Closed sajith closed 1 year ago
make sense to me
The current topology and request json files were all generated by me manually for testing. I'm waiting for the completion of the Kytos-LC integration so we can generate a suite of realistic testing files.
I compared the topology files with some "better" diffing tools (plain old diff is not very helpful) such as difftastic and jsondiff.com. The big difference is that the the topology that broke during the demo had some really high latency values.
... and so on. You get the idea. :-)
Perhaps this will work: https://jsoncompare.com/#!/diff/id=3eb161670c9ab0c9934c33dbfc53d5aa/
Great! Oh, poor TE solver.... So, need to add the realistic upper bound on the latency and lower bound on the bandwidth as part of the validation....
We have topology files in three places:
sdx-controller/samples/topologies
sdx-controller/tests/data
pce/tests/data/topologies
With topology files under sdx-controller/samples/topologies and the sample test request that we've been testing with (namely
test_sample.json
), TESolver fail to find a solution. With the topology files in pce/tests/data/topologies, things work as expected (a solution is found, and it breaks down to links in three domains). This has been quite confusing.I am proposing four steps to avoid future confusion (once the current release target is met):
pce/tests/data/topologies
: This works. In order to avoid confusion, it would be helpful to have only one version around. Let's make sure to use just these topology files when running tests "live" on the test bed.sdx-controller/samples/topologies
: Let us remove this altogether.sdx-controller/tests/data
: The tests in tests/test_Solver.py are now basically unit tests for pce. Let us move the tests and test data topce/tests
.Related to: