Internet Core Routing - Ethernet

Alex:

Wrong. None of the vendors wants to help you what so ever. What they want to
do is to have you spend your money with them by claiming that they want to
help you.

Cool, I respect that attitude. At least now I know how to deal with you. The glass is half empty.

> Wouldn't you rather have a reliable, redundant L2 core with isolated
> failures and fast recovery?

No. I hate complicated words and complicated concepts. It reminds me NASA
spending 10M to invent a pen that writes in 0 gravity instead of using a
pencil.

You are the guy that posted about IBGP Nailed Routes. I love that idea. I understand your love of L3 and your design is great; however, as a service provider, you can't always rely on IP. What about Global Ethernet? The NY, Paris, Tokyo, Seoul, Singapore, Frankfurt, Moscow, Bankok, Sydney, Hong Kong, London, Cairo, Los Angeles, Jerusalem, Toronto, Mexico City, Rio (Can I manage this one?) Internet Exchange? MPLS VPLS over Ethernet Anybody? Use your nailed routes trick in the middle (MPLS side) and L2 peering exchanges in each city. Only problem with this is complexity. Do this alone or with few partners and you may be able to pull it off completely L2 without the cost of complexity. Simple like you said. Gotta work on the reliability part though. Test. I'd like to test in Paris, London, Jerusalem, and Toronto. I haven't been to those places yet and I got this cool digital camera. Please don't shoot when you see a Mexican (American) who can't speak Spanish.

> Ethernet is my digital wrapper of choice because of vendor chipsets
> supporting it and quality product.

Yeah, ethernet over what, carrier pigeons?

That will work, but I don't think carrier pigeons are Internet Core unless there has been major pigeon development I'm unaware of. Good question.. 1G (.z and .ad aggregates), 2.5G (POS), 10G (.az), 20G (.ad), 40G (.ad), beyond are more of the realm we are in nowadays huh? I suppose you can use WDM equipment and get an 802.3ad 100G trunk someday, but then again the assumption is that single flows would not be more than 10G :slight_smile:

Bobby

>Wrong. None of the vendors wants to help you what so ever. What they want
>to do is to have you spend your money with them by claiming that they
>want to help you.
Cool, I respect that attitude. At least now I know how to deal with you.
The glass is half empty.

Very simple. Put a good product in front of me and you wont get this
attitude. Tell me that you can do a line-rate filtering when you cant and
be bitch-slapped.

Should you, as a service provide, not want to be up every second day at 4am
because a new super-cool feature that you have deployed broke your entire
network, a dose of healthy skeptisism would do wonders to your nerves, your
customer service nerves and *gasp* customers not disputing charges for
services.

> > Wouldn't you rather have a reliable, redundant L2 core with isolated
> > failures and fast recovery?
>
>No. I hate complicated words and complicated concepts. It reminds me NASA
>spending 10M to invent a pen that writes in 0 gravity instead of using a
>pencil.

You are the guy that posted about IBGP Nailed Routes. I love that idea. I
understand your love of L3 and your design is great; however, as a service
provider, you can't always rely on IP.

You are absolutely correct. One should not rely on IP for everything.

One also probably should not use Starwars Network Management Protocolv2
running over an unprotected network (which uses hubs in public places) to
not just read telemetry data but also issue march commands.

However, using IP as a transport between two encrypting SNMP gateways to do
the same thing would do just fine.

What about Global Ethernet? The NY, Paris, Tokyo, Seoul, Singapore,
Frankfurt, Moscow, Bankok, Sydney, Hong Kong, London, Cairo, Los Angeles,
Jerusalem, Toronto, Mexico City, Rio (Can I manage this one?) Internet
Exchange?

And see a spectacular cascading failure from NY affecting traffic in Cairo.
Brilliant fireworks. I bet New York City's New Year celebration would be
nothing compared to the fireworks at the NOC screens. However, should your
company offer $500,000 per minute outage insurance policy, I am quite
certain that a few of companies I consult for may be interested.

MPLS VPLS over Ethernet Anybody?

Eik.

Use your nailed routes trick in the middle (MPLS side) and L2 peering
exchanges in each city. Only problem with this is complexity.

No, the only problem with this is lack of clues on the designer's side.

Do this alone or with few partners and you may be able to pull it off
completely L2 without the cost of complexity. Simple like you said.

I highly suggest before one jumps into this, he or she should take a
few math and physics classes. It may open even some permanently shut eyes.

Alex

Thus spake "Bob Martinez" <bobmartinezzz@hotmail.com>

You are the guy that posted about IBGP Nailed Routes. I love that idea. I
understand your love of L3 and your design is great; however, as a service
provider, you can't always rely on IP.

As an ISP, your job is to provide IP. If you -- much less your customers --
can't rely on your IP service, you'll soon be toast in the marketplace.

What about Global Ethernet? The NY, Paris, Tokyo, Seoul, Singapore,
Frankfurt, Moscow, Bankok, Sydney, Hong Kong, London, Cairo, Los
Angeles, Jerusalem, Toronto, Mexico City, Rio (Can I manage this one?)
Internet Exchange?

And one errant BPDU can take down your entire network. You're going to need a
bunch of PhD-level guys in your NOC with that design.

Simple like you said. Gotta work on the reliability part though.

You're just now thinking about reliability after you've finalized your design?
Stop, you're killing me...

S