Comments on: Hosting in Australia Doesn’t Necessarily Give You Low Latency

By: JosephB

JosephB — Tue, 27 Oct 2009 03:52:43 +0000

Doesn’t look good. Still, it does look to be specific to a local LAN/VLAN or host as opposed to a wider issue with the DC etc.

By: scotty

scotty — Tue, 27 Oct 2009 01:01:56 +0000

Here's the reverse traceback from my home ADSL connect (Exetel, Sydney POP).

HOST: my-home-box                Loss%   Snt   Last   Avg  Best  Wrst StDev
 1. 220.233.xxx.yyy               0.0%    10   23.6  22.3  21.2  23.6   0.8
 ...
 4. 220.233.2.37                  0.0%    10   22.6  22.5  21.6  24.0   0.7
 5. 218.100.2.13                  0.0%    10   24.6  23.4  22.1  24.6   0.8
 6. 202.60.84.98                  0.0%    10   27.1  29.6  23.0  71.4  14.7
 7. 118.127.9.xx                  0.0%    10   42.6  60.6  41.0 224.6  57.6
 8. 202.60.89.xx                 10.0%    10  195.7 192.1 179.7 210.1   9.5
 9. 202.60.89.xx                 10.0%    10  196.7 188.9 170.3 210.3  10.9

(6) to (7) would be from Sydney to Brisbane. (9) is my VPS. The latency looks pretty similar from this view.

By: JosephB

JosephB — Mon, 26 Oct 2009 23:55:45 +0000

I’m doubtful that it’s between the VE host and the first router considering it should be 100mbps.

Although if someone bought colo with a fixed pipe, say 10m/bits and the pipe is full, then it could look like that.

What’s the inbound or reverse traceroute look like?

By: scotty

scotty — Mon, 26 Oct 2009 22:58:27 +0000

Yes. Overselling bandwidth is bad regardless whether it’s in US or in Australia. However I am surprised that there’s any overselling on the first hop between the host computer and the next router in the loop — inside an Australian DC. You probably know more about this than I do. I am just trying to point out that the argument “host in Australia for low latency” is mute, if the pipe is congested.

By: JosephB

JosephB — Mon, 26 Oct 2009 14:40:29 +0000

That’s pretty awful congestion, yuk.

Uber bandwidth limits are no good if you can’t use it :-)