[LINK] Broadband for a Broad Land

[Scott Howard]

On Sun, Jan 2, 2011 at 7:52 PM, Tom Koltai <tomk at unwired.com.au> wrote:

> I think you are correct Richard, I used to calculate each "hop" as being
> 34 ms.


When was this?  1983?

Tracing the route to www.l.google.com (74.125.224.48)

  1 192.168.250.250 0 msec 0 msec 4 msec
  2 bras10-l0.pltnca.sbcglobal.net (151.164.184.66) 36 msec 36 msec 40 msec
  3 64.164.107.2 40 msec 32 msec 32 msec
  4 151.164.93.239 32 msec 36 msec 36 msec
  5 151.164.101.130 40 msec 40 msec 44 msec
  6 74.125.48.181 40 msec 36 msec 88 msec
  7 216.239.49.250 24 msec 24 msec 24 msec
  8 64.233.174.15 28 msec 24 msec 20 msec
  9 www.l.google.com (74.125.224.48) 20 msec 20 msec 16 msec


That's 9 hops in (no more than) 16 ms.   In practice, basically all of that
latency will be being introduced over the first hop (DSL).  eg, from a
system connected to the Internet using Ethernet, the 7 hops to the same host
take ~2 ms :

[scott at willers ~]$ traceroute www.google.com
traceroute: Warning: www.google.com has multiple addresses; using
74.125.19.104
traceroute to www.l.google.com (74.125.19.104), 64 hops max, 40 byte packets
 1  gw1-vlan94.scl.kjsl.com (198.137.202.29)  0.747 ms  1.194 ms  0.927 ms
 2  vl2.core1.scl.layer42.net (69.36.225.129)  1.302 ms  1.319 ms  1.978 ms
 3  po1-900.core1.sv1.layer42.net (69.36.239.61)  1.985 ms  1.343 ms  1.802
ms
 4  eqixsj-google-gige.google.com (206.223.116.21)  1.029 ms  1.140 ms
1.089 ms
 5  216.239.49.168 (216.239.49.168)  2.724 ms  3.845 ms  3.131 ms
 6  209.85.251.94 (209.85.251.94)  3.599 ms  12.225 ms  2.159 ms
 7  nuq04s01-in-f104.1e100.net (74.125.19.104)  2.084 ms  2.815 ms  2.021 ms

(For anyone that's wondering, the times to the intermediate routers are
generally artificially higher than they would be to a host at the equivalent
hop due to the way traceroute works and the way routers handle it).



> Therefore 10 hops on a single ISP with older style routers
> attempting to define a set of QOS rules adds 340 ms latency to each
> packet.
>

Perhaps you'd like to legitimize your comments by naming even a single ISP
that is using these "older style routers" of which you speak?

  Scott.