Speedtest confusion

Tech Talk
1

I’m sure someone here will provide an answer to my confusion.
The retirement community I live in just upgraded our internet connection to the Australian NBN from a single copper connection with download speed of about 40mbps and upload speed about 10 mbps, to a single fibre connection, with a 100mbps modem (they will be installing a 200 mbps modem in the next few weeks). Several residents have been asked to monitor the internet speed and reliability.

So I did some speedtests. Mostly using Ookla speedtest, but also some others like DSL Reports, and Telstra speedtest. Results make little or no sense. I get upload speeds that are fairly consistent of about 90mbps, but get some around 15mbps. But download speeds vary from 15 - 85mbps, and change constantly. They vary significantly depending on the device I use to test. A Fetch DVR (like a Tivo) connected via ethernet shows 28mbps down, but others show numbers varying between 20 - 80mbps down.

And to add to the confusion, I get different results from Ookla depending on server I choose, and my iPhone 11 and wife’s iPhone iPhone 11 Pro. If I force the Ookla app to use the same server, I get pretty much same results. But if I allow the app to pick server automatically, it picks different servers, and the numbers are no where near each other.

And to confuse things more, my wife’s iPhone Ookla app thinks we are 42kms from Sydney (we are 500kms), but Maps and Google Maps both show correct location. My iPhone appears correct. I just checked my Apple TV and Mac, and Ookla thinks they are in Sydney. All on the same router within my house.

Asked one of the other residents considered tech wise to do speed test. Came back with 21mbps down, 30 mbps up.

So my take away is this. Totally confused by location error. Using a speedtest is almost useless unless one takes the time to make sure app is configured to use the same testing server route, and the route chosen will result in different results. Some devices must have speed limitations, probably result of the hardware. And, when village gets results from random residents, the results will be almost useless.

So, how does one get a useful measurement of the actual speed of the connection from home/network to “the internet”?

2

In most cases there will be many network hops between you and any other useful server, and at any of those points, there can be congestion and delay. The point of a speed test wanting a local connection is to avoid as many of these interconnection points as possible.

Once you reach a certain speed threshold, the speed of your connection is virtually useless information unless you take into account the service being consumed. Speed test apps are optimized for one thing, making you feel like you have fast speeds. In reality, on a day by day basis, you’re connecting all over the planet (or certainly continent) for services, from news, to YouTube videos, to online banking to lord knows what else. The server (on the other end) providing you with the service plays a bigger part in your experience of speed than most anything else. The network links between you and the server are the next limitation on your perceived speeds. And finally, the device(s) you are using (and its local network conditions (congested WiFi?)) play the final part. Time of day or day of week also plays a part, as people wake up, get home from work, or something important happens (a new Netflix series debuts?)

I wouldn’t get too concerned about the variability… it’s too difficult to control for all the variables. As long as it’s not getting worse than before it’s probably fine.

2 Likes
3

@PHolder is absolutely correct. Speeds are only as good as what you are connecting to. The only controlled speeds that you have are on your own lan and even that varies wildly at times. I have 1 gig fiber at home but most of the time I don’t see more then 100-300 connections to most things.

2 Likes
4

My tips to get the most useful speed measurements:

  • Use a 1 Gbps or faster wired network connection on the measuring device, preferably directly connected the router or modem. 100 Mbps wired connections, and most WiFi connections, add a large number of variables that can lower the measured speed of 100 Mbps and higher internet connections.
  • Run the speed tests at times when the least number of users are actively streaming/downloading from the internet.
  • For continuous monitoring a good, inexpensive, low labor solution is to install a Fingbox (With the wired connection directly to the router).

I receive a monthly connection status report from a Fingbox connected to my router. Whenever I think there is problem, I query the current speed via the Fing app on one of my phones/tables/PCs. Over the past year almost every time I’ve checked it has turned out to be LAN or WiFi issues causing the problem, not my provider (Spectrum 100Mbps cable).

2 Likes
5

Your mention of Fingbox got my attention. It is offered here in Australia, but I cannot remember ever seeing a website for a commercially available product that says less of what it can do.

I need a device that can be attached via ethernet to my router that will constantly monitor that the retirement village provided internet is up. I have been using a 15 year old Sony Vaio running XP to run an app called Net Uptime Monitor, which pings 3 selected IP’s at a set interval (I use 2 seconds), and reports any outages longer than 2 seconds, but the Vaio physically died several weeks ago. I thought about getting a Raspberry Pi, and writing my own monitor code. Even wrote the code using Python3 on my Mac, and testing it using a virtual Raspberry Pi on VirtualBox. But setting up a Raspberry Pi is not as simple as just buying the board. Ends up over $300 for what I want to use.

I can run Net Uptime Monitor on my Bootcamp based Windows 10 system, but have to use wifi connection during the day. Not interested in monitoring internet uptime over wifi as that does not indicate the village network performance. Also locks me out of my MacOS environment.

From your experience, can the Fingbox provide this internet uptime monitoring feature?

6

I had no problems with variations in speedtest results within what I consider reasonable limits. But, I guess I have now learned I need to do several tests with different servers to come up with some visibility into what speed we are getting from our village network at a given moment in time.
An interesting (to me) result of doing all this testing is finding ping results from Cloudflare (1.1.1.1) are showing 1 - 2 ms vs ~30 ms from most other IPs.

7

I think they figure most buyers of their product will skim the low on details sales/marketing page and head to the knowledge base, owners manual, and forum for details.

Since you have now said that you need monitoring at 2 second intervals, not just periodic speed measurements, I don’t know if a Fingbox is right for your application.

In case you want to look into it further here’s the main Fingbox help page:

Here’s the internet speed monitoring knowledge base article.

1 Like