I use a CRON job and shell script to save logs from each night for a weeks'
worth, keep Sunday for the last 5 weeks, and the 1st & 15th of each month
for the past 6 months. It's just interesting to me to see the changes that
occur from time to time (usually things get "closer" and have less hops over
time, as ISPs peer at more places).
Anyway, the best thing to do is look up their ip space on one of the looking
bgp glasses and find out their ASN and how many different directly-connected
paths they're announcing out (how many ISPs they're connected to).
route-server.cerf.net shows the best 3
route-views.oregon-ix.net shows the best 22
Use 'show ip bgp ' to find their ASN and such info. Here's an
example with one of our netblocks:
route-server>sh ip bgp 63.107.123.0
BGP routing table entry for 63.107.123.0/24, version 294979
Paths: (3 available, best #3)
Advertised to non peer-group peers:
216.128.28.34
1740 1239 18506
134.24.88.55 from 134.24.88.55 (134.24.127.27)
Origin IGP, metric 20, localpref 100, valid, external
1740 701 18506
192.157.69.5 from 192.157.69.5 (134.24.127.201)
Origin IGP, metric 20, localpref 100, valid, external
1740 701 18506
198.32.176.25 from 198.32.176.25 (134.24.127.39)
Origin IGP, metric 20, localpref 100, valid, external, best
Our ASN is 18506, and 701 is UUNET and 1239 is Sprint, both of which are our
directly connected ISPs. If you compare on oregon-ix, you'll see that we're
only announced out those two ISPs in the 22 best routes it keeps. I would
expect a good co-location place to be connected 3-5 of the top ISPs, but
that's just a guess off the top of my head. I think I've seen a really
high-end co-lo that boasted connections to the top 12 ISPs.
Anyway, my guess is that your corp. ISP is either manipulating the BGP
announcements (prepending or other means) and therefore influencing a longer
(but perhaps bigger/faster connection) route, whereas they're taking the
shortest hop announcement from the co-lo.
Again, look up your corp. netblock as well, and see how it's being
announced.
It could also be that the co-lo isn't taking full routes from all of their
ISPs, which would result in the sub-optimal routing you fear.
Just looking up the netblock of your MX, it does look like your ISP is
prepending to influence routing far away from some of their links (Telstra
AS16779) and their AT&T (AS7018) link, somewhat away from their UUNET
(AS701) link and through the Sprint link (AS1239 has no prepending):
6395 7018 6993 6993 6993 6993
216.140.8.63 from 216.140.8.63 (216.140.8.63)
Origin IGP, metric 483, localpref 100, valid, external
Community: 6395:10
8297 6453 1239 6993
195.219.96.239 from 195.219.96.239 (195.219.96.239)
Origin IGP, localpref 100, valid, external
1221 16779 3549 6993 6993 6993 6993 6993 6993 6993
203.62.248.4 from 203.62.248.4 (203.62.248.4)
Origin IGP, localpref 100, valid, external
6066 701 6993 6993
204.29.239.1 from 204.29.239.1 (150.225.14.1)
Origin IGP, localpref 100, valid, external
I would guess that your ISP InterNAP knows what they're doing with all these
prepends and is purposely influencing return traffic toward their bigger
pipes.
--
Jason Roysdon, CCNP+Security/CCDP, MCSE, CNA, Network+, A+
List email: [EMAIL PROTECTED]
Homepage: http://jason.artoo.net/
wrote in message
[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> I am hoping someone could provide me some experienced perspective for the
> following situation:
>
> We utilize a somewhat 'new on the block' co-location facility, and while
> they otherwise provide fantastic service I have some questions about the
> routing performance.
> Over the past few weeks, I have noticed a degradation of service on our
> colocator-provided connection. (significant latency, and loss of packets)
> As a result, I have been tracerouting our corporate offices from our
> co-location facility (only 30 miles away) and it takes anywhere from 13 to
> 16 hops to reach it's destination. I have been doing this on a
> semi-scientific basis (whenever I remember) and the results are usually
the
> same, but closer to 16 hops than 13. When I traceroute from our
corporate
> offices to our co-location facility the results are usually 6 to seven
hops
> using the same semi-scientific methodology as stated above.
>
> My concerns are that end-user experience are being affected by apparent
> sub-optimal routing.
>
> The question I ask of the Grand-Master BGP geniuses is: do I have a valid
> complaint regarding sub-optimal routing from our co-locator?
>
> Thanks!
> Matthew
> FAQ, list archives, and subscription info:
http://www.groupstudy.com/list/cisco.html
> Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=588&t=495
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
