Ikoula / Starter IKL Speedtest

 Speedtest Monster v.1.2.6 beta (30 Sep 2019) 
 Region: Global  https://bench.monster/speedtest.html
 curl -LsO bench.monster/speedtest.sh; sh speedtest.sh -Global

 CPU Model            : Intel(R) Celeron(R) CPU G3930 @ 2.90GHz
 CPU Cores            : 2 Cores @ 2900.000 MHz x86_64 2048 KB Cache
 OS                   : CentOS 7.7.1908 (64 Bit)
 Kernel               : Dedicated / 3.10.0-1062.1.1.el7.x86_64
 Total Space          : 4.3 GB / 1844.2 GB 
 Total RAM            : 665 MB / 7708 MB (2310 MB Buff)
 Total SWAP           : 0 MB / 1023 MB
 Uptime               : 13 days 8 hour 1 min
 Load Average         : 0.02, 0.06, 0.05
 ASN & ISP            : AS21409, Ikoula Ripe
 Organization         : 
 Location             : Boulogne-Billancourt, France / FR
 Region               : Île-de-France

## dd: sequential write speed (1.0GB):

 1st run   : 186 MB/s
 2dn run   : 190 MB/s
 3rd run   : 188 MB/s
 Average   : 188.0 MB/s

## Global Speedtest

 Location                       Upload           Download         Ping   
 Speedtest.net                  855.63 Mbit/s    608.00 Mbit/s    5.744 ms
 USA, New York (AT&T)           198.95 Mbit/s    204.02 Mbit/s    78.478 ms
 USA, Chicago (Windstream)      9.10 Mbit/s      150.45 Mbit/s   106.498 ms
 USA, Dallas (Frontier)         147.97 Mbit/s    233.23 Mbit/s   136.106 ms
 USA, Miami (Frontier)          166.90 Mbit/s    294.34 Mbit/s   116.219 ms
 USA, Los Angeles (Spectrum)    53.21 Mbit/s     123.40 Mbit/s   148.918 ms
 UK, London (Community Fibre)   710.48 Mbit/s    696.86 Mbit/s    10.593 ms
 France, Lyon (SFR)             627.65 Mbit/s    308.72 Mbit/s     9.295 ms
 Germany, Berlin (DNS:NET)      352.53 Mbit/s    176.80 Mbit/s    28.186 ms
 Spain, Madrid (Adamo)          590.05 Mbit/s    169.37 Mbit/s    21.714 ms
 Italy, Rome (Vodafone)         232.73 Mbit/s    517.54 Mbit/s    28.028 ms
 Russia, Moscow (MTS)           134.62 Mbit/s    127.35 Mbit/s    60.419 ms
 Israel, Haifa (013Netvision)   14.96 Mbit/s     267.22 Mbit/s    78.561 ms
 India, New Delhi (Spectra)     13.06 Mbit/s     2.84 Mbit/s     292.364 ms
 Singapore (FirstMedia)         17.44 Mbit/s     4.90 Mbit/s     164.700 ms
 Japan, Tokyo (ATCC)            19.78 Mbit/s     10.35 Mbit/s    220.702 ms
 Australia, Sydney (Yes Optus)  18.51 Mbit/s     14.19 Mbit/s    308.124 ms
 RSA, Randburg (Cool Ideas)     101.81 Mbit/s    264.33 Mbit/s   175.217 ms

 Timestamp    : 2019-10-01 20:33:55 UTC

 Share result:
 · http://www.speedtest.net/result/8639257788.png

I like the output of this script, especially since it tests BOTH upload and download speeds. Is this your project?


No, This is improved version of script/superbench.sh at master · oooldking/script · GitHub

But it has so much potential left behind

1 Like

It’s using speedtest-cli… There’s a reason it’s not used much by benchmark scripts.


Yeah, I realized that after I saw the share link at the bottom of the output. Would be nice if there’s an alternative out there that can test both download and upload speeds. Maybe I’ll write something up that uses iperf or something. Problem is, there’s only a few public iperf servers out there.

what is wrong with that script?

I think it’s awesome and the real measure between the server and real visitors

the iperf, speedtest from some servers is not a real visitors experience

speedtest-cli, apart from being wildly inaccurate and showing completely unrealistic numbers (0.2Mbps upload to a server located 0.2ms away on an uncontested gigabit pipe) introduces too many variables into the equation.

If you see low numbers, you have no idea whether
a) the speedtest server is just crap
b) the speedtest server isn’t able to handle your speeds (since it can be on a 10Mbps line and you wouldn’t know)
c) your connection is bad
d) the issue is somewhere in the middle
e) requires quite a bit of CPU at higher speeds

With iperf, you know the IP of the server, which means you can traceroute it to troubleshoot any issues. The iperf server operators also generally publish how fast the connection is.

Not to mention iperf can utilize multiple threads, which simulates real world usage much much better (a server generally serves multiple users), while being able to max out any connection (assuming both sides are fast enough)

Also, iperf does IPv6 :joy: