[j-nsp] Advice on a 100Gbps+ environment

[Mark Tinka]

On Tuesday, July 02, 2013 05:09:44 PM Julien Goodwin wrote:

> Other than the amount of CPU capacity on the MX80 (which
> has been done to death here) do you really have RE's
> fail often enough to be a problem?

When we had chassis-based core switches, we slowly moved 
away from dual control planes to a single one, given the 
work being done (Layer 2 switching only) and the reliability 
that has gone into the platforms over time.

So much so that now, our core switching happens on 1U 
chassis that have multi-rate Ethernet ports, i.e., Cisco 
4500-X or Juniper EX4550. The only time we'll ever see 
chassis-based core switches again is if we're doing some 
kind of data centre core aggregation in a specific PoP, if 
ever, since routers are now either small and powerful or big 
and powerful enough that the days of needing 300 to serve 
your edge in one PoP are long gone.

On the routing side, however, we've had cases where routers 
have been rock solid for years, and others have had a bad 
batch of RE's, nearly all at the same time.

Moreover, for software upgrades (even in planned windows), 
it's always good to have a secondary RE to which you can 
dump traffic until you complete an upgrade cycle on the 
other.

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/juniper-nsp/attachments/20130702/d4ebb25b/attachment.sig>