|
|
|
|
|
by andrewvc
5953 days ago
|
|
What problems does this solve? Are internet core routers currently running near peak, or is this just getting ready for what traffic patterns will look like in 2 - 3 years? Will this lower transit costs? Will it lower costs for end-users? All of the news coverage I've heard is that 'this thing makes the net faster', but no opinions from the people who will actually work with these things. |
|
I glanced at the released documentation a little. The thing to remember is that this is a multi-cabinet router for _really_ big telcos. It scales up to 1152 slots. Since a single shelf system is 16 slots. 1152 slots is 72 shelfs. Judging from the picture a rack can hold 2 shelves. So that's 36 racks for one router to do 322tbps.
In the telco world it's all about density per-rack. Colo space is expensive and so is cooling and power. How much switching/routing capacity can you get in one rack? Well it looks like Cisco now has about 9tbps per-rack. Which is not bad but it puts things a bit more in perspective.
Cisco also tends to measure their bandwidth a bit differently than us mortals. They take full-duplex line-rate and double it. I've seen them do this on previous products so I'm only assuming they've done this here.
So what we really end up with is a router that does 4.5tbps full-duplex per-rack. Still not bad. But not as mind blowing a number as 322tbps. And not so far away from the competition either.