SYSU-Workflow-Lab / Burst-Aware-Autoscaling

Burst-aware Autoscaler

Source code of paper Multi-Level ML Based Burst-Aware Autoscaling for SLO Assurance and Cost Efficiency submit to TSC.

The overview of the proposal shown as a MAPE loop: (a) Cloud-based system that provides Monitoring and Execution; (b) Burst-aware autoscaler for Analysis and Planning includes Workloads Prediction, Burst Detection & Handling, Resource Estimation, and Estimation Enhancement.

Require

python  3.6.0
torch  1.8.2
gym  0.22
pandas  1.4
numpy  1.22
sklearn  1.2

Running

Load generate using k8

cd k6
bash background.sh 

Autoscale

python -u main.py