Resilient Overlay Networks

Motivation: A reselient overlay network is an architecture that allows
distributed applications to detect and recover from path outages and periods
 of degraded performance withing serveral seconds.The key design goal in RON
 is to develop techniques to allow end-hosts and applications to
cooperatively
 gain improved reliability and performance from the Internet.

Key Points:
1) Problems with ISP based routing is:
 -Users cannot select routing metrics.
 -Sophisticated routing only within each ISP.
 -Only ISPs assemble measurements.
 -Hop-by-hop model is error-prone

2) The key design goal is to provide for fast failure detection and
recovery, tighter integration with applications and a fine-grained routing
policy so that nodes can make more effective decisions about routing.

2) RON approach attempts to deal with the problem by moving routing control
 towards end systems. Thus, it takes advantage of small scale and bases
 decisions on end-to-end monitoring.

3) Frequently measure all inter-node paths by exchanging routing
information.
Then, route along app-specific best path consistent with routing policy.

4) RONs improve packet delivery reliability. An interesting finding is that
Single-hop indirection works well both for fault recovery and for latency
improvements.

5) The scalability of RON is currently an issue as it scales to about 50
nodes
 but many systems could benefit from such a number as well.

6) RON does provide the possibility of BGP transit policies. It has been
argues that
given the scale of RON (50 nodes) the misbehaving peers can be dealt with at
a
administrative level rather than a technical one, although if it becomes
more scalable
then more efficient ways would be needed to be able to control misbehaviour.

7) The RON architecture was able to improve latency and throughput in
near-outage
 conditions.
Received on Tue Oct 17 2006 - 04:57:44 EDT