Authors of draft-ietf-spring-ipv6-use-cases,

Section 2.3 SPRING in Data Center:
The current text in the section only describes why some DCs are transitioning 
from IPv4 to IPv6 natively, but not having very convincing points on why 
Segment Routing is needed (even with IPv6 addresses). Most data centers use 
private addresses, even AWS Cloud DCs use IP4.

I think the real use case of segment routing in Data Center (regardless IPv4 or 
IPv6 address space) is

-        There could be very large number of parallel paths among leaf nodes 
(via Spine nodes). It is more efficient for leaf nodes to use designated (a set 
of) Spine nodes as outer destination addresses to avoid being placed on random 
paths (via ECMP) between leaf nodes and spine nodes.  The designated spine 
nodes can easily replace the outer destination address (i.e. the spine node 
address) with either Aggr nodes (going up) or the original leaf addresses 
(going down) .

Here is an old academic paper on why this approach is more effective in DC.

http://delivery.acm.org/10.1145/1600000/1592576/p51-greenberg.pdf?ip=12.111.81.80&id=1592576&acc=PUBLIC&key=5A3314F2D74B117C%2E5A3314F2D74B117C%2E4D4702B0C3E38B35%2E4D4702B0C3E38B35&__acm__=1519340096_d0ed599df20bcdcf17f61ce05325f3f2



Section 2.1 SPRING in the Small Office:

You stated that the IPv6 small office will have multiple egress points. Why? Is 
it because the small office is connected to multiple locations? How is it 
different from today's small office environment?

Don't most small offices have uplinks to one or two PEs?

Linda Dunbar
_______________________________________________
spring mailing list
spring@ietf.org
https://www.ietf.org/mailman/listinfo/spring

Reply via email to