About The Project

This project is based on OMNeT++'s routing example, using cTopology to calc the shortest paths.

This project aims to implement different congestion algorithms.

Notice

• No packet loss
• No handling of package disorder
• Socket only supports Simplex Communication

Why not use NS3/INET?

Too complicated for testing congestion control algorithms. As a simulation, the code doesn't need to be implemented exactly like that in the linux kernel, and a faithful representation of the network layers is also unnecessary.

We don't even need ip or mac addresses. Things should be kept as simple as possible.

Roadmap

• Packet type: DATA and ACK
• label packet ECN and cut the window into half
• AIMD algorithm should should split the bandwidth equally
• DCTCP
• Aggregate packets one level forward and backward
• Aggregate packets for more levels
• aggregatorIndex hash collision, then resend packets
• FatTree(k), k means pod number
• job deployment policy
• resend policy when diorders happen