Task grouping and optimized deep learning based VM sizing for hosting containers as a service

Journal of Cloud Computing

Advances, Systems and Applications

Table 8 Comparative Assessment of Deep-ConvLSTM_FPO Scheme With Task Size 300

Number of Iteration	VM size selection technique	Hurst exponent+Markov transition	ICLB	OM-FNN	VM size IaaS multi-tenant public cloud	Deep-ConvLSTM GD	Deep-ConvLSTM ADAM	Proposed Deep-ConvLSTM_FPO
Resource Utilization
5	0.1428	0.1223	0.1095	0.1058	0.1040	0.0997	0.0993	0.0987
10	0.1338	0.1286	0.1220	0.1187	0.1133	0.1120	0.1101	0.1025
15	0.1233	0.1195	0.1175	0.1164	0.1158	0.1137	0.1117	0.1025
20	0.1495	0.1217	0.1185	0.1181	0.1179	0.1147	0.1127	0.1041
Response Time
5	51830	42142	34677	20456	17531	981	734	650
10	51791	43258	34657	20583	17500	1021	786	527
15	51872	45125	34719	20725	17493	1089	848	511
20	52590	47853	35405	20892	17849	1156	891	482
Task Rejection Rate
5	0.4013	0.3954	0.3658	0.3569	0.3414	0.3368	0.3318	0.3254
10	0.4125	0.3999	0.3714	0.3688	0.3515	0.3486	0.3458	0.3303
15	0.4185	0.4014	0.3799	0.3627	0.3699	0.3647	0.3618	0.3401
20	0.4014	0.3985	0.3854	0.3800	0.3479	0.3447	0.3418	0.3387
Makespan
5	0.7689	0.7245	0.6865	0.6578	0.6367	0.6257	0.6145	0.6087
10	0.7866	0.7367	0.7076	0.6867	0.6578	0.6468	0.6367	0.6256
15	0.8087	0.7578	0.7146	0.6967	0.6798	0.6678	0.6624	0.6578
20	0.8256	0.7790	0.7468	0.7156	0.6865	0.6846	0.6825	0.6755