Hi Frantisek,
sorry, was away for a day...
Thanks for testing, all of these tests show a pretty atrocious latency-under-load/working-conditions for the upload, indicating that this is unlikely the the peerings of your ISP. My guess is your access link and/or your modem or router. Could you describe your network set-up a bit and whether/what measures you took to counter bufferbloat?
Regards
Sebastian
On May 25, 2022, at 23:52, Frantisek Borsik <frantisek.borsik@gmail.com> wrote:
Hello Sebastian,
Thanks. Let’s do this. Kind of servers I see:
So picked a few:
https://www.speedtest.net/my-result/i/5129305856 (t-mobile cz)
https://www.speedtest.net/my-result/i/5129307905 (O2 telefonica cz)
https://www.speedtest.net/my-result/i/5129308857 (local ISP)
https://www.speedtest.net/my-result/i/5129310554 (Germany)
https://www.speedtest.net/my-result/i/5129311633
(Poland)
Sent on the move.
F.
https://www.linkedin.com/in/frantisekborsik
+421919416714
Skype: casioa5302ca
On 24 May 2022 at 9:44:25 AM, Sebastian Moeller (moeller0@gmx.de) wrote:
Hi Frantisek,
in the speedtest app, you can drag up a "drawer" from the bottom, which among others will reveal a "Change Test Server" "link" following that you can manually select from a multitude of servers, just selexct one that is unlikely to be housed by Cogent (like Vodafone.cz). Sure that still does not rule out that Cogent is in the network path*, but if you repeat the test with a few different servers and all show around the same high upload latency this would point to your access link/ISP being the likely responsible part.
Regards
Sebastian
*) The best way to check a path requires both traceroute/mtr to and from the remote endpoint (or at least to/from the network/data center the remote end is housed in). Personally I like to use to following for the forward trace (from local to remote, here goggle's anycasted DNS server 8.8.8.8 as example):
mtr -ezb4 -o LSNBAWVJMXI 8.8.8.8
For the reverse traceroute/mtr either use a looking glas server (if the network offers these) or try to find a node in https://www.globaltraceroute.com that is housed in the correct network and location; this is a bit approximate, but what can we do, short of adding the traceroute/mtr functionality into the actual speedtest so it is always available.
On May 24, 2022, at 09:31, Frantisek Borsik <frantisek.borsik@gmail.com> wrote:
Hey, not sure how I can do this kind of testing :-) but if You tell me, I will do it, happily.
F.
https://www.linkedin.com/in/frantisekborsik
+421919416714
Skype: casioa5302ca
On Tue, May 24, 2022 at 9:09 AM Sebastian Moeller <moeller0@gmx.de> wrote:
Hi Frantisek,
On May 24, 2022, at 06:41, Frantisek Borsik via Rpm <rpm@lists.bufferbloat.net> wrote:
Good, good - most of the time.
https://www.speedtest.net/my-result/i/5127045302
This test is against fdcservers.net which appear to be single-homed in Cogent's network, do you see the same high upload latency under working conditions when measuring against servers in other networks (that are not connected via Cogent)?
Regards
Sebastian
P.S.: Not intended as dig against Cogent it is just that some ISP networks are known to have issues with connectivity with Cogent, so testing with servers in other networks can be a useful check.
Sent on the move.
F.
https://www.linkedin.com/in/frantisekborsik
+421919416714
Skype: casioa5302ca
_______________________________________________
Rpm mailing list
Rpm@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/rpm