An In-depth Benchmarking of Evolutionary and Swarm Intelligence Algorithms for Autoscaling Parameter Sweep Applications on Public Clouds

Table 10

RPD-oriented metric values obtained for applications FPA and WRF. For all metrics, positive values represent favorable results (savings respecting MOEA-NSGA-III).

Application	Size	Autoscaler	Makespan RPD (%)	Cost RPD (%)	Task failures RPD (%)
FPA	30	MOEA-SMPSO	−40.68	−125.58	−100.00
		MOEA-SMS-EMOA	0.81	−379.07	0.00
		MOEA-E-NSGA-III	0.90	6.98	0.00
	100	MOEA-SMPSO	−67.08	−58.69	−100.00
		MOEA-SMS-EMOA	2.43	−57.28	0.00
		MOEA-E-NSGA-III	1.78	15.96	0.00
	300	MOEA-SMPSO	−29.28	−43.70	−100.00
		MOEA-SMS-EMOA	−33.65	−93.41	0.00
		MOEA-E-NSGA-III	4.09	9.17	0.00
WRF	30	MOEA-SMPSO	−1.07	11.92	−100.00
		MOEA-SMS-EMOA	−3.15	40.07	0.00
		MOEA-E-NSGA-III	7.66	11.56	0.00
	100	MOEA-SMPSO	−5.17	−17.06	−1042.50
		MOEA-SMS-EMOA	−5.57	47.87	−732.50
		MOEA-E-NSGA-III	3.14	39.32	100.00
	300	MOEA-SMPSO	−1.69	−51.20	−89.11
		MOEA-SMS-EMOA	−7.95	−27.69	100.00
		MOEA-E-NSGA-III	3.95	32.35	100.00

indicate the best value.