While recent years have witnessed a steady trend of applying Deep Learning (DL) to networking systems, most of the underlying Deep Neural Networks (DNNs) suffer two major limitations. First, they fail to generalize to topologies unseen during training. This lack of generalizability hampers the ability of the DNNs to make good decisions every time the topology of the networking system changes. Second, existing DNNs commonly operate as "blackboxes" that are difficult to interpret by network operators, and hinder their deployment in practice. In this paper, we propose to rely on a recently developed family of graph-based DNNs to address the aforementioned limitations. More specifically, we focus on a network congestion prediction application and apply Graph Attention (GAT) models to make congestion predictions per link using the graph topology and time series of link loads as inputs. Evaluations on three real backbone networks demonstrate the benefits of our proposed approach in terms of prediction accuracy, generalizability, and interpretability.

Security threats arising in massively connected Internet of Things (IoT) devices have attracted wide attention. It is necessary to equip IoT gateways with firewalls to prevent hacked devices from infecting a larger amount of network nodes. The match-and-action mechanism of Software Defined Networking (SDN) provides the means to differentiate malicious traffic flows from normal ones, which mirrors the past firewall mechanisms but with a new flexible and dynamically reconfigurable twist. However, vulnerabilities of IoT devices and heterogeneous protocols coexisting in the same network challenge the extension of SDN into the IoT domain. To overcome these challenges, we leverage the high level of data plane programmability brought by the P4 language and design a novel two-stage deep learning method for attack detection tailored to that particular language. Our method is able to generate flow rules that match a small number of header fields from arbitrary protocols while maintaining high performance of attack detection. Evaluations using network traces of different IoT protocols show significant benefits in accuracy, efficiency and universality over state-of-the-art methods.

5G technologies promise to revolutionize mobile networks and push them to the limits of resource utilization. Besides better capacity, we also need better resource management via virtualization. End-to-end network slicing not only involves the core but also the Radio Access Network (RAN) which makes this a challenging problem. This is because multiple alternative radio access technologies exist (e. g. ,LTE, WLAN, and WiMAX), and there is no unifying abstraction to compare and compose from diverse technologies. In addition, existing work assumes that all RAN infrastructure exists under a single administrative domain. Software-Defined Radio Access Network (SD-RAN) offers programmability that facilitates a unified abstraction for resource sharing and composition across multiple providers harnessing different technology stacks. In this paper we propose a new architecture for heterogeneous RAN slicing across multiple providers. A central component in our architecture is a service orchestrator that interacts with multiple network providers and service providers to negotiate resource allocations that are jointly optimal. We propose a double auction mechanism that captures the interaction among selfish parties and guarantees convergence to optimal social welfare in finite time. We then demonstrate the feasibility of our proposed system by using open source SD-RAN systems such as EmPOWER (WiFi) and FlexRAN (LTE).

Since the early development of Software-Defined Network (SDN) technology, researchers have been concerned with the idea of physical distribution of the control plane to address scalability and reliability challenges of centralized designs. However, having multiple controllers managing the network while maintaining a “logically-centralized” network view brings additional challenges. One such challenge is how to coordinate the management decisions made by the controllers which is usually achieved by disseminating synchronization messages in a peer-to-peer manner. While there exist many architectures and protocols to ensure synchronized network views and drive coordination among controllers, there is no systematic methodology for deciding the optimal frequency (or rate) of message dissemination. In this paper, we fill this gap by introducing the SDN synchronization problem: how often to synchronize the network views for each controller pair. We consider two different objectives; first, the maximization of the number of controller pairs that are synchronized, and second, the maximization of the performance of applications of interest which may be affected by the synchronization rate. Using techniques from knapsack optimization and learning theory, we derive algorithms with provable performance guarantees for each objective. Evaluation results demonstrate significant benefits over baseline schemes that synchronize all controller pairs at equal rate.

Software defined networking (SDN) can be beneficial in mobile ad hoc networks (MANETs) to increase flexibility, provide programmability and simplify management. The high dynamics in mobile networks, however, raise new reliability challenges to the conventional centralized control plane of SDN. To increase reliability, methods such as placing multiple controllers in the network have been considered that add redundancy in the control plane in a brute force manner. However, these methods cannot by themselves fundamentally solve the reliability problem. To address this issue, this paper complements the controller placement methods with a new architecture that has a hybrid structure splitting the routing decision logic between the controllers and the data plane nodes. Specifically, the controllers can break the routing path into segments, similar to the segment routing technique, and broadcast the list of segment labels to the data plane nodes. The latter are able to make the actual forwarding decisions for each segment in a distributed manner, e.g., by running an existing MANET protocol like OLSR. Experiments on a testbed built from commercial mobile devices with integrated SDN functionality highlight the feasibility and benefits of the proposed architecture.

Fog architectures at the network edge are becoming a popular research trend to provide elastic resources and services to end-users, where the processing capacity resides at the network periphery as opposed to traditional data-centers. Despite their momentum, the control plane of these architectures remains complex and challenging to implement. To enhance control capability, in this work, we propose to use Software Defined Networking. SDN moves the control logic off data plane devices and onto external network entities, the controllers. We provide a proof-of-concept implementation of a multi-controller edge system and measure traffic delay and overheads. The results reveal the sensitivity of delay to the location of controllers and the magnitude of inter-controller and controller-node overheads. Guided by the above, we model the problem of determining the placement of controllers in the edge network. Using linearization and supermodular function techniques, we present approximation solutions which perform close to optimal and better than state-of-the-art methods.

Fog architectures at the network edge are becoming a popular research trend to provide elastic resources and services to end-users, where the processing capacity resides at the network periphery as opposed to traditional data-centers. Despite their momentum, the control plane of these architectures remains complex and challenging to implement. To enhance control capability, in this paper, we propose to use software defined networking (SDN). SDN moves the control logic off data plane devices and onto external network entities, the controllers. We provide a proof-of-concept implementation of a multi-controller edge system and measure traffic delay and overheads. The results reveal the sensitivity of delay to the location of controllers and the magnitude of inter-controller and controller-node overheads. Guided by the above, we model the problem of determining the placement of controllers in the edge network. Using linearization and supermodular function techniques, we present approximation solutions which perform close to optimal and substantially better than state-of-the-art methods. Finally, we analyze the interplay between various performance and reliability objectives.