Model-driven Application Refactoring to Minimize Deployment Costs in Preemptible Cloud Resources

1. Attached files:        

This archive contains 1440 MATLAB files, each one containing the results of a single experiment.
The name of each file follows the following format:

                       A_B_C_null_D_0.95_E_F.mat

Where the fields A, B, C, D, E, and F are described as follows.

A: number of users.
Considered values are: 2500, 5000, 10000, 20000, 40000.

B: variation in the profile of the requests.
Considered values are: ref, var1, var2, var3, var4, var5.
ref -> reference experiment, no users are halved or doubled
var1 -> users in class 1 halved, other users doubled 
var2 -> users in class 2 halved, other users doubled 
var3 -> users in class 3 halved, other users doubled 
var4 -> users in class 4 halved, other users doubled 
var5 -> users in class 5 halved, other users doubled 

C:  variation of the replaceability set of the application server.
Considered values are: ref, k1, k2, k3, k4, k5.
ref -> reference experiment, no rates are halved or doubled
k1 -> only possible substitution has rate k1 halved and other rates doubled
k2 -> only possible substitution has rate k2 halved and other rates doubled 
k3 -> only possible substitution has rate k3 halved and other rates doubled 
k4 -> only possible substitution has rate k4 halved and other rates doubled 
k5 -> only possible substitution has rate k5 halved and other rates doubled 

D: variation of the design constraints.
Considered values are: none, 4, 34, 234.
none -> no components can be replicated
4 -> only the application server can be replicated
34 -> only the application server and the database server can be replicated
234 -> all the components can be replicated

E: Optimization algorithm.
Considered values are: norefactoring, replacement, reassignment, full
norefactoring -> experiment with no refactorings
replacement -> experiment with only replacement refactoring
reassignment -> experiment with only reassignment refactoring
full -> experiment with replacement and reassignment refactorings

F: Experiment seed.
Considered values are from 1 to 20

2. Data format:

MATLAB data format, can be loaded from MATLAB using the following command:

results = load(filename);

results is defined as a structure with the following fields:

results.cost
    Type:    scalar, positive real number.
    Desc:    hourly cost in US dollars.

        
results.time
    Type:    scalar, positive real number.
    Desc:    total time (in seconds) needed by the algorithm to compute the solution.

results.evaluations
    Type:    scalar, positive integer number.
    Desc:    number of constraints evaluations needed by the algorithm to compute the 
        solution.

results.d
    Type:    matrix, non negative positive real number. 
    Desc:    association matrix between rented resources (columns) and application 
        components (rows). The sum of all the elements of this matrix is equal to
        the ECUs used by the application.