Makespan Optimisation in Cloudlet Scheduling with Improved DQN Algorithm in Cloud Computing

Algorithm 1

Pseudocode of the proposed reward function.

	Input: //The current state of VMs
	Input: //VMs speeds in MIPS
	Input: //The cloudlet to be scheduled in MIPS
	Input: //The selected virtual machine
	Output: //The reward
	Function rewardCurrentState():
	//Get the number of VMs
	//List of makespans initialized with zeros with a length of
	forIterate over VMs
	do
	//Make a copy of the current VMs state
	//Calculate the run time of the given task in
	//Add task’s execution time to
	//Calculate the makespan of the copied state
	end
	//Sort VMs executions times in descending order
	//Get the unique length of makespans to exclude makespans with the same value
	//Create an empty dictionary to store each VM rank.
	//The highest rank is given to a VM
	for//Iterate over VMs
	do
	//makespan’s value from sorted makespans
	not in //Check the makespan’s value if it is not in the rank dictionary
	then
	//Assign the rank’s start value to
	//Decrement rank’s start value
	end
	end
	//List of scores initialized with zeros with a length of
	for//Iterate over makespans
	do
	//Get
	//Score the VM based on makespan’s rank
	end
	return//Select the score based on the given action
	End Function