What horrors lurk in the future: Networks without sysadmins
We chat to Cisco's Alvaro Retana on routing, in and out of the data center
Interview At Cisco Live Sydney, your humble hack at Vulture South got to spend time with distinguished engineer and CCIE number 1609 Alvaro Retana to talk about the future of routing, networks that can run themselves, and the internet at large.
The Register: What are the problems we're trying to solve in the world of routing today … what are the things people bring to you and say “this isn't working, how can we fix it?”
Retana: Scalability. Mobility – not just mobility in terms of having more phones, that's part of scalability. Ultra-mobility in terms of things like auto-forming networks. Some applications are very specific, like the military for example, where networks are forming every second.
The Register: The military wants battlefield networks to form themselves around the soldier, and HPC clusters that serve information.
Retana: Right. A self-forming network leads not only to military applications, but more generally. Networks are getting complex, and so you want to simplify them, which makes them more autonomous, with more intelligent auto-configuration.
So the topic of mobility for me goes into the complexity of the network, and how can we make better auto-configuring, auto-optimising, and auto-healing networks.
And the other thing is security, of course. Everyone's worried “is this my route? Is my traffic going in the right direction? Is someone else looking at my traffic?”
The Register: Why are we still asking “please make routing secure, and stop traffic getting black-holed”?
Retana: When the internet was small, nobody thought it was going to be big. That was one part. The other part is that really good security and operations don't necessarily see each other in the eye. If I do really good security, operations will be harder.
And people are in this to make money – if you're a service provider you're in this to make money.
The Register: In 1986, the security was guaranteed because if I am at UCLA and want to route to Berkeley, I pick up the phone and arrange it.
Retana: Right – it used to be a lot simpler. Now it's more complex, and the mechanisms, which are still being worked out, are not necessarily those that will give me incremental benefits.
If I [create a new route], what's in it for me? Even if both of us do it, what about the rest of the world? Some of the mechanisms require everyone to get involved. We can't say: “Next Friday at 12pm, we're going to do this.” We need to figure out long-term strategies [for adding routes].
It's going to be a slow process. For security, it's going to be a slow process as well. That combination of security, scalability and autonomy – what's it going to look like 20 years from now, and how are we going to get there?
It's not going to be "I'm going to come up with a new BGP," and that's it. Because we're not going to just switch everyone to something else. My opinion is that we have to look, see where we want to be, and start evolving, little by little.
The Register: We should write the requirements document before we build something?
Retana: That would be a fabulous idea. We're all used to “it just works” – plug in another machine, plug in another autonomous system, and it works. You don't have to ask permission, you don't have to tell anyone you're going to do this, it just works.
But somehow, if we want to evolve the network, we need to somehow evolve it with a purpose, and that is not easy.
The Register: Because even if you say “we're going to have a three-month conference about the future of the internet, and look for a consensus” … consensus won't happen.
Retana: Even if we come up with something that 80 per cent of everyone agrees on…
The Register: There will be enough people 'hmmming' over it…
Retana: Yes, enough people 'hmmming'! Even if we get to that [consensus], there will be an evolution. We need to take five, ten, twenty different steps. It's a slow process, working that over the next ten or twenty years.
Maybe twenty years from now, someone else will have the same discussion [about improving routing], just passing the buck to someone else, which is something we hope will not happen.