SearchNetworking chats with Roger Geerts, Juniper Networks' director of systems engineering for Australia and New Zealand, about life, the universe and very fast networks.
Ian Yates: What are your thoughts on 10G Ethernet over copper? Now that the 10GBase-T standards are ratified do you think we.ll see a rapid uptake?
Roger Geerts: There's still a bit to sort of out there. Some people are still struggling to run gigabit over copper. CAT5 installations are pretty good, but I think to extend that to 10GigE . only over a short distance perhaps. What we.re seeing is definitely with 10GigE it.s all optical.
Yates: I've had it put to me that people would like to use 10GBase-T in the data centre. So if all you.re doing is linking a storage box to a server or a switch in between half a dozen servers and you.re not going to need more than two or three metres, is that where you think you might first see copper working?
Geerts: Yeah I would suspect so. That's where gigabit copper first started out as well. You expect 10G would be in a similar vein. Then if it's only short runs, even if you need to change to copper, it's not a great cost. I certainly don't see 10G to the desktop in the very near future.
Yates: Which means fibre optic is still going to dominate at 10G speeds. Is it still a lot more expensive than copper?
Geerts: It's just a more costly proposition to run fibre, to get the right people to terminate the ends. We found that just putting our labs together. When you do it on a large scale, obviously it's a fairly costly exercise. But once you require that type of bandwidth it becomes necessary, really.
Yates: With pre-made fibre patch leads and some smart kit for field installs I would have thought fibre terminations had become more cost effective and more reliable surely?
Geerts: I would agree with that, but I think that at the same time the copper installations have also dropped, right, so it's a two way street. There's still a difference there.
Yates: Are the carriers already braying for more than 10G?
Geerts: Well, we've already released a 40G switch based on SONET as opposed to Ethernet. The jury is still out, as you probably know, on whether the next step for the internet is 40G or 100G. I don't think their argument is over yet. I think Ethernet has just been one of those things where it's basically going up in factors of 10. It has been good, but whether the increment from 10G to 100G is too far for people to jump, that's the question.
There's a couple of things to consider there obviously, just the backplanes for some of the devices that are already out there. It's hard to get too many useful 100G interfaces into the existing products and obviously 40G is pretty well understood already with SONET OC-768. It's still I think early days on that, but we know that we've got some customers who do want 40G SONET already and if they're an Ethernet customer, then we do have very good methods of doing link-bonding. That's where there is a bonding of multiple 10G links together to achieve those sorts of rates.
The other problem is that if you've got to cross transport layers, then you've got the obvious question of whether your transport infrastructure can support 40G and what's the cost of upgrading that? It might be substantially more cost effective to aggregate multiple 10G links.