Abstract: Fueled by the explosive growth of ultra-low-latency and real-time applications with specific computing and network performance requirements, the computing force network (CFN) has become a hot research subject. The primary CFN challenge is to leverage network resources and computing resources. Although recent advances in deep reinforcement learning (DRL) have brought significant improvement in network optimization, these methods still suffer from topology changes and fail to generalize for those topologies not seen in training. This paper proposes a graph neural network (GNN) based DRL framework to accommodate network traffic and computing resources jointly and efficiently. By taking advantage of the generalization capability in GNN, the proposed method can operate over variable topologies and obtain higher performance than the other DRL methods.

图神经网络与深度强化学习结合的算力网络资源分配方法

[1]Almasan P, Suárez-Varela J, Rusek K, et al., 2022. Deep reinforcement learning meets graph neural networks: exploring a routing optimization use case. https://arxiv.org/abs/1910.07421

[2]Badia-Sampera A, Suárez-Varela J, Almasan P, et al., 2019. Towards more realistic network models based on graph neural networks. Proc 15^th Int Conf on Emerging Networking Experiments and Technologies, p.14-16.

[3]Barbarossa S, Sardellitti S, Lorenzo PD, 2014. Communicating while computing: distributed mobile cloud computing over 5G heterogeneous networks. IEEE Signal Process Mag, 31(6):45-55.

[4]China Mobile, Huawei Technologies, 2019. Technical White Paper on Computing-Aware Networking.

[5]Du ZP, Li ZQ, Duan XD, et al., 2022. Service information informing in computing aware networking. Proc Int Conf on Service Science, p.125-130.

[6]Ferriol-Galmés M, Suárez-Varela J, Barlet-Ros P, et al., 2020. Applying graph-based deep learning to realistic network scenarios. https://arxiv.org/abs/2010.06686

[7]Geyer F, 2017. Performance evaluation of network topologies using graph-based deep learning. Proc 11^th EAI Int Conf on Performance Evaluation Methodologies and Tools, p.20-27.

[8]Gilmer J, Schoenholz SS, Riley PF, et al., 2017. Neural message passing for quantum chemistry. Proc 34 ^th Int Conf on Machine Learning, p.1263-1272.

[9]Guo SY, Dai Y, Xu SY, et al., 2020. Trusted cloud-edge network resource management: DRL-driven service function chain orchestration for IoT. IEEE Int Things J, 7(7):6010-6022.

[10]Kiani A, Ansari N, 2018. Edge computing aware NOMA for 5G networks. IEEE Int Things J, 5(2):1299-1306.

[11]Li M, Yang L, Yu FR, et al., 2019. Joint optimization of networking and computing resources for green M2M communications based on DRL. Proc IEEE Global Communications Conf, p.1-6.

[12]Liu B, Mao JW, Xu L, et al., 2021. CFN-dyncast: load balancing the edges via the network. Proc IEEE Wireless Communications and Networking Conf Workshops, p.1-6.

[13]Mao YY, You CS, Zhang J, et al., 2017. A survey on mobile edge computing: the communication perspective. IEEE Commun Surv Tutor, 19(4):2322-2358.

[14]Moreira R, Silva F, Frosi P, et al., 2018. A flexible network and compute-aware orchestrator to enhance QoS in NFV-based multimedia services. Proc IEEE 32^nd Int Conf on Advanced Information Networking and Applications, p.512-519.

[15]Nordhaus WD, 2001. The Progress of Computing. Cowles Foundation Discussion Papers.

[16]Orlowski S, Wessäly R, Pióro M, et al., 2010. SNDlib 1.0—survivable network design library. Networks, 55(3):276-286.

[17]Ren YL, Chen XY, Guo S, et al., 2021. Blockchain-based VEC network trust management: a DRL algorithm for vehicular service offloading and migration. IEEE Trans Veh Technol, 70(8):8148-8160.

[18]Ruiz L, Gama F, Ribeiro A, 2021. Graph neural networks: architectures, stability, and transferability. Proc IEEE, 109(5):660-682.

[19]Rusek K, Suárez-Varela J, Mestres A, et al., 2019. Unveiling the potential of graph neural networks for network modeling and optimization in SDN. Proc ACM Symp on SDN Research, p.140-151.

[20]Suárez-Varela J, Carol-Bosch S, Rusek K, et al., 2019. Challenging the generalization capabilities of graph neural networks for network modeling. Proc ACM SIGCOMM Conf Posters and Demos, p.114-115.

[21]Sun PH, Lan JL, Li JF, et al., 2021. Combining deep reinforcement learning with graph neural networks for optimal VNF placement. IEEE Commun Lett, 25(1):176-180.

[22]Suzuki T, Yasuda Y, Nakamura R, et al., 2020. On estimating communication delays using graph convolutional networks with semi-supervised learning. Proc Int Conf on Information Networking, p.481-486.

[23]Tran TX, Hajisami A, Pandey P, et al., 2017. Collaborative mobile edge computing in 5G networks: new paradigms, scenarios, and challenges. IEEE Commun Mag, 55(4):54-61.

[24]Wang LN, Gu RT, Li ZK, et al., 2021. Computing-aware proactive IP-optical integrated network restructuring for edge computing. Proc 19 ^th Int Conf on Optical Communications and Networks, p.1-3.

[25]Wu ZH, Pan SR, Chen FW, et al., 2021. A comprehensive survey on graph neural networks. IEEE Trans Neur Netw Learn Syst, 32(1):4-24.

[26]Xie M, Yu K, Wu X, 2023. Adaptive routing of task in computing force network by integrating graph convolutional network and deep Q-network. Proc 8 ^th IEEE Int Conf on Network Intelligence and Digital Content, p.242-247.

[27]Xu ZY, Jian T, Meng JS, et al., 2018. Experience-driven networking: a deep reinforcement learning based approach. Proc IEEE Conf on Computer Communications, p.1871-1879.

[28]Yang BX, Chai WK, Xu ZC, et al., 2018. Cost-efficient NFV-enabled mobile edge-cloud for low latency mobile applications. IEEE Trans Netw Serv Manag, 15(1):475-488.

[29]Yao H, Duan X, Fu Y, 2022. Computing-aware routing protocol for computing force network. Int Conf on Service Science, p.137-141

[30]Yu S, Chen X, Zhou Z, et al., 2021. When deep reinforcement learning meets federated learning: intelligent multitimescale resource management for multiaccess edge computing in 5G ultradense network. IEEE Int Things J, 8(4):2238-2251.

[31]Zhang ZW, Cui P, Zhu WW, 2022. Deep learning on graphs: a survey. IEEE Trans Knowl Data Eng, 34(1):249-270.