vm-autoscaling

Cloud management system using the libvirt API
Detailed problem statement: here

Monitoring

• Continuous monitoring of CPU utilisation by server VMs
• usage % = 100 * (cpu_time 2 - cpu_time 1) / (time 2 - time 1)
• Overload if cpu usage above a threshold for some number of consecutive iterations

Autoscaling

• Scale up from N to N+1 server vms using libvirt api
• Notify the client to use the new vm to mitigate overload

Setting up

• Install virt-manager, libvirtd, qemu and virsh
• Use virt-manager to create two VMs
• Configure network using steps given in config file
• Add server code files to respective vms and configure them to run on boot up
• Start vm1 and run python monitoring.py and python client.py
• Monitoring will show cpu usage in real time
• Change load using 'C' key in the terminal where client.py is running to see autoscaling in action