Aims of the scheduler prototype include the following:
- Same APIs as a production scheduler
- Easy to add new algorithms
- Able to replay production logs
- Support unit tests for core scheduler and for database (Redis) integration
(NUM_NODES=3
NUM_WORKERS_PER_NODE=1
for trace in $(ls traces/test/*.json); do
python replaytrace.py $NUM_NODES $NUM_WORKERS_PER_NODE 0.001 .0001 trivial fifo true $trace;
done)
- install anaconda
- pip install ortools numpy