Recent research on Software-Defined Networking (SDN) strongly promotes the adoption of distributed controller architectures. To achieve high network performance, designing a scheduling function (SF) to properly dispatch requests from each switch to suitable controllers becomes critical. However, existing literature tends to design the SF targeted at specific network settings. In this paper, a reinforcement-learning-based (RL) approach is proposed with the aim to automatically learn a general, effective, and efficient SF. In particular, a new dispatching system is introduced in which the SF is represented as a neural network that determines the priority of each controller. Based on the priorities, a controller is selected using our proposed probability selection scheme to balance the trade-off between exploration and exploitation during learning. In order to train a general SF, we first formulate the scheduling function design problem as an RL problem. Then a new training approach is developed based on a state-of-the-art deep RL algorithm. Our simulation results show that our RL approach can rapidly design (or learn) SFs with optimal performance. Apart from that, the trained SF can generalize well and outperforms commonly used scheduling heuristics under various network settings.
Cloud computing provides on-demand access to computational resources while outsourcing infrastructure and service maintenance. Edge computing could extend cloud computing capability to areas with limited computing resources, such as rural areas, by utilizing low-cost hardware, such as singleboard computers. Cloud data centre hosted machine learning algorithms may violate user privacy and data confidentiality requirements. Federated learning (FL) trains models without sending data to a central server and ensures data privacy. Using FL, multiple actors can collaborate on a single machine learning model without sharing data. However, rural network outages can happen at any time, and the quality of a wireless network varies depending on location, which can affect the performance of the Federated Learning application. Therefore there is a need to have a platform that maintains service quality independent of infrastructure status. We propose a self-adaptive system for rural FL, which employs the Greedy Nominator Heuristic (GNH) based optimisation to orchestrate application workflows across multiple resources that make up a rural computing environment. GNH provides distributed optimization for workflow placement. GNH utilises resource status to reduce failure risks and costs while still completing tasks on time. Our approach is validated using a simulated rural environment -composed of multiple decentralized controllers sharing the same infrastructure and running a shared FL application. Results show that GNH outperforms three algorithms for deployment of FL tasks: random placement, round-robin load balancer and simple greedy algorithm.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.