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

Table 9

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

Application	Size	Autoscaler	Makespan RPD (%)	Cost RPD (%)	Task failures RPD (%)
MPG	30	MOEA-SMPSO	−58.10	8.19	−100.00
		MOEA-SMS-EMOA	1.65	−60.23	100.00
		MOEA-E-NSGA-III	2.46	11.11	100.00
	100	MOEA-SMPSO	−54.90	−135.59	−100.00
		MOEA-SMS-EMOA	−57.47	−175.45	0.00
		MOEA-E-NSGA-III	10.90	14.86	0.00
	300	MOEA-SMPSO	17.77	19.28	−195.00
		MOEA-SMS-EMOA	−25.24	−69.56	5.00
		MOEA-E-NSGA-III	16.56	18.18	100.00
GMS	30	MOEA-SMPSO	−32.44	14.11	2.05
		MOEA-SMS-EMOA	3.78	−28.07	−23.97
		MOEA-E-NSGA-III	2.19	23.50	36.76
	100	MOEA-SMPSO	−35.30	6.88	−61.20
		MOEA-SMS-EMOA	−5.66	−32.07	51.80
		MOEA-E-NSGA-III	10.37	5.57	54.36
	300	MOEA-SMPSO	−32.50	6.97	−203.20
		MOEA-SMS-EMOA	−32.73	−40.27	−89.36
		MOEA-E-NSGA-III	10.11	15.52	33.99

indicate the best value.