On 03/14/2013 04:31 PM, NAPIERALA, MARIA H wrote:
Dave,
What are the intended use cases for i2rs?
I think we're all here to come up with those. The charter lists the
following:
o Interactions with the Routing Information Base (RIB). Allowing read
and write access to the RIB, but no direct access to the Forwarding
Information Base (FIB).
o Control and analysis of the operation of the Border Gateway Protocol
(BGP) including the setting and activation of policies related to
the protocol.
o Control, optimization, and choice of traffic exit points from
networks based on more information than provided by the dynamic
control plane.
o Distributed reaction to network-based attacks through rapid
modification of the control plane behavior to reroute traffic for
one destination while leaving standard mechanisms (filters, metrics,
and policy) in place for other routes.
o Service layer routing to improve on existing hub-and-spoke traffic.
o The ability to extract information about topology from the network.
Injection and creation of topology will not be considered as an
initial work item.
Some use cases I can come up with for such a system include:
* Polling the state of the network to determine the exact path the RIB
believes a packet should take, edge to edge.
* Traffic engineering by identifying an overloaded link and overriding
the next hop for some subset of traffic. Or even more generalized
traffic engineering: an autonomous controller that uses analytics to
determine what endpoints currently need the most bandwidth and
automatically allocate bandwidth to those applications, whether across
the backbone or across the datacenter.
* Setting up policy routing for traffic matching a particular
application to optimize performance; for example, pulling best-effort
traffic to a longer path when load on a link is causing jitter in real
time media.
* Finding routes that are not active in the RIB, due to higher
administrative distance, and overriding that decision. For example,
imagine a situation where you are running OSPF and RIP on a router.
Among other things, you receive a default route and another network, say
192.0.2.0/24, via both protocols. You want to believe the RIP route for
192.0.2.0/24 and OSPF for default. Using I2RS, you set a lower distance
for only the RIP route. (There's lots of mechanics with this one,
including I2RS telling you when the table has changed, so you can
intercede again if you have to.)
* Attack mitigation by routing traffic to a victim IP to either the bit
bucket or a middlebox that can filter bad traffic from good.
* Modifying preference on BGP routes in real time to balance traffic
ratios, prefer financially less expensive links, manipulate your 95th
percentile traffic rate, etc.
* Tracking a route oscillation or similar event as it happens to help
determine the source.
I'm sure there are plenty more...
_______________________________________________
i2rs mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/i2rs
