What should be the maximum throughput of the wired network card?
I saw that for a Raspberry Pi 3+ that has a Gigabit interface, it’s just 300Mbps,
but since I’m using a Raspberry Pi CM4 that is more powerful, and the network card is on the Yellow’s board, it should be a better value, I just don’t know how much.
It’s to measure my Internet speed. I have a 1Gbit/s fiber connection, and I have SpeedTest testing my network each 15 min, but it tops a little more than 700 Mbit/s, so I would like if that’s correct or maybe there is something wrong with my link. The HA Yellow is connected to my Wi-Fi router and the Wi-Fi router is connected to the ISP router (that also has Wi-Fi, but it’s an awful equipment otherwise). Since I can’t test the speed test from my Wi-Fi router, I would like to test from HA Yellow. I hope I was able to make myself clearer now.
Nothing is wrong with your link.
Something is wrong with your method.
Speedtest only makes sense if you are running it on a network where every other equipment is disconnected and the device running Speedtest have all other services shut down.
That way you are testing the link.
If you do not control the setup in this way, then you have no way of knowing what your link is composed of.
Speedtest only show the traffic that it receive and transmit. It does not show what other devices receive and transmit or what other services on your device receive and transmit.
Speedtest is also a test that tries to congest a link to reach a result, so at the moment you are congesting your link around 6% of the time with no useful outcome, which will affeft all the other stuff using your network. Your actions are probably the worst cause for throughput issues at the moment.
You backbone may be 1Gb Fiber but that speed test means that you are loosing some speed based on the server you are connecting for the test to and the route that it goes through. i.e your bottleneck is most likely the router you are using to connect to the ISP one if your ISP has not upgraded you to a better main modem yet).
For a fiber backbone you want the main modem to have a 2.5GbE or 10GbE main port which then either goes to other 10GbE or 2.5GbE ports so that you are not limited at all other than the speed plan you are paying for from the ISP.
Camera’s and basic hardware are more than fine on 1GbE ports.
300Mb/s is also what 2.4Ghz wifi tends to peak out at and a lot of IoT and NoT devices that use it will be using since its low burst data.
TLDR you are never going to get 100% out of your WAN connection ever, there will always be a bottleneck somewhere.
Your internal connection is what you can control and you upgrade past 1GbE ports and switches in places where you need them the most.
If you are not doing a test in a controlled environment, then it can be HA itself using bandwidth, any device on the network using bandwidth, any network device on the route that is underpowered, the ISP’s link to your POI that is faulty (less likely), any ISPs your route to the Speedtest server that your route goes through that is congested.
It is a wild guess what is the cause and your ISP will dismiss any support request if you do not control your own end of the test. They will also dismiss it if the ISPs on the route to the Speedtest server is the cause. They provide you with a line to the internet, not a route to a specific server.
The design of the protocol will make it impossible.
80% is often considered max of a ethernet connection.
Unless you have that speed all the way to each end point that needs it, then it is a waste of money to get those ports.
Cameras and basic hardware can often run on 100Mbit/s with ease.
Often workstations can too. Unless you download or upload a lot and time is a critical factor, then 1Mbit/s is more than enough.
The way the sliding window works in TCP/IP you will often not get over 1Mbit/s anyway, because the other end and the route to that will get packet drops and therefore lower the window and the speed. You would need to download large data amount from sources that have a really connection and route for it to make a difference.
That is what I mean by the route and majority of the time in home environments a poorly setup router or mesh can cause congestion issues internally.
Correct, speedtest picks the closest host by default so you want to test with more than one and run at least 3 tests to get an overall average (also closest is not always the best because as we stated due to congestion and hop issues at the time and its good practice to check via hosts not listed as your provider).
Hence why I stated at the end where they would need them the most i.e to a NAS or media system where you have a dedicated connection internally that can take advantage of it if you do heavy content creation or work with data sets to speed up the access to it, which at that point your limiting factor will be storage read and write speeds based on the OS of the device and how its configured.
It’s why you see most have a 2.5/5/10GbE WAN port and then that goes into separate 1/2.5GbE LAN ports to spread out the bandwidth to all the ports that get connected but the network admin in me wants ALL the ports to be the same even if I can’t yet take advantage of them heh.
Can attest to this since I am currently due to cost on a 100down/40up
plan with my main connection being an NBN coaxial modem for the backhaul then its fed into my main router via bog standard 1GbE ethernet ports on all my devices.
I have no issues using openvpn for access into my LAN when not home to get a live feed of my cameras that are set to 1440p30.
The main bottleneck externally at that point is what connection I am on with the phone.
Sorry again getting carried away… just a habit of making sure things are as clear as possible to those reading and learning.