[101516] in North American Network Operators' Group
Re: can the memory technology save the routing table size scalability problem?
daemon@ATHENA.MIT.EDU (Christopher Morrow)
Tue Jan 8 22:52:39 2008
Date: Tue, 8 Jan 2008 22:48:34 -0500
From: "Christopher Morrow" <christopher.morrow@gmail.com>
To: "yangyang. wang" <wyystar@gmail.com>
Cc: nanog@merit.edu
In-Reply-To: <10fbbf5a0801081825n65cc69bdsb8a4e34a7b7cbb@mail.gmail.com>
Errors-To: owner-nanog@merit.edu
On Jan 8, 2008 9:25 PM, yangyang. wang <wyystar@gmail.com> wrote:
> As we known, the DFZ RIB size expand rapidly. It may be resolved via router
> architecture improvement, such as adding memory chips or compressing RIB. or
> via changing routing and addressing scheme, which one will be the long-term
> essential approach?
There are at least 2 problems to be addressed (given that you believe
'too many routes will crush the dfz routers')... both number of routes
and speed/number of updates. So, adding more MEMORY (pick your type
DRAM/SRAM/bah) is only solving one of the problems. Additionally, most
moderm DFZ-placed platforms aren't necessarily 'RAM' limited, some of
the limits exist in hardware forwarding elements (TCAM/CAM/ASIC
systems).
Anyway, Suresh's followup has a decent info as well, you might locate
the RRG and RAM/RAWs working group meeting outputs as well:
(report)
http://tools.ietf.org/html/rfc4984
-Chris
