<WOPR>Would. you. like. to. play. a. game?</WOPR>

bartonski · 12-09-2009, 12:21 AM

Ubee's 'Persistent Routing' thread prompted me to traceroute a few different addresses. I figured that yahoo.com and google.com should be far enough away geographically that I might see some differences in routing (ok, honestly, I don't know where they're located, and I'm guessing California, but that's not entirely relevent.)

I live in Louisville, Kentucky (USA). Louisville, KY has a population of around 1,000,000 -- larg but not huge, and as far as I know, not particularly special in terms of internet connectivity. I'm guessing that both Yahoo and Google are in Western California, (2403 miles/3867 km from my house in Louisville, KY driving to Googleplex in Santa Clara California). I was two hops between my ISP's gateway router and the first router on either Yahoo's or Google's networks respectively, the intervening routers both belonged to Time Warner.

That either means that two hops on Time Warner's network gets me to Google's front door, or that Google owns some of its own long haul network, which connects with major routers (and likewise, Yahoo is doing the same thing).

The other major possibility is that Google and Yahoo have data centers positioned around the country (ahem. around the world), so that I'm likely to be no more than a couple hops off of one of their data centers...

Anyway, here's the game: run a traceroute from your machine to google.com. Post the ip address that you're tracerouting to, and all of the hops from your ISP's gateway router all the way to the target IP address.

I would like to figure out which of the scenarios above is actually correct:

a small number of hops directly to a single point on google's network
a small number of hops to one of a whole bunch of routers, ending at the same IP address or
a small number of hops to one of a whole bunch of routers, ending at different IP addresses.

I'll go first:

Code:

traceroute to google.com (74.125.45.100), 30 hops max, 60 byte packets
 <snip>
 5  po-ch-1.border2.iglou.com (204.252.74.10)  42.355 ms  44.179 ms  46.109 ms
 6  216.84.104.153 (216.84.104.153)  50.142 ms  40.119 ms  43.716 ms
 7  chcg-peer-02.twtelecom.net (66.192.254.70)  55.795 ms  51.029 ms  51.502 ms
 8  209.85.254.120 (209.85.254.120)  35.046 ms 209.85.254.128 (209.85.254.128)  30.369 ms  32.875 ms
 9  209.85.242.215 (209.85.242.215)  54.258 ms 72.14.239.90 (72.14.239.90)  56.174 ms 209.85.242.215 (209.85.242.215)  64.044 ms
10  209.85.254.243 (209.85.254.243)  62.109 ms 209.85.254.241 (209.85.254.241)  111.986 ms  68.020 ms
11  209.85.253.137 (209.85.253.137)  80.124 ms 209.85.253.133 (209.85.253.133)  80.818 ms 209.85.253.141 (209.85.253.141)  81.234 ms
12  yx-in-f100.1e100.net (74.125.45.100)  81.646 ms  83.720 ms  87.182 ms

Code:

$ whois 209.85.254.120

OrgName:    Google Inc.

One other thing: 209.85.254.120 shows 32.875 ms latency.

The maximum theoretical speed of a signal in a fiber optic cable travels at about 2/3 of the speed of light, i.e. 2e10^5 km/s or 200 km/ms. Given 3867 km one way, this would be 2*3867km / 200 km/ms = 38ms latency, if packets were traveling as fast as possible across fiber optic, without any penalty introduced by intermediate routers. This means that if packets are not landing directly at google's doorstep at the googleplex in California.

If someone has real numbers for latency introduced by distance, please post... one of my telecom buddies had a good rule of thumb, but I don't remember what it was... I'm guessing around 50km/ms round trip (half the theoretical maximum) or around 10ms for every 300 miles round trip.

Yahoo isn't really worth looking at, they apparently have ICMP blocked for their internal networks, so you can't see any hops beyond the gateway router.

everest40 · 12-09-2009, 04:45 AM

Here's the output of a traceroute from my home network to Google:

Code:

traceroute to www.google.com (74.125.19.99), 30 hops max, 60 byte packets
 <snip>
10  pos-0-0-0-0-pe01.529bryant.ca.ibone.comcast.net (68.86.86.122)  34.265 ms  34.536 ms  34.500 ms
11  75.149.230.190 (75.149.230.190)  35.577 ms  36.518 ms  36.481 ms
12  216.239.49.168 (216.239.49.168)  39.633 ms  40.706 ms  40.083 ms
13  209.85.251.94 (209.85.251.94)  40.626 ms  39.010 ms  39.475 ms
14  nuq04s01-in-f99.1e100.net (74.125.19.99)  35.895 ms  36.412 ms  36.970 ms

#10 and 11 appear to be owned by Comcast (my ISP), and #12 is owned by Google.

everest40 · 12-09-2009, 04:54 AM

Just realized, your traceroute was to google.com and mine was to www.google.com. Here's the output of a traceroute to google.com:

Code:

traceroute to google.com (74.125.45.100), 30 hops max, 60 byte packets
<snip>
10  pos-0-0-0-0-pe01.529bryant.ca.ibone.comcast.net (68.86.86.122)  33.266 ms  33.655 ms  33.619 ms
11  75.149.230.190 (75.149.230.190)  34.595 ms  35.444 ms  35.336 ms
12  216.239.49.170 (216.239.49.170)  33.780 ms  32.490 ms 216.239.49.168 (216.239.49.168)  31.664 ms
13  209.85.243.247 (209.85.243.247)  94.386 ms  89.998 ms 209.85.249.140 (209.85.249.140)  106.942 ms
14  72.14.232.215 (72.14.232.215)  92.513 ms 209.85.254.243 (209.85.254.243)  95.420 ms  95.383 ms
15  209.85.253.141 (209.85.253.141)  103.296 ms 209.85.253.145 (209.85.253.145)  103.251 ms 209.85.253.141 (209.85.253.141)  102.680 ms
16  yx-in-f100.1e100.net (74.125.45.100)  95.202 ms  95.169 ms  93.637 ms

Like last time, #10 and 11 appear to be owned by Comcast and #12 appears to be owned by Google.

catkin · 12-09-2009, 07:04 AM

Some very shy nodes between me and Google but here's the output FWIW

Code:

traceroute google.com
traceroute: Warning: google.com has multiple addresses; using 74.125.53.100
traceroute to google.com (74.125.53.100), 30 hops max, 38 byte packets
 1  192.168.1.1 (192.168.1.1)  0.415 ms  0.401 ms  0.386 ms
 2  * * *
 3  * * *
 4  * * *
 5  * * *
 6  * * *
 7  * * *
 8  * * *
 9  * * *
10  * * *
11  * * *
12  * * *
13  * * *
14  * * *
15  pw-in-f100.1e100.net (74.125.53.100)  355.124 ms  342.206 ms  349.838 ms

bartonski · 12-09-2009, 07:31 AM

Quote:

Originally Posted by catkin

Some very shy nodes between me and Google but here's the output FWIW

Code:

traceroute google.com
traceroute: Warning: google.com has multiple addresses; using 74.125.53.100
traceroute to google.com (74.125.53.100), 30 hops max, 38 byte packets
 1  192.168.1.1 (192.168.1.1)  0.415 ms  0.401 ms  0.386 ms
 2  * * *
...  
14  * * *
15  pw-in-f100.1e100.net (74.125.53.100)  355.124 ms  342.206 ms  349.838 ms

Ah, that's a shame, I would have liked to have seen some of the hops through India.

catkin · 12-09-2009, 07:44 AM

Quote:

Originally Posted by bartonski

Ah, that's a shame, I would have liked to have seen some of the hops through India.

And so you shall, courtesy of Mr Gates

Code:

C:\Documents and Settings\C>tracert google.com

Tracing route to google.com [74.125.45.100]
over a maximum of 30 hops:

  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2    43 ms    44 ms    41 ms  59.92.0.1
  3    65 ms    64 ms    64 ms  218.248.255.90
  4    64 ms    64 ms    63 ms  115.113.128.17.static-mumbai.vsnl.net.in [115.113.128.17]
  5    67 ms    66 ms    66 ms  172.31.4.93
  6    66 ms    65 ms    66 ms  172.31.4.93
  7   337 ms   337 ms   338 ms  59.163.16.234.static.vsnl.net.in [59.163.16.234]
  8   349 ms   339 ms   349 ms  72.14.232.110
  9   352 ms   341 ms   339 ms  66.249.94.74
 10   336 ms   357 ms   335 ms  209.85.254.155
 11   341 ms   352 ms   341 ms  209.85.255.37
 12   337 ms   345 ms   336 ms  209.85.248.131
 13   346 ms   346 ms   349 ms  216.239.43.81
 14   350 ms   348 ms   349 ms  72.14.239.90
 15   347 ms   349 ms   345 ms  72.14.232.215
 16   351 ms   363 ms   355 ms  209.85.253.133
 17   350 ms   346 ms   347 ms  yx-in-f100.1e100.net [74.125.45.100]

Trace complete.

Which raises the question why a WXP tracert generated more information than a Slackware traceroute ...

bartonski · 12-09-2009, 07:56 AM

Quote:

Originally Posted by catkin

And so you shall, courtesy of Mr Gates

Code:

  4    64 ms    64 ms    63 ms  115.113.128.17.static-mumbai.vsnl.net.in [115.113.128.17]
  5    67 ms    66 ms    66 ms  172.31.4.93
  6    66 ms    65 ms    66 ms  172.31.4.93
  7   337 ms   337 ms   338 ms  59.163.16.234.static.vsnl.net.in [59.163.16.234]

Wow, that's a hell of a jump in latency, especially since whois says that both routers are in Mumbai...

Quote:

Which raises the question why a WXP tracert generated more information than a Slackware traceroute ...

Yeah, that got my attention too.

di11rod · 12-09-2009, 01:25 PM

Code:

Quad-Damage:~ $ traceroute google.com
traceroute: Warning: google.com has multiple addresses; using 74.125.53.100
traceroute to google.com (74.125.53.100), 64 hops max, 40 byte packets
 1  192.168.0.1 (192.168.0.1)  0.787 ms  0.201 ms  0.248 ms
 2  * * *
 3  ge-1-3-0-160.aggr01.austtx.grandecom.net (66.90.139.54)  15.685 ms  18.078 ms  24.252 ms
 4  so-2-0-1-0.core01.smrctx.grandecom.net (24.155.121.9)  13.454 ms  10.602 ms  15.273 ms
 5  so-5-0-0.core01.dllstx.grandecom.net (24.155.121.135)  17.891 ms  19.593 ms  16.634 ms
 6  ge-0-0-0.core02.dllstx.grandecom.net (24.155.121.128)  31.953 ms  18.951 ms  25.556 ms
 7  66.90.138.77 (66.90.138.77)  39.323 ms  51.527 ms  26.572 ms
 8  72.14.233.65 (72.14.233.65)  25.382 ms  22.313 ms  39.359 ms
 9  72.14.233.111 (72.14.233.111)  59.844 ms  81.526 ms  56.569 ms
10  209.85.249.189 (209.85.249.189)  58.315 ms  56.685 ms  67.576 ms
11  72.14.232.85 (72.14.232.85)  97.266 ms  95.032 ms  92.872 ms
12  209.85.250.144 (209.85.250.144)  100.756 ms  100.387 ms 216.239.48.34 (216.239.48.34)  82.410 ms
13  64.233.174.129 (64.233.174.129)  85.078 ms 64.233.174.123 (64.233.174.123)  81.389 ms  94.048 ms
14  72.14.232.10 (72.14.232.10)  96.929 ms 72.14.232.70 (72.14.232.70)  111.762 ms 72.14.232.2 (72.14.232.2)  82.894 ms
15  pw-in-f100.1e100.net (74.125.53.100)  105.523 ms  82.352 ms  81.626 ms

bartonski · 12-09-2009, 03:43 PM

Quote:

Originally Posted by di11rod

Code:

Quad-Damage:~ $ traceroute google.com
...
 7  66.90.138.77 (66.90.138.77)  39.323 ms  51.527 ms  26.572 ms
 8  72.14.233.65 (72.14.233.65)  25.382 ms  22.313 ms  39.359 ms
...

Hop 7 is Grande Communications Networks, Inc, Hop 8 is one of Google's routers. I'm assuming that Grande Communications is your ISP?

No daylight there...

catkin · 12-10-2009, 01:35 AM

Quote:

Originally Posted by bartonski

Wow, that's a hell of a jump in latency, especially since whois says that both routers are in Mumbai...

[Regards WXP trace giving more output than Linux's] Yeah, that got my attention too.

Turns out Windows and Linux implement route tracing differently as explained here and here. The Linux command has a -I (letter I for ICMP) option that makes it behave similarly to the Windows version -- useful when the default Linux packets are firewalled. Here's tracing from Tamil Nadu again (output prettified into columns for legibility. The 3rd figure for step 12 was missing before editing)

Code:

traceroute -I google.com
traceroute: Warning: google.com has multiple addresses; using 64.233.181.103
traceroute to google.com (64.233.181.103), 30 hops max, 38 byte packets
 1  192.168.1.1 (192.168.1.1)                                   0.441 ms   0.405 ms   0.400 ms
 2  59.92.0.1 (59.92.0.1)                                      42.780 ms  85.597 ms 101.447 ms
 3  218.248.255.6 (218.248.255.6)                              85.415 ms  90.903 ms  40.732 ms
 4  218.248.246.130 (218.248.246.130)                          60.986 ms  71.373 ms  58.985 ms
 5  115.113.128.17.static-mumbai.vsnl.net.in (115.113.128.17)  72.797 ms 110.392 ms 259.650 ms
 6  59.163.25.242.static.vsnl.net.in (59.163.25.242)           59.956 ms  60.883 ms 126.649 ms
 7  59.163.16.238.static.vsnl.net.in (59.163.16.238)           68.561 ms  60.380 ms 157.884 ms
 8  216.239.43.214 (216.239.43.214)                           107.347 ms  63.835 ms  61.557 ms
 9  72.14.232.100 (72.14.232.100)                              87.343 ms  
                                 72.14.232.93 (72.14.232.93)             139.132 ms  86.983 ms
10  209.85.241.86 (209.85.241.86)                             229.315 ms  
                                 66.249.94.38 (66.249.94.38)             123.646 ms 
                               209.85.241.86 (209.85.241.86)                        196.596 ms
11  209.85.250.255 (209.85.250.255)                           123.636 ms 
                             209.85.250.237 (209.85.250.237)             127.075 ms 
                             209.85.241.173 (209.85.241.173)                        156.383 ms
12  * 209.85.243.142 (209.85.243.142)                         164.814 ms 171.900 ms
13  74.125.206.6 (74.125.206.6)                               204.970 ms 163.635 ms 
                             74.125.206.134 (74.125.206.134)                        129.533 ms
14  ni-in-f103.1e100.net (64.233.181.103)                     128.546 ms 125.615 ms 168.760 ms

bartonski · 12-10-2009, 06:24 AM

As one of my co-workers is prone to say "Ok, I've learned my fact for the day, can I go home now?" (Ironic, because I'm writing this before I head to work).

[.me edits .bashrc]

Code:

alias traceroute='traceroute -l'

[/edit .bashrc]

Code:

$ traceroute www.google.com
The specified type of tracerouting is allowed for superuser only

Oops. that's not what I want. I'll keep that in my bag of tricks rather than my .bashrc.

[.me edits .bashrc]

Code:

$ vim +/traceroute ~/.bashrc
dd
wq

[/edit .bashrc]

Code:

$ exec bash
$ traceroute www.google.com
traceroute to www.google.com (64.233.169.105), 30 hops max, 40 byte packets
 1  router.localhost (192.168.1.1)  1.116 ms  1.563 ms  2.018 ms
 2  10.0.0.2 (10.0.0.2)  14.016 ms  14.222 ms  14.429 ms

That's better.

Thanks, Catkin, I think that you win cool tidbit of information of the day.

bartonski · 12-14-2009, 04:42 PM

When I read this, I couldn't help thinking about everything that I wrote about speed of light in fiber optic cables above...