Loose Ends
printFlowOriginNode
in bothAnalyzeMFCLP
andAnalyzeSimpleLPFlow
is targeted at a single traffic mode, respectively MAWP/MIWP and ATA. However, what is printed as flow from that origin node does not make sense / fails for other traffic modes. This should be made generic. E.g.int n = ((int) Math.sqrt(tuples.size())) + 1;
inAnalyzeSimpeLPFlow.printFlowOriginNode
only works for ATA.FileBidirGraph
is old and should be improved to handle e.g. different node weights.- Make
Configuration
more intelligent. - Within
Graph
, makeshortestPathLen[]
completely handled by itself, in a similar fashion as done with the node weights. - Consider renaming
Graph
toBiDirGraph
, and adding the ability to create directed graphs (e.g. addDirGraph
), abstractingGraph
further.
Testing
- Achieve near-perfect test coverage.
- Write an end-to-end tests that run both modes in their entirety, automatically testing if the resulting objective value is correct for the given circumstances.
- Check platform compatibility with common versions of Windows, Mac OSX, and Linux.
Graph Variety
- Hypercube
- BCube
- DCell
- Dragonfly
- FlattenedButterfly
- VL2
- ... others
New Features
- Add mode for static topology bounds calculator (
ch.ethz.topobench.graph.utility.BoundsCalculator
). - Random traffic mode?
Too Ambitious
- Multi-lingual support for the command line