TY - GEN
T1 - Deep Reinforcement Learning based Cloud-native Network Function Placement in Private 5G Networks
AU - Kim, Joonwoo
AU - Lee, Jaewook
AU - Kim, Taeyun
AU - Pack, Sangheon
N1 - Funding Information:
ACKNOWLEDGEMENT This work was supported in part by Samsung Research in Samsung Electronics and in part by National Research Foundation (NRF) of Korea Grant funded by the Korean Government (MSIP) (No. 2020R1A2C3006786).
Publisher Copyright:
© 2020 IEEE.
PY - 2020/12
Y1 - 2020/12
N2 - With the advantages of satisfying service requirements and providing high security, standalone private fifth generation (5G) network is perceived as a promising technology for vertical industries. However, to manage the cloud-native network functions (CNFs) in an effective manner, a sophisticated control plane management scheme should be designed in standalone private 5G networks. In this paper, we propose a deep Q-network based CNF placement algorithm (DQN-CNFPA), that jointly minimizes the cost occurred in launching and operating CNFs on edge clouds and the back-haul control traffic overhead. In addition, DQN-CNFPA learns spatiotemporal patterns in service requests and places CNFs in consideration of future cost leveraged by the previous CNF placement strategy. Evaluation results demonstrate that DQN-CNFPA can reduce the cost per hour up to 11.2% compared to the scheme without learning spatiotemporal service request patterns.
AB - With the advantages of satisfying service requirements and providing high security, standalone private fifth generation (5G) network is perceived as a promising technology for vertical industries. However, to manage the cloud-native network functions (CNFs) in an effective manner, a sophisticated control plane management scheme should be designed in standalone private 5G networks. In this paper, we propose a deep Q-network based CNF placement algorithm (DQN-CNFPA), that jointly minimizes the cost occurred in launching and operating CNFs on edge clouds and the back-haul control traffic overhead. In addition, DQN-CNFPA learns spatiotemporal patterns in service requests and places CNFs in consideration of future cost leveraged by the previous CNF placement strategy. Evaluation results demonstrate that DQN-CNFPA can reduce the cost per hour up to 11.2% compared to the scheme without learning spatiotemporal service request patterns.
UR - http://www.scopus.com/inward/record.url?scp=85102937106&partnerID=8YFLogxK
U2 - 10.1109/GCWkshps50303.2020.9367481
DO - 10.1109/GCWkshps50303.2020.9367481
M3 - Conference contribution
AN - SCOPUS:85102937106
T3 - 2020 IEEE Globecom Workshops, GC Wkshps 2020 - Proceedings
BT - 2020 IEEE Globecom Workshops, GC Wkshps 2020 - Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2020 IEEE Globecom Workshops, GC Wkshps 2020
Y2 - 7 December 2020 through 11 December 2020
ER -