Reading for Lecture Oct 28

Resilient Overlay Networks
RON is an architecture that allows distributed Internet applications to detect and recover from path outages and periods of degraded performance within several seconds. It is an application layer overlay on top of the existing Internet routing substrate.

The main goal of RON is to enable a group of nodes to communicate with each other in the face of problems with the underlying Internet paths connecting them. It aggressively probes and monitors the paths connecting its nodes to detect link failures and find the best paths. Information exchanged between nodes is metrics including latency, packet loss rate, and available throughput.

RON takes advantages of underlying Internet path redundancy on time-scales of a few seconds, reacting responsively to path outages and performance failure. It is closely tied to the application using it and it more readily integrates application specific path metrics and path selection policies. It demonstrate fast recovery from failure and improved latency and loss-rate even over short-time scale.

For each link, the latency estimate is updated as: lat ← α . lat + (1-α). new_sample. The overall latency for a path is the sum of the individual virtual link latencies: lat_path = Σlat_link. The loss rate metric is multiplicative on a path. The score = (1.5^0.2)/(rtt*p^0.5) is used to prevent large route oscillations from single packet losses.

Active network vision and reality lessons from a capsule-based system

Active networks help deploying new services rapidly but enabling the programmability at each router in the networks. Some active networks provide extensibility to individual devices; this kind of systems is well suited to the task of imposing policy or functionality at particular network location like boundary devices. Another style of system provides programmability across multiple odes for control tasks rather than new data transfer service.

ANTS does not a priori restrict who can program which nodes. It aims to allow each user to construct new data transfer services across the wide-area. ANTS is based on an aggressive “capsule” approach in which code is associated with packets and run at selected IP routers that are extensible.

In ANTS, applications obtain customized network services by sending and receiving special types of packets called capsules via a programmable router referred to as an active node. The field type to determine the service, hence, which code to run. If a node does not have that code, it requests from previous node. The ANTS places of limit of 16KB on the code each service to limit the impact of code transfers on the network.

Does processing capsule affect the performance at each node? With 16Kb code, what services can be supported? Do functionalities also depends on infrastructure at each node? Does active network violate end-to-end argument?