFP+ ports and L3 routing at 10Gb line rate from a £300 box is bonkers.

https://mikrotik.com/product/crs317_1g_16s_rm

Got an email from zen at the end of my contract offering to put me on their faster broadband offering for about £5 less than I’m paying currently.
Spoke to their tech support as I’d gotten the slower offering (37mbps) as they’d mentioned I might not be able to get the full speed of the faster one (77mbps). Tech support confirmed this, saying at most I’d get a bump to 49mbps, but wouldn’t know until I upgraded.
Spoke to their retention team (I didn’t have any intention of changing provider, as I’d get the same speed with all of the ones working off the open reach infrastructure) and they confirmed what tech support had mentioned, and knocked a tenner off my monthly payment and gave me a new 12mth contract.

Pretty happy I chose zen, with the honesty they have about their product/s and the actual real world speeds you will get from them.

Cheers. Any suggestions on the best way to test network speed then?

Cheers. Any suggestions on the best way to test network speed then?

Test for what?

Sorry if I missed it upthread but didn't see what the problem is that you are trying to diagnose.

Are you experiencing slow network performance on a particular device?

The variables are too many to get any accurate picture.

When I've changed things and cared to get an indicative understanding of what a specific device has in terms of speed... then I use https://fast.com/ https://speed.cloudflare.com/ https://www.google.com/search?q=speed+test (Google's search engine provides a speed test built in) and https://www.speedtest.net/

I will run each test on each provider a few times, repeat at different times in the day, and then use that to understand the likely average speed.

The problems:

• With Wi-Fi there is way too much local and potentially temporary interference to get a real measure
  
• Depending on the device you're testing, the device may have a max throughput, i.e. my phone outperforms the Raspberry Pi, and both are beaten by my Macbook, which in turn is trounced by my ThinkPad... all on the same network in roughly the same spot
  
• If you're doing local testing the server device may have limitations
  
• Anything else happening on the network may interfere with the test
  
  

Basically I don't bother trying to get an accurate number... and I understand my experience differs based on what service I am using the network for. So I only try to get an idea of what the network may achieve, and where they may be issues (typically when I should cable something instead of WiFi).

My internet speed right now (Windows PC, tests run in latest stable Firefox, on 1Gbps ethernet connected to Virgin Cable 1Gbps package):

The variance is huge... but from that I can say roughly 500Mbps down and 7Mbps up with a latency around 15ms.

After having played about with my network settings I was testing my network speed (using iperf 3) on some computers connected via wi-fi (to a machine connected by ethernet) and getting speeds of 80-90Mbps.

This, from Aggi's original post, threw me off a bit.

Is he testing point to point on network or testing internet speeds?

I think he's testing Wi-Fi locally.

But I'm not sure what that will tell him beyond Wi-Fi is shit, and the variance is huge.

If @aggi is moving large files locally... cabled ethernet or nothing (let the storage speed be the limiting factor)... and if they are just internetting, then the internet speed and service being used likely limits the max speed more than the WiFi does.

Ordered the Unifi Dream Machine...

@Velocio and @Stonehedge

My internet speed is fine even on wireless devices, I just ran a speed test on my 200Mbps connection and got this for instance:

I'm trying to work out how fast the local network speed is on my wi-fi devices to see whether it's worth using ethernet. I thought iperf3 was the right tool to test that but running it gave speeds less than half of my internet speed which doesn't seem right so I'm wondering the best way to test that local network speed.

Edit: Pondering is mainly for stuff like remote desktop, streaming Stream, etc to a Shield, etc

Oh... I just do a far easier thing for that... I find a nice huge file, a 20GB blu-ray will do it... and I transfer it first one way, then the other.

I use the calculated time after a few minutes of transfer to tell me whether it's worth cabling.

For the Nvidia Shield... if you want to watch 4k things... you really need to cable from storage to the Shield.

Also... just discovered the Ofcom speed checked, which reports almost 1Gbps for me... that's such a difference from the others that I re-ran the others... and now I'm seeing between 850Mbps and 1Gbps across all of them.

:meh

I suppose that is the obvious that should have occurred to me. Too busy trying to think up a clever solution.

Shield, storage and plex server are all wired (I'm very thankful that the people who rewired my house before I bought it included ethernet into each room). It's just my desktop gaming PC that isn't and possibly should be.

Oh... I just do a far easier thing for that... I find a nice huge file, a 20GB blu-ray will do it... and I transfer it first one way, then the other.

This reminded me of a time I tried this method in the late 90s.

I tried transferring a 500Mb file between two ethernet cabled computers (about 10m apart) using FTP.

It was only ~450Mb and obvs corrupted at the destination when completed. Never did quite get my head around how that happened. I assume something, somewhere, maxed out and there was no error checking :D

I'm (finally) setting up VLANs (using a nice intelligible guide from here ).

I'm doing it in stages, so that I don't do what I always do, and break the network, requiring starting from scratch again.

LAN1 is still intact, as has pretty much everything attached to it.

I'm gradually transferring my stuff onto a VLAN 10 for trusted machines, and everything else onto a Guest VLAN 20

The problem I have, though, is that the Guest VLAN assigns IPs in the 169.254.x.x range, rather than the 192.168.20.x that the router should be assigning.

Could this be a firewall thing?

Using Edgemax router, EdgeSwitch, Unifi controller

169.254 is the autoconfig network used when there is no response from a DHCP server. It sounds like your router doesn't have an interface on VLAN 20 or the DHCP server isn't configured on the router or VLAN 20 isn't trunk/tagged to and from your switch and router.

iperf is the correct tool to test network performance, but the results depend on the variables you supply.

Which commands did you run on the server & client, and how are those machines connected to your network (wireless in particular; channel width/link rate/hardware etc)?

VLAN 20 isn't trunk/tagged to and from your switch and router.

Thanks - I though (hoped) that this might be it!

I had some ports excluded for the guest VLAN on one switch, and I didn't even have VLAN 20 on the other switches

They are all tagged now, but still no IP being assigned.

The only difference in dhcp for the two VLANs is the domain name, and guest has a shorted lease time.

On the switches, they're now all the same, and on the unifi controller they're also the same, other than rate limiting on the guest VLAN, and they use a wifi.

I'm wondering if I need to add port profiles in the unifi controller and tag the guest network?

Or if it's something fucky with the firewall rules or dnsmasq

Also, it turns out the MS Teams doesn't work if you use a VLAN. I have no idea how to troubleshoot that.

So the Team thing is because DNS wasn't working on the VLANs

I did get a guest VLAN working by attaching it to the eth1 interface, but I cannot get one working on switch0

I honestly don't know if this is a problem or not

I've just discovered iPerf .

25Mbps on my network.

Lovely.

[Edit] Just noticed the iperf recommendation up there ^

Server is just

iperf3 -s

client is

iperf3 -c 192.168.xxx -t 60

Connects to ubiquiti gear via 5GHz.

Interestingly I ran it again just now and the results were more what I expected (~ 200MBps). Not entirely sure what has changed but I guess that's one of the issues with wi-fi.

You can also try:

iperf3 -c 192.168.xxx -t 60 -P 4

...which will run the test using 4 parallel streams, more accurately simulating a multi-stream web download.

Adding -R reverses the client-server traffic flow; always good to test both ways.

Cheers, very useful.

iperf seems to be one of those things where everything you come across online assumes you already know how to use it.

Some of you guys who know me IRL will know that I've been a life long radio nerd. When I was a kid I used to build and sell pirate radio transmitters to DJs and studios in London before going legit and getting my radio amateur license in the mid 90s (long since lapsed). Now days, I spend my spare time contributing to various IoT and amateur satellite projects but my real passion is radio communications. Anyway, setting the scene but I digress.

I noticed a long time ago that every single wifi device I have ever dissected was fitted with terrible quality antennas. Most wifi antennas as supplied by manufacturers are adequate at best, utlising crude coax sleathed dipoles (aka bazooka antennas) or very basic PCB based dipoles. The problem with these antennas is that while designed to work across both wifi bands, they are quite narrow band in design, slightly directional and overall very innefficient which means they don't do very well on either 2.4Ghz or 5Ghz.

If people have wifi signal problems at home they tend to look for a cheap solution and often go to Amazon or eBay where there are plenty of aftermarket antennas for sale for between £5 and £20 but unfortunately these are invariably worse than stock antennas. At best they are missold (common to see old 3G antennas labelled as Wifi) and at worse they are so badly made they barely work at all (commonly using extremely cheap high loss cable).

I did a few experiments recently and made a few dual band wifi antennas and was blown away by how much better they work than the usual shit that gets bundled with routers and PCIe cards. Funnily enough, good quality components and proper tuning doesn't seem to happen in the factory at Netgear, Huawei etc.

If anybody is having any signal problems and have devices that use external antennas, I'd be glad to knock one up for the cost of parts and P&P to see if it helps avoid buying a mesh setup when it could just be the quality of the antenna letting you down. Would expect that costs will be lower than £10 including P&P.