flowsim

A.k.a. FlowSimulator A.k.a. FlowsImproved

An generic event-driven flow-level network simulator for fast prototype implementation. The goal is to evaluate rate control transport schemes for data center networks.

Topologies:

• FatTree
• SpineLeaf

Traffic arrival models:

• Poisson
• On/Off

Load balancing:

• ECMP
• Hedera
• Conga

How to run the code?

1. "cd FlowSim/TestHuawei"
2. "python Runner.py [-r routing_scheme]"
   The routing_scheme currently can be "ECMP" or "LB" or "Qlearning". E.g., "python Runner.py -r Qlearning".

Where to implement Qlearning algorithm?

Please open FlowSim/Routing/Qlearning_SpineLeaf.py, and check function "def select_action(self, state)".

a brief introduction to the code:

• Routing
  • ECMP_SpineLeaf.py: Code of the ECMP routing scheme on SpineLeaf topology.
  • LB_SpineLeaf.py: Code of one load balancing (LB) scheme on SpineLeaf topology.
  • Qlearning_SpineLeaf.py: Code of the Qlearning routing scheme on SpineLeaf topology.
• Topology
  • SpineLeaf.py: the code of spine-leaf topology.
• Src
  • Link.py, Node.py: Code of links and nodes (servers and switches) of data center topologies, respectively.
  • Flow.py, Coflow.py: Code of flow and coflow, respectively.
  • FlowScheduler.py: The code to do flow scheduling of all input flows with varying start times and flow sizes.
  • Topology.py: The code of a class called "Topology" which defines a general data topology. All the specific topologies in Topology directory should be a subclass of "Topology" class.
  • Routing.py: The code of a class called "Routing" which defines a general routing scheme. All the specific routing schemes in Routing directory should be a subclass of "Routing" class.
  • Simulator.py: The code of simulator.
• TestHuawei
  • Input: Input data.
  • Output: Output data.
    • LogInfo: trace of state updates.
  • TestFlowScheduler.py: the code todo input and output.
  • TestSimulator.py: the code to run simulator.py.
  • Runner.py: the script to run simulations with different options.