Xiao Liu PhD Thesis.pdf - Faculty of Information and Communication ...

Ant Colony Optimisation
Input: Integrated Task-to-VM list L{ ( Ti , VM j ) | i = 1,2,.., n;
j = 1,2,.., m};
Tasks { T i , Cost i , Time i , DAG{
T i > T i+ 1}};
Virtual Machines { VM j,
Price j , Speed j}.
Output: Optimised Scheduling Plan
//Optimising the overall makespan and cost
//Initialisation of pheromone and other parameters for ACO algorithm
// Package based random generation
1) INITIALISATION(ACO);
2) While (stopping condition is not met)
{
// initialise each ant
for each ant
{
// select heuristics from duration-greedy, cost-greedy and overall-greedy
3) SELECTION (Heuristics);
// build tackling sequence TS based on the input DAG task graphs
4) BUILDING( TS,
DAGs);
}
// construct solutions
5) While (not end of TS)
{
// choose the resource for the next activity based on its earliest start time ,
earliest end time, and the bias of B ij (mapping resource R j to activity a i )
6) CHOOSE ( ai,
ai.
est,
ai.
eet , R{
R
j},
Bij
);
// update the est and eet for all the subsequent activities;
7) UPDATE( est,
eet);
// update local pheromone for both duration and cost
8) LOCALUPDATE( dτ
ij
, dτ
ij
);
}
// update the global pheromone based on the makespan and cost of the best-sofar
solution
bs bs
9) GLOBALUPDATE
( dτ
ij
, dτ
ij
, makespan , cos t );
// return the best-so-far solution and record into the SolutionSet
10) RETURN( Solution,
SolutionSet);
}
// Selecting the BestSolution according to user preferences
11) BestSoluti on = COMPARE( SolutionSet , UserPref );
// deploy the scheduling plan
12) Deploy(L).
Figure 8.4 ACO based Rescheduling Algorithm
8.3.4 Other Representative Metaheuristic Algorithms
Besides GA and ACO introduced above, other representative metaheuristic
algorithms often used in workflow scheduling/rescheduling include PSO (Particle
Swarm Optimisation) [102], GRASP (Greedy Randomised Adaptive Search
Procedure) [100] and SA (Simulated Annealing) [5]. In this thesis, we do not intend
to investigate all of the metaheuristics algorithms. Readers interested in an overview
of representative metaheuristics based algorithms may refer to [6]. Meanwhile, we
136