In the context of this PhD Thesis defense, Mohammed Osman presents his work realized at CTTC.
Abstract:
The SDN (Software-Defined Networks) architecture separates the data and the control planes of the networks. It logically centralizes the control of a network in a central point that is an SDN controller, which acts as a brain of the network and is in charge of telling each network node how to forward incoming packets by installing the appropriate forwarding rules. One of the main advantages it brings is programmability through this single entity (the logical controller) with which network management applications must interact to apply their policies. Through agreedupon APIs, the network managers can exploit the full potential of SDN.
SDN generally assumes ideal control channels between the SDN controller and the network nodes, which may not be the case in challenging environments that are becoming more common due to dense deployment of small cells (SCs) with reduced coverage in 5G and beyond 5G deployments. In 5G and beyond 5G use cases, cost-effective wireless transport networks are required to connect the SCs. In this context, mmWave technology is a good player to connect the SCs as mmWave provides larger radio spectrum chunks that in turn provide larger bandwidth and higher data rate.
To manage the dense deployment of SCs in the mobile networks, on the network management/control front, network programmability and virtualization are also an integral part of 5G and beyond 5G networks. In this regard, to provide end-to-end connectivity, management and orchestration of all the segments of the networks ranging from RAN (Radio Access Network), transport network to the core is vital. On the transport networks side (the main focus of the dissertation), SDN plays an important role as SDN enables programmability and virtualization in the network.
Though SDN Provides huge flexibility in network management by splitting the control plane from the data plane, it has some limitations in wireless networks context as separation of the control plane from the data plane introduce the extra points of failure in the SDN paradigm (e.g., control communication channel failure, SDN controller failure). In the wide-area networks (WAN) scenarios where in-band channels (e.g., microwave or mmWave links) are responsible to carry control traffic between the forwarding nodes and the SDN controller, the assumption of the availability of a reliable network may not be possible as the performance of the wireless link changes with the environmental conditions, which leads to a high risk of experiencing channel impairments, which might cause centralized SDN operation failure by affecting communication between the transport component of SCs and the SDN controller.
To overcome SDN from failure, the dissertation presents a hybrid SDN scheme that explores the benefits of centralized and distributed operations depending on control communication channel conditions. Our hybrid SDN approach combines both centralized and distributed modes in the same node to form a hybrid control plane architecture. We introduce a local agent in the node that is composed of a monitoring framework to detect reliability of the control communication channel and a decision module that conceive a novel control logic switching algorithm to make a decision whether to operate in a centralized or distributed mode. We evaluate the proposed solution under a variety of unreliable network conditions (e.g., link impairments, control packet loss) to investigate the operational performance of the hybrid SDN during high loss conditions. The experimental results show that the proposed hybrid SDN solution substantially improves the aggregated throughput, particularly when control channel packet loss ratios increase, which in turn keeps the network operational in hard conditions where the centralized SDN would result in a non-operational network.
Virtual defense at 10:00h: Click to this link
