I'm trying to set up a script I can schedule to run hourly or so that appends something like [timestamp] [speed] (and ideally [variance])to a log file that I can turn into pretty graphs. The command I have now is:
The highest speed you will see is the maximum speed that your connection and the server's connection can offer. If your connection is 512KB and the place where you are downloading is 400KB, your max connection will be 400KB because it is the max for the server you are downloading from.
Wget Check Download Speed
Download Zip 🔥 https://urluso.com/2y7ZcC 🔥
You need to have at least 4 or 5 different testing sources to have a more accurate speed. Never test only from the same site as this can be affected by your distance to it, any problem in the server and the connections to it, etc. Always test from different servers.
Actually, Ookla, the provider of Speedtest released a command-line utility that is measuring your speed against a huge number of servers spread around the world. You can find instructions on how to install it on this link and you can use it fairly simple by executing:
where -s sets the server ID against which you want to test your Internet speed, -f is defining the format of the output. I think the most useful information is generated when you use json/json-pretty format for the output because a lot of the information of the test setup isn't printed if you are using the csv/tsv format. Both -s and -f are just optional but if you want to automate your measurement they might be useful.
It's a more generic solution that can be consumed by other applications. I managed to pinpoint an ISP network issue with it that resulted in doubling my SFTP downstream speed, which resulted in increased happiness.
With the increase in people staying at home and spending more time on the Internet, ISPs have seen traffic loads higher than ever. If you noticed your network speed was slower at times, this global overload is the reason.
The tools in this guide help you check the Internet and LAN speed on a Linux machine. The article uses Ubuntu 20.04 for instructions, but the utilities work for any Linux distribution.
I get 190~200Mbps dowload and 100Mbps upload on fast and speedtest-cli, and +20MB/sec downloads in pacman and curl.
But the problem is that in wget, firefox or any other browser, I consistently get either 500KB/sec or up to 11.5MB/sec downloads.
To test it, I've used the largest file (1.4GB) on the fastest arch mirror I could get for my region:
curl and pacman ALWAYS get +20MB/sec, while wget and firefox ALWAYS get the speeds mentioned.
An interesting thing is that after testing it sometimes, wget/firefox dl speeds seem to alternate between up to 11.5MB/sec and up to 500KB/sec.
I should mention that I'm using an extremely light wm (evilwm) and that these things happen even on console, so it isn't a question of wm bloat slowing down the dl speeds.
I've also stopped testing using that server constantly to avoid abusing it.
Without using wget or explicitly telling Firefox to download anything, the results are pretty much perfect: 20.20ms ping, 213.49Mbps download and 99.01Mbps upload - even higher than my nominal 200Mbps dl and 100Mbps ul.
But if I try to download anything anywhere, it alternates between 500KB/sec and up to 11.5MB/sec - ALWAYS, while curl and pacman get +20MB/sec consistently.
Sorry for taking so long to reply, it was a long week... Anyway, you were right on the money:
Downloading the 1GB test files from this site, every single program - wget, firefox, curl, etc. - got 18~20MB/sec or more, but then trying to dl anything from any good mirror near me or even pretty much any file on a number of sites I've visited would get the usual results: 500~900KB/sec on wget and firefox, 20MB/sex or more on curl.
Then I followed your suggestion and after this:
rtorrent seems to have its own thing, with speeds rising rapdily to 18~20MB/sec until close to the end of the file, seemingly pausing for a few seconds, then coming crashing down to 5~7MB/sec where it stays until the end of the downloads.
No idea why.
It happens on both wireless and ethernet, makes no difference.
Oh, and I also tested it again on Windows: The Debian iso gets +16MB/s, Ubuntu's get +14~15MB/s, testfile's get +18MB/s but the arch linux files get lower speeds, going from 500KB/s to 5-7MB/s.
Which is strange, since pacman and curl'ing the arch files used to get 20MB/s constantly.
Yes, I've disabled windows fast-start - in fact, I'm running my system from a pendrive and changed boot order in UEFI: it would be impossible to do that with fast-start active.
And I've run the tests again with IPv6 off, both directing to /dev/null and to my home folder, which is on RAM; the results are pretty much the same: downloading the arch linux files with firefox, wget or curl - these last 2 both to /home/x or /dev/null - it starts at 20MB/s, falls to ~7MB/s, goes up to 14MB/s, down to 10~11MB/s, up and down between 15MB/s and 7~9MB/s and finishes at 23MB/s or 11MB/s. testfile.org gets +20MB/s through and through.
In short, pretty much the same.
Every single time I got precisely the same results, so I startx'ed into my wm (evilwm) and repeated the tests - with wget and curl as before, but now also including firefox, and once again, exactly the same results:
I then rebooted back into graphical.target, and stil got the same speeds.
The next day, still wired and using the same initial config you suggested, testfile and the arch zst still got the same speeds, but ubuntu and debian speed had fell a bit.
Disconnecting the cable and using wifi however, the difference in speeds was very clear: testfile didn't change with wget, but curl got 6-10MB/s; arch zst got 14-20MB/s, averaging on 14, both with wget and curl; ubuntu and debian oscillated between 0 and 10MB/s; firefox got about the same.
Downloading to /dev/null, /home or the pendrive had no clear difference in speed wired or wireless, except that, in both cases using firefox, "a few seconds remain" would crash from 18 seconds to a minute and a half when downloading to the pendrive.
Today all speeds were 100% again, wired and wireless (arch zst +20MB/s on wifi), except ubuntu that oscillated sometimes and was in general a bit slower.
So while these didn't solve entirely the issue, turning off ipv6 really raised the speeds.
But the most interesting thing is how consistent is the behaviour of the oscillation/low speeds:
1. testfile pretty much ALWAYS* has +20MB/s;
2. arch zst files now ALWAYS have 24.1MB/s through and through, thousands of times wired; but can oscillate between 14-22MB/s (or worse) wireless;
3, Ubuntu and Debian can oscillate even wired, but will ALWAYS be worse on wireless - usually a LOT worse;
*The very first time you run the test can sometimes oscillate/have low speeds, but then you can test hundreds of times under different circunstances and you'll get 100%. Also there's always a message like error 301 or 206 at th beginning of every test with wget, but that never has any effect.
Unless there're meaningful differences in routing on your side (IPv6/4 or completely different route to the target IP) that's down to the targeted host, I suggest to focus on a single, speedtest oriented destination per IP stack (4/6)
It took me a while to answer because I've had a lot of instability in my internet connection lately, which made it difficult to make accurate measurements and zero in on where the issues are, but eventually I managed to do it, I think.
To see if it was an issue with a faulty pendrive as you mentioned, I've repeated the tests on a backup of my system that hasn't been updated in a while and also with the latest arch linux iso - the former on an entirely different model and brand of pendrive (HP x795w) and the latter on another pendrive of the same model and brand (Sandisk Ultrafit), all USB 3,0 pendrives connected on USB 3.0 ports and confirmed to be fucntioning as USB 3.0 with "lsusb -t". The results were the same as on my system:
1. testfile.org gets +20MB/s almost* always, wired or over wifi, through and through;
2. arch zst gets 24.1MB/s all the way with no variation** wired and 20-24MB/s wireless;
Ubuntu and debian isos are WAY more affected by wireless and network issues:
3. Ubuntu's speeds are 16-18MB/s when it's good and 6-8MB/s when bad wired, and 10-12MB/s wireless when good but oscillating between 0 and 6-8MB/s wirelessly when bad constantly.
4. Debian's speeds are 14-16MB/s when it's good and 6-8MB/s when bad wired and 10-12MB/s wireless whan good but oscillating between 0 and 2-4MB/s wirelessly when bad constantly.
*The first time it usually gets 6-8MB/s, then it's 20MB/s all the way, except for the occasional drop to 12-14 for 3 or 4 seconds then back to 20MB/s;
**Arch zst will get 500KB/s-1.5MB/s randomly but very consistently, from beginning to end of the tests, but that is an issue with my ISP or the Arch mirror in question, affecting even Windows and Android devices when it happens;
In short, testfile and arch zsts don't seem to be really affected by the wireless issue, but ubuntu and debian isos very much so.
Oh, ipv6 off seems to help with the arch mirror on Linux, not the other downloads or Windows speeds.
Any ideas on what could be happening and how to proceed?
testfile.org seems to produce consistent results across the board and it does have an IPv6 and an IPv4 (check the IPs resolved by curl or wget), so it's a good metric and afaict does not(?) seem to indicate any issues?
The test basically downloads a 10MB file using the wget command to pull a file from a remote server. The speed is displayed in the terminal when the file download has completed, so you know how fast your connection is running.
Arch served me really well over the past years but recently when I formatted my computer to install Arch again with AwesomeWM I saw that the download speed within pacman is really slow. This is not just in pacman rather in yay too but I won't take yay into consideration since it is not supported by Arch. Before my format I used to get 2-3 Mbps everytime I used to download using pacman but now the speed is around 200-300 Kbps and sometime fluctuating to 700 Kbps. I know such question has previously been asked and even answered but most of such answers contain people solving the issue after the update but in my case no update actually worked. 006ab0faaa