Fast DRL-based scheduler configuration tuning for reducing tail latency in edge-cloud jobs

Journal of Cloud Computing

Advances, Systems and Applications

Table 2 Percentages of reduced tail latency under DAG batch workloads driven by Alibaba trace 2018

Scenario	Job scheduling scenarios			TRP		BRA	EP	ILP	IPAP	LRP	MRP	NAP	NLP	NPAPP	RLP	TTP
1	DAG batch workloads	Daytime	Original resources	No policy	0%	26.25	29.52	35.03	28.46	27.92	20.42	30.55	23.60	30.55	27.38	29.52
2				Restart rate	80%	18.10	20.17	26.92	18.80	17.03	13.64	22.13	15.56	21.49	19.49	21.16
3				Restart rate	70%	12.50	10.85	19.57	13.30	12.09	7.80	12.09	10.85	12.09	13.70	12.50
4		Night	Original resources	No policy	0%	23.38	28.14	27.94	27.54	30.53	24.82	27.74	26.41	27.44	27.13	27.54
5				Restart rate	80%	17.09	22.88	21.72	21.00	21.84	19.03	21.60	20.64	21.95	21.12	22.19
6				Restart rate	70%	9.91	14.10	11.75	11.75	13.67	11.30	12.35	13.23	14.24	13.38	15.36