Skip to content

Latest commit

 

History

History

prediction

README.md

Prediction Engine

We have implemented two models: ARIMA Model and Recurrent Neural Network

Setup

  • Install requirements using Pipfile
# python install -r Pipfile
  • Setup Prediction Service
# cp prediction.service /etc/systemd/system/
  • Setup KafkaConsumer and KafkaProducer:

In producer.py and consumer.py, enter the IP address of the bootstrap server where Kafka service is running

  • Start the prediction engine
# systemctl daemon-reload
# systemctl start prediction.service
  • Serve the output files using python http server

Firewall

To allow services through firewall iptables setup needs to be done. Use the following commands to add iptables exception (run as root)

iptables -I INPUT -p TCP -s 0.0.0.0/0 --dport 8000 -j ACCEPT # python simple http

Run python http server

# cd outputs
# python -m http.server -b 0.0.0.0

The outputs can be viewed at http://IP_Address:8000

Prediction & Scaling Policy

Prediction policy:

if (average)>0.70*(len(memory_host.keys())*1024):
    status='up'
elif (average)<0.30*(len(memory_host.keys()))*1024:
    status='down'
else:
    status='no'

Scaling Policy:

ARIMA RNN AutoScale
UP NO UP
NO UP UP
DOWN NO NO
NO DOWN NO