ANSWERED Speed or latency rule?

[FrankU 0]

Hi all,

 

The Eddie client shows servers based on speed by default. Would selecting latency not provide the best results?

 

Obviously selecting latency rates servers in my home country more favourably (~3-10ms) but rating by speed shows the highest rated servers elsewhere in Europe (>10ms).

 

Just looking for views on whether I should stick with the default of speed which selects a server outside my home country.

 

Thanks for your feedback! Couldn't find anything on forums regarding whether it's best to stick with client default of speed.

[Staff 9972]

Hi all,

 

The Eddie client shows servers based on speed by default. Would selecting latency not provide the best results?

 

 

Hello,

 

not really, round trip time weighs significantly, even more than available bandwidth (if you mean that for "speed"). Of course if you have a saturated server by bandwidth, the weight of the penalty for such saturation will make the server rating very low regardless of any other parameter, and it is almost impossible that it will be "best server" for any zone.

 

If you want an empiric rule to quickly decide and that is in most cases effective, once you discard heavily populated servers, go for the one with the lowest round trip time from you.

 

Kind regards

[serenacat 83]

Related and not worth raising a new thread and not really a fault and just for info ...

Latency, and in the worst case packet loss, seems to better reflect actual performance. Even from the same geolocation, but with different datacenters or backbone routing.

For about a week, there has been a large difference in latency from Australia to datacenters in HK.

 

Using Hadar pacswitch.com

Tracing route to www.abc.net.au [150.101.195.201]
over a maximum of 30 hops:

  1   301 ms   151 ms   151 ms  10.50.0.1
  2   383 ms   524 ms   388 ms  10.119.0.1
  3   538 ms   415 ms   382 ms  125.63.60.254
  4   385 ms   394 ms   409 ms  ge1-11-242.cor3.syd7.cine.net.au [210.5.35.38]
  5   389 ms   390 ms   392 ms  10ge2-3-55.br1.syd6.cine.net.au [210.5.37.181]
  6   465 ms   410 ms   430 ms  as4739.nsw.ix.asn.au [218.100.52.5]
  7   419 ms   447 ms   388 ms  ae0.br1.syd7.on.ii.net [150.101.33.15]
  8   387 ms   379 ms   390 ms  150.101.195.201

Trace complete.

 

Using Alnilam or other hkserverworks.com
Tracing route to www.abc.net.au [150.101.195.201]
over a maximum of 30 hops:

  1   363 ms   356 ms   359 ms  10.50.0.1
  2   687 ms   689 ms   689 ms  10.119.0.1
  3   692 ms   686 ms   692 ms  125.63.60.254
  4   692 ms   680 ms   691 ms  ge1-11-242.cor3.syd7.cine.net.au [210.5.35.38]
  5   690 ms   689 ms   689 ms  10ge2-3-55.br1.syd6.cine.net.au [210.5.37.181]
  6   687 ms   689 ms   715 ms  as4739.nsw.ix.asn.au [218.100.52.5]
  7   745 ms   684 ms   689 ms  ae0.br1.syd7.on.ii.net [150.101.33.15]
  8   687 ms   689 ms   689 ms  150.101.195.201

Trace complete.
 

whois 125.63.60.254 -> networkpresence.com.au

10.119.x.y would be local routing in the datacenters, but note the large delay in hkserverworks.

I first thought this was different undersea cable routing.Testing with other URLs in Au such as my ISP has consistent results.

Eddie says Anilam latency=334ms load=15% users=10; Hadar latency=146ms load=9% users=26

As usual, Antares Sg is best "score" with latency=124ms load=27% users=133

 

FYI