Agent-based cloud simulation model for resource management

Journal of Cloud Computing

Advances, Systems and Applications

Table 3 Statistical mean differences observed between the staged evaluations of each algorithm, conducted independently within both homogeneous and heterogeneous cloud environments

Test Methods	Levene	Shapiro	ANOVA	Kruskal	Environment
random (rnd)	p>0.999	(W=0.992, *p=0.326*)	*p>0.999*	p>0.999	Homogeneous, {1, 2, 3}SE
random (rnd)	p=0.961	(W=0.986, p=0.041)	p=0.997	p=0.976	Heterogeneous, {1, 2, 3}SE
first-fit (ff)	p=0.996	(W=0.988, p=0.013)	p=0.999	p>0.999	Homogeneous, {1, 2, 3}SE
first-fit (ff)	p=0.976	(W=0.955, p=4.9e-8)	p=0.989	p=0.996	Heterogeneous, {1, 2, 3}SE
balanced-fit (bf)	p=0.937	(W=0.980, p=6.6e-3)	p=0.864	p=0.851	Homogeneous, {1, 2, 3}SE
balanced-fit (bf)	p=0.804	(W=0.976, p=1.8e-3)	p=0.829	p=0.872	Heterogeneous, {1, 2, 3}SE
max-utilization (mu)	p=0.985	(W=0.987, p=6.9e-3)	p=0.889	p=0.924	Homogeneous, {1, 2, 3}SE
max-utilization (mu)	p=0.767	(W=0.993, *p=0.204*)	*p=0.611*	p=0.794	Heterogeneous, {1, 2, 3}SE
min-energy (me)	p=0.893	(W=0.978, p=1.4e-4)	p>0.863	p=0.794	Homogeneous, {1, 2, 3}SE
min-energy (me)	p=0.986	(W=0.977, p=1.1e-4)	p=0.995	p=0.996	Heterogeneous, {1, 2, 3}SE

The performance difference between the algorithms were tested using either ANOVA (parametric) or Kruskal-Wallis (non-parametric) methods, depending on whether the Levene’s test for homogeneity of variance and Shapiro-Wilk method for normality residuals passed (p > 0.05 indicates no violation). If both tests pass, ANOVA results are more trustworthy; otherwise, Kruskal-Wallis results prevail. In all cases, p-values were above the significance threshold (0.05) from both ANOVA and Kruskal-Wallis tests, accepting the null hypothesis and suggesting no significant differences between evaluation stages for each algorithm in both homogeneous and heterogeneous cloud environments