[Zope] access to zope.com ? a bad router

Michel Pelletier michel@digicool.com
Fri, 9 Apr 1999 10:23:06 -0400


> -----Original Message-----
> From: Robert OConnor [mailto:rocon@pivot.net]
> Sent: Friday, April 09, 1999 9:05 AM
> To: zope@zope.org
> Subject: [Zope] access to zope.com ? a bad router
> 
> 
> Again today I couldn't access the zope.com website.
> The trace reveled:
> 
>   1  1184 ms   109 ms   105 ms  stan-m010.cybertours.com 
> [208.153.18.2]
>   2   110 ms   110 ms   110 ms  stan-r1.pivot.net [206.98.61.126]
>   3   115 ms   115 ms   109 ms  prtl-r0-e3-0.cyberTours.com 
> [206.98.60.253]
>   4   120 ms   120 ms   114 ms  borderx1-hssi4-0.Boston.cw.net
> [204.70.179.109]
> 
>   5   125 ms   129 ms   119 ms  core-fddi-0.Boston.cw.net 
> [204.70.2.33]
>   6   125 ms   120 ms     *     core4.WestOrange.cw.net [204.70.4.77]
>   7   129 ms   124 ms   125 ms  sprint3-nap.WestOrange.cw.net
> [204.70.10.226]
>   8     *      125 ms   119 ms  pnnj-01.core.exodus.net 
> [192.157.69.25]
>   9     *        *      135 ms  dcr01-h10-0-0.jrcy01.exodus.net
> [209.1.169.209]
> 
>  10   135 ms   130 ms   127 ms  bbr01-p1-3.jrcy01.exodus.net
> [216.32.173.122]
>  11   130 ms   135 ms   131 ms  bbr01-p5-0.hrnd01.exodus.net
> [209.185.249.214]
>  12   140 ms   124 ms   134 ms  bbr02-p6-0.hrnd01.exodus.net 
> [209.185.249.2]
>  13   220 ms   204 ms   205 ms  bbr02-p0-1.irvn01.exodus.net 
> [209.185.9.206]
>  14   205 ms   204 ms   205 ms  dcr02-p0-0-0.irvn01.exodus.net
> [209.185.249.9]
>  15   210 ms   205 ms     *     rsm-ir-b-vl920.lan.exodus.net
> [209.67.128.102]
>  16   269 ms     *        *     zope.org [209.67.167.110]
>  17   204 ms   205 ms     *     zope.org [209.67.167.110]
>  18     *      205 ms   204 ms  zope.org [209.67.167.110]
> 
> Is there a problem with an exodus.net router.  Is exodus.net
> the zope ultimate provider?
> 
> This is the third time this week.
> 

Bob(o),

I don't want to be biased here, because Scott does a great job for us
and I might get the feeling of backing him despite hard facts about his
provider (Exodus).  I do have a bit of background in this, because I
used to own an ISP.

We have had occasional problems getting responsiveness from codeit, but
nowhere near as bad as it used to be when we had our old connection.
The old connection was slow or dead all the time, day or night.  With
Erols, our new provider, maybe two or three times a week, at about
2-3pm, for a few minutes, the response is slow.  I write this off as bad
Internet weather.  Our old provider, however, just plain stank, and so
did their provider, Cable and Wireless (cw.net).

The thing about traceroutes is they're deceptive.  From your perspective
it looks like it's Exodus's fault, but where, really, is that packet
getting dropped?  Remember, traceroute's use ICMP, and ICMP 'dropped
packet' datagrams are not generated for other ICMP datagrams.  So it may
look like an exodus router is unreachable or is dropping packets, when
it was really a Cable and Wireless router that dropped the ICMP *reply*.
There is really quite a bit of action going on, and no way to tell where
the packes are really getting dropped.  Sometimes, a router may seem
100% reachable when you ping (traceroute) it, but if you ping or
traceroute *through* it, the chances or a packet getting droped increase
alot.  This makes it look like the problem is beyond that router, when
it is really that router.  If we ran a traceroute to *you*, I wouldn't
doubt it if it looked exactly the opposite situation.  So who's fault is
it?  It could be many things.  I don't want to get into the problems
with overload at NAP exchange points, and then there's the political
issues: Provider X is sending much more packets over Y's network than Y
is over X's, so Y throttles back their router to toss the difference and
'level' the 'field'.

I do know these things:

1) Our old provider went through Cable and Wireless, and the access
sucked *all the time*.

2) We now use Erols, and the access is good, *most of the time*.

3) Every and all exchange points (NAPs) on the Internet are 100%+
capacity during peak hours.

4) Traceroutes are deceptive for all measurements, except unidirectional
hop counts (packets usualy never take the same way back because of 'hot
potato' routing)

5) Only hosting networks who can afford multiple large pipes (45Mb+) to
dispersed providers can boast near perfect availability all the time.
Codeit has a 10Mb connection to Exodus.  The provider I ran had two T1s
(3Mb).

The ** are not meant to be emphasized, but to be mark off as qualifiers,
nothing is ever true all the time on the Internet.

Do you dial in?  If so, perhaps you can put some time into helping us
reasearch this.  Often, if you call up a dial up provider and say "hey,
I need a really solid connection to this one site, can I just try you
out for a day to see if your better than what I got"?  They'll sign you
up and let you cancel the next day without charging you.  Make a deal.
Make sure they don't use the same NSP (National Service Provider, ie
Cable and Wireless) as you do.  Now, try the access from the other
provider.  Compare the ping times and traceroutes and see what comes up.

-Michel

> -bobo connor  bob@rocnet.com
> 
> 
> 
> _______________________________________________
> Zope maillist  -  Zope@zope.org
> http://www.zope.org/mailman/listinfo/zope
> 
> (For developer-specific issues, use the companion list,
> zope-dev@zope.org - http://www.zope.org/mailman/listinfo/zope-dev )
>