GAE - Flask ML Auto Scaling - Load Testing with Locust CLI.

Live Google app engine url:

https://flask-gae-ml-autoscale-273601.uc.r.appspot.com/

Demo link [YouTube]

https://youtu.be/fkMp7T5c2Js

Team Members:

* Niraj Bista
* Stephen Hackman
* Saif Murad
* Richmond Otchere

Machine Learning Approach

  * We adopted the random forest model on a list of features to predict whether cereal is healthy or not. 
  * We didn't ask users to input values for all these values. We randomly generated values for some of the  features.
  * The front end asks users to input values for calories, fiber and sugar in grams for their cereal.
  * The prediction of a healthy cereal is base on the model working on the background to give a predicted value of whether the cereal is a healthy choice or not.

Deploy Appengine App

$ gcloud app deploy

List all app version

$ gcloud app versions list
SERVICE  VERSION.ID       TRAFFIC_SPLIT  LAST_DEPLOYED              SERVING_STATUS
default  20200403t133612  0.00           2020-04-03T13:43:20-04:00  STOPPED
default  20200403t142656  1.00           2020-04-03T14:29:37-04:00  SERVING

Browse app

$ gcloud app browse

You can stream logs from the command line by running:

  $ gcloud app logs tail -s default  

Locust Load testing instructions.

Install Locoust load testthrough cli

sudo python3 -m pip install locustio

Run Locust load test

$ locust

app instances list

$ gcloud app instances list
SERVICE  VERSION          ID                                                                        VM_STATUS  DEBUG_MODE
default  20200404t000535  00c61b117c0e358b4f39351e31c7cbbb4a42eeccf5c8e79c67ffadb5bc90c36d16176c28  N/A
default  20200404t000535  00c61b117c1316418d0ddc5f63ba320f4e5f6c33da5f92b06ee4448c1b1620aa73d82e0c  N/A
default  20200404t000535  00c61b117c23d72c2dc77c392bb5d7f0524a288cba7c32483c6b10fd7a12dace6a89e525  N/A
default  20200404t000535  00c61b117c3da9a1cae3f6fad5c7d9c7e964d1d0d3a7c4cf4e82209f217db46e0e1b1466  N/A
default  20200404t000535  00c61b117c4630017427a019c195e79fddc6689690148d9541ecb402543f41b3265e15    N/A
default  20200404t000535  00c61b117c7a46d0f08f90e179410464eb31977df1a3c34c4e172bf2d7cb59c6c994d458  N/A
default  20200404t000535  00c61b117ce32c112aaf8289288201a9845bea7e76c0f6c742a9cb737e124c4fb2904fc1  N/A

Run Locoust in cli

$ locust -f locustfile.py --no-web --host https://flask-gae-ml-autoscale-273601.uc.r.appspot.com -c 10 -r 1

Instance Autoscaled graph.

Cereal healty or not healthy predictions.