Args:
- G (networkx Graph): Graph with 2^k nodes, with k \in \mathbb{N}. Ring structure
- D (np.ndarray): Demand matrix for every node pair in G. Demand given in % of total traffix
Find a "good" algorithm, in terms of run-time (efficiency) and correctness (near-optimal). For the following objective:
- Each node has at most (log_2 |V| + 1) edges.
- undirected Graph
- Ring structure needs to stay
| Name | Status |
|---|---|
| BS_CHORD | ✔️ |
| BS_binarySwap | ✔️ |
| BS_halving | ✅ |
| ILP | ✅ |
| BS_demandAware | ✅ |
| BS_bothwayRecursive | in progress |
| BS_bothwayPhase | in progress |
BS := Binary Search
:heavy_check_mark: := ready and tested
✅ := needs optimization
- uniform Distribution
- Zipfian Distribution
- Random
- Single Pair's heavy traffic
Using the Gurobi optimizer, I want to construct an LP that constructs the optimal network graph from a Demand matrix.
In this subchapter I want to keep track of some issues I encoutnered and how to solve them.
In this subchapter I keep track of ideas for future work.