Software Defined Networking has afforded numerous benefits to the network users but there are certain persisting issues with this technology, two of which are scalability and privacy. The natural solution to overcoming these limitations is a distributed SDN controller architecture where multiple controllers are deployed over the network, with each controller orchestrating a certain segment of the network. However, since the centralized control is the key attribute of SDN that allows it to be so beneficial, a centralized logical view of the network will have to be maintained by each of these controllers; this can be done through synchronization of the distributed controllers, where each controller communicates with the others to ensure that they remain informed about the entire network. There is however a network cost associated with constantly having to update each others about different aspects of the network, which will become a greater issue in dynamic wireless networks. To minimize this network cost, there is a need to consider not only when to get the update information from the neighboring controllers, but also where to dynamically place the controllers such that the network costs may be minimized. The placement should take into consideration both communication for synchronization among the distributed controllers and communication of the controllers with the network devices that they manage. In this work, we show that our multi-objective deep reinforcement learning-based method performs the best at achieving different application goals by developing policy for controller synchronization as well as placement, outperforming different other possible approaches, under a wide variety of network conditions.

The growing number of AI-driven applications in mobile devices has led to solutions that integrate deep learning models with the available edge-cloud resources. Due to multiple benefits such as reduction in on-device energy consumption, improved latency, improved network usage, and certain privacy improvements, split learning, where deep learning models are split away from the mobile device and computed in a distributed manner, has become an extensively explored topic. Incorporating compression-aware methods (where learning adapts to compression level of the communicated data) has made split learning even more advantageous. This method could even offer a viable alternative to traditional methods, such as federated learning techniques. In this work, we develop an adaptive compression-aware split learning method (“deprune”) to improve and train deep learning models so that they are much more network-efficient, which would make them ideal to deploy in weaker devices with the help of edge-cloud resources. This method is also extended (“prune”) to very quickly train deep learning models through a transfer learning approach, which tradesoff little accuracy for much more network-efficient inference abilities. We show that the “deprune” method can reduce network usage by 4× when compared with a split-learning approach (that does not use our method) without loss of accuracy, while also improving accuracy over compression-aware split-learning by up to 4 percent. Lastly, we show that the “prune” method can reduce the training time for certain models by up to 6× without affecting the accuracy when compared against a compression-aware split-learning approach.

A logically centralized controller architecture for SDN deployments is a well understood and implemented method, however because of issues such as scalability, privacy and more, there is a need to develop and implement a robust physically distributed SDN controller architecture. In a distributed SDN environment, a centralized logical network view needs to be maintained, which means the distributed controllers need a robust method of remaining informed about other controller's network through synchronization. This is specially true in mobile, wireless networks with changing controller and network environment, so to this end we develop a deep reinforcement and transfer learning based method that provides the controllers with an efficient policy for synchronizing with other controllers and maintaining a logically centralized view in such networks. We show that our application-centric method performs well for different kinds of applications including shortest path routing and load balancing, outperforming a reinforcement learning based method as well as a round robin method.

In this article, we introduce a fog system design for processing data collected from edge devices, such as mobile, sensor, and extended (mixed, augmented, virtual) reality equipment. Our system enables the network to provide hardware-accelerated processors for resource-intensive computations on data gathered from remote locations, such as 5G and beyond mobile networks. By splitting heavy computations into pipelines, and distributing them among processors in the edge, fog and the cloud, our design benefits from the processing power of the cloud, while utilizing fog devices with a lower network latency. We implement our design, and use it for distributed training and inference with industry-grade deep learning models for computer vision. We deploy our architecture in infrastructure including cloud and edge servers supporting GPU-accelerated computations. We benchmark pipelines in various deployment settings to study the overhead that they introduce. Our contributions are a new design for wide-area data processing, a framework that realizes this design and provides means of developing applications that are optimized in terms of infrastructure and hardware. These contributions are complemented with our benchmark results, which reveal the potential causes of processing overhead.

As the containerization of network services is expanding towards the Radio Access Network (RAN), the operators seek to benefit from the paradigm of cloud-native services through a wide ecosystem of practices that are applicable to such resources. Such practices include the dynamic scaling of the services in response to the demand, with the network service being assigned more/less resources, or replicated, for accommodating the incoming demand. In such cloud-native environments, proactive decisions can be accomplished through Machine Learning models, which are efficiently trained for specific metrics that reflect the network demand. In this work, we use a real cloud-native telecommunications network and real traffic patterns, and evaluate four different Machine Learning methods for predicting the incoming demand. The decisions made based on the predictions regard the scaling of the base station (gNB/eNB) and the core network entities that deal with the User-Plane traffic (UPF/SPGW-U). Our results show that higher accuracy for such predictions can be accomplished using the tree-based methods over the Neural Network-based solutions, when each method is used for making accurate pro-active decisions for scaling of the under-study network functions.

The evolution of the different network functions to a cloud-native configuration creates fertile ground for the efficient management and reconfiguration of the network. Through the wide application of softwarization and virtualization, cloud-native approaches can extend even to the RAN, that has been dominated by monolithic non-configurable hardware equipment in the past generations of mobile network access. As such, a cloud-native deployment can cover the end-to-end 5G network architecture, from the Core Network to the base stations, with the respective services benefiting from several advanced features, such as automatic scaling of the deployed functions based on monitored metrics. Through the application of Machine Learning, the evolution of the metrics can be predicted and thus the respective functions can be pro-actively scaled. In this work, we use an end-to-end real-world cloud-native deployment of a 5G network, and deal with two different types of scaling, applied at three different parts of the network: vertical scaling for the base station, and horizontal scaling for control and user plane functions of the core network. We use a real-world dataset for replicating traffic over our setup and closely monitor the evolution of metrics from different parts of the network. By applying Machine Learning methods, we accurately predict the future network load and use it to decide on the pro-active allocation of resources for the RAN and the Core Network.

Existing software-defined SDx architectures highly depend on a centralized control plane and hence can face substantial reliability challenges in software-defined coalition (SDC) settings, in which the centralized control plane can be weakly connected to the data plane, or even disconnected from the data plane due to high dynamicity. On the contrary, distributed control planes (e.g., OLSRv2) provide autonomy but lose flexibility and global policy guarantees. In this paper, we present Magnalium, a novel system to achieve high reliability in SDC networks by composing multiple control planes in real-time. Magnalium introduces a novel, unified composition framework that uses a distributed verification to systematically generate forwarding rules in accordance with desired policy requirements. Magnalium also introduces several supporting components to address challenges in wireless environment and resource management. We conduct data-driven simulations, showing that Magnalium benefits from both centralized and distributed control planes and even reduces downtime by 65% over the most reliable individual control plane.