Research Plan

1. Simulation

Move simulation to NS. This has several benefits:

• Flexible simulation environment: easy to configure and simulate various topologies and group sizes.
• Allows direct comparison and verification of results from other researchers using NS.
• Can use "reference" topologies, as described in the RM mailing list.

Use Internet topology generation tool (Georgia Tech) to generate topologies for simulation.

Use above simulation infrastructure to collect results on:

• Performance: duplicates, recovery latency
• Router state with various group sizes and topologies
• Router state propagation overhead:
• what is the overhead with various group sizes?
• how often is state updated?
• Adaptability:
• How fast is state updated after membership changes?
• How do membership changes affect performance?

Use simulation to evaluate/refine mechanism for detecting replier failure.

2. Exploration of other Router-assist schemes

Identify other proposed router-assist schemes (RM list, literature).

Systematicaly evaluate other schemes using our simulation infrastructure (NS + Internet topology generation tool) and analysis.

Compare with our scheme and identify strengths/weaknesses of other schemes

1. Performance (e.g., duplicates, latency).
2. Impact on network (state and processing in routers).
3. Incremental deployment.
4. Trade-offs (e.g., trade performance for easier deployment)

3. Characterize router-assist schemes and refine our scheme

Use simulation results to improve/refine our scheme.

Use results and insight gained from other schemes to further refine our scheme by perhaps borowing good ideas.

Define solution space, based on performance and trade-offs, and characterize each examined router-assist scheme accordingly.

4. Implemention

Continue with our implementation and use it as a vehicle to evaluate the complexity of integrating our scheme into the existing network infrastructure.

• finish modifications to mrouted.
• measure overhead in terms of instruction count of all new components.
• deploy a small testbed overlayed over the MBone and run experiments.

Prepare a release package to be distributed to other interested parties. This will encourage the addition of more sites to our testbed and hopefully make it lsrge enough to run some large scale experiments.

5. Reliable multicast protocol

Design a reliable multicast protocol, which will reside at the application layer and uses our services:

• gap-detection
• nack-based
• buffering at the application

Define the protocol API.

Simulate and evaluate the performance of the protocol. Refine protocol based on results.

Implement the protocol and include the code in the release package.

6. Congestion Control

Explore the potential of the proposed services as the basis for and efficient congestion control mechanism:

• by providing faster feedback to the source
• by controlling layer subscription at the routers in a layered scheme

Investigate alternatives for congestion control mechanisms, e.g.,

• TCP friendliness
• Layering

Simulate and evaluate the performance of the various mechanisms.

Select and add the most suitable mechanism to the reliable multicast protocol implementation.