NOTES.md

Optimization Notes

Because it's a complex environment that knits together multiple runtimes + sophisticated Python code, the fcs-etl-application container can have some painfully long run times, and it's also a bear to push to DockerHub because it generates a lot of layers.

Baseline Performance

• Docker runtime:
  • Version 18.02.0-ce, build fc4de44
  • CPUs: 4
  • RAM: 2 GB
• Host Specs:
  • OS: MacOS X 10.12.6
  • Model: MacBookPro11,5
  • Processor
    • Name: Intel Core i7
    • Speed: 2.5 GHz
    • Total Cores: 4
  • Memory
    • 2x 8 GB DDR3 RAM 1600 Mhz
• Network:
  • 365 Mbps down / 380.8 Mbps up

$ time docker build --no-cache -t sd2e/fcs:optimize .
Successfully built 70659799d467
Successfully tagged fcs:dev

real    7m21.246s
user    0m0.560s
sys 0m0.359s

$ time docker push sd2e/fcs:optimize

real    1m43.977s
user    0m0.219s
sys 0m0.116s

Order of Operations

1. octave repo added before the update/install.
2. octave needs to be installed before TASBE
3. Numpy and SciPy need to be installed via apt
4. TASBE should be updated as frequently as the python code