The questions stand

Mailman List Mirror (Read Only)
1

That happened here in Columbus with a provider whom I won't name since they
are very imfamous for threatening legal action at the drop of a hat. The
fact of the matter is that they claim to be carrying some ungodly portion
of the internet traffic on their "private network" thereby providing faster
connections for their customers. Well, this is America folks. Lets face
it. If there is opportunity for someone to make a quick buck, they will.
These guys ended up with DS1's to the big-4, put in an OC3 circuit
provisioned for a single DS3 to a local ISP that was a partner in the
project and at the time had a 10MB partial DS3 circuit to one of the big-4
and advertised their network as "OC3 Fiber Optics based" and "Faster than a
Teier One".

Lets see how they are performing today, shall we?

[somebox]:~# traceroute customer.of.blah-12345-we-are-a-nap.net
traceroute: Warning: Multiple interfaces found; using x.x.x.x @ eth0
traceroute to customer.of.blah-12345-we-are-a-nap.net (y.y.y.y), 30 hops
max, 40 byte packets
1 (default gateway) (x.x.x.1) 1.069 ms 0.454 ms 0.434 ms
2 (Tier One provider) (x.x.x.21) 1.456 ms 1.295 ms 1.268 ms
3 aads-nap.att.net (198.32.130.60) 42.083 ms 55.715 ms 51.440 ms
4 br2-a3110s2.cgcil.ip.att.net (192.205.31.122) 50.639 ms 54.973 ms
51.153 ms
5 br1-h30.sl9mo.ip.att.net (12.127.15.210) 59.831 ms 51.588 ms 44.678 ms
6 ar1-a300s1.sl9mo.ip.att.net (12.127.4.5) 42.652 ms 33.012 ms 35.015 ms
7 (y.y.y.y) 50.081 ms 42.371 ms 43.084 ms
8 customer.demarc.blah-12345-we-are-a-nap.net (y.y.y.y) 69.753 ms
74.542 ms OSPF? (y.y.y.z) 89.058 ms
9 router.customer.of.blah-12345-we-are-a-nap.net (y.y.y.y) 92.875 ms
87.486 ms 95.569 ms
10 www.customer.of.blah-12345-we-are-a-nap.net (y.y.y.y) 86.939 ms
58.300 ms 64.930 ms

Wow. Now, that's some lightning speed. <sic>

And NOW, lets look at someone with a real Tier One provider:

[somebox]:~# traceroute www.mcs.net
traceroute: Warning: Multiple interfaces found; using x.x.x.x @ eth0
traceroute to www.mcs.net (205.164.8.12), 30 hops max, 40 byte packets
1 (Default Gateway) (x.x.x.1) 0.495 ms 0.415 ms 0.385 ms
2 (Tier One provider) (x.x.x.21) 2.424 ms 1.630 ms 1.529 ms
3 aads.good.net (198.32.130.49) 15.131 ms 13.261 ms 13.930 ms
4 mcsnet.chicago.good.net (207.98.189.130) 14.775 ms 15.296 ms 16.193 ms
5 www.mcs.net (205.164.8.12) 14.592 ms 14.750 ms 14.504 ms
[somebox]:~#

The Tier one's who want the business will maintain adequate bandwidth for
their downstreams. Time and time again, people who haven't been interested
in investing in the facilities to provide Tier One access have cut corners
to line their pockets at the expence of their downstreams who far too often
have no clue and accept the poor performance of their circuit as "normal."

BTW: Karl, lightning fast as always to your site.

2

Yes, John, I've been reading NANOG long enough to be aware that there
are hundreds of ways to do it wrong.

There are about a dozen ways to do it right, too.

There are only about 40 ISPs in this area, OC-3 is probably overkill...
not to mention whether it's actually _available_ reasonably, but you'll
notice that's where I started.

Cheers,
-- jra

3

Lots of this has to do with your constituents.

If you're a dial-only provider, you can afford not to care (much) about
latency. You ALWAYS care about packet loss, but latency can go to the dogs
and your customers won't know much about it.

If you serve DS1 customers, its a different ballgame. Now you HAVE TO
provide good network infrastructure, or people consider your plant to
"suck".

Then again, my background and engineering efforts have been geared towards
the Fortune-50 marketplace and OLTP environments where ANY packet loss or
significant hysteresis in the transport times is considered unacceptable.

We build 'em "a bit different" around here.