Recent years have seen an evolution of software-defined networking (SDN) control plane architectures, starting from simple monolithic controllers, over modular monolithic controllers, to distributed controllers. We observe, however, that today's distributed controllers still exhibit inflexibility with respect to the distribution of control logic. Therefore, we propose a novel architecture of a distributed SDN controller, providing maximum flexibility with respect to distribution and improved manageability. Our architecture splits control logic into lightweight control modules, called controllets, based on a micro-kernel approach, reducing common controllet functionality to a bare minimum and factoring out all higher-level functionality. Lightweight controllets also allow for pushing control logic onto switches and enable local processing of data plane events to minimize control latency and communication overhead while leveraging SDN's global view to maximize control decision quality. Controllets are interconnected through a message bus supporting the publish/subscribe communication paradigm with specific extensions for content-based message filtering. Publish/subscribe allows for complete decoupling of controllets to further facilitate control plane distribution. Furthermore, we identify crucial requirements for practical on-switch deployments, where we employ lightweight virtualization techniques to ensure a safe control plane operation. We evaluate both, the scalability and performance properties of our architecture, including its deployment on a white-box networking hardware switch.

Recent years have seen an evolution of SDN control plane architectures, starting from simple monolithic controllers, over modular monolithic controllers, to distributed controllers. We observe, however, that today's distributed controllers still exhibit inflexibility with respect to the distribution of control logic. Therefore, we propose a novel architecture of a distributed SDN controller in this paper, providing maximum flexibility with respect to distribution. Our architecture splits control logic into light-weight control modules, called controllets, based on a micro-kernel approach, reducing common controllet functionality to a bare minimum and factoring out all higher-level functionality. Light-weight controllets also allow for pushing control logic onto switches and enable local processing of data plane events to minimize latency and communication overhead. Controllets are interconnected through a message bus supporting the publish/subscribe communication paradigm with specific extensions for content-based OpenFlow message filtering. Publish/subscribe allows for complete decoupling of controllets to further facilitate control plane distribution. We evaluate both, the scalability and performance properties of our architecture, including its deployment on a White Box networking hardware switch.

Recent developments in networking hardware and software-defined networking have enabled full distribution of network control to reduce control latency and increase reliability. However, both, hardware and software of current white-box networking hardware are highly heterogeneous, which limits the deployment and operation of switch-local control applications. Furthermore, switch-local control raises yet unconsidered security concerns. In this paper, we present our concept of in-forward-element processing, which leverages the open access to the control plane of white-box networking hardware to deploy control logic directly onto switches. We combine local control applications with lightweight virtualization to cope with networking hardware heterogeneity and to achieve required isolation properties and ease of management. Beyond distributed network control, we show this scheme is also beneficial for implementing switch-local virtual network functions (NFV), processing packets. Highlighting the practicability of the concepts, we provide an overview of the current white-box networking hardware and software landscape and their compatibility with lightweight virtualization technologies. To this end, we perform an empirical evaluation of NOS-virtualization combinations on such hardware and compare the results with respect to incurring virtualization overhead.

With the increasing popularity of software-defined networking (SDN), ternary content-addressable memory of switches can be directly accessed by a publish/subscribe middleware to perform filtering operations at low latency. In this way, three important requirements for a publish/subscribe middleware can be fulfilled, namely, bandwidth efficiency, line-rate performance, and low latency in forwarding messages between producers and consumers. Nevertheless, it is challenging to sustain line-rate performance in the presence of dynamically changing interests of producers and consumers. In this paper, we realize a scalable, SDN-based publish/subscribe middleware, called PLEROMA, that performs efficient forwarding at line-rate. Moreover, PLEROMA offers methods to efficiently reconfigure a deployed topology in the presence of dynamic subscriptions and advertisements. We evaluate the performance of both the data plane and the control plane of PLEROMA to support our claim. Furthermore, we evaluate and benchmark the performances of SDN-compliant hardware and software switches in the context of our middleware.

Updating a network is an essential and continual task in the management of today's softwarized networks. When applying updates on distributed network elements, desired network properties, such as drop- and loop-freeness, might be transiently violated. Although being crucial, update consistency has yet been less considered in network management. In this paper, we argue for incorporating the particularities of update consistency into the reconfiguration process of continuous network management. We present a generic management architecture allowing for an appropriate selection of an update mechanism and its parameters based on expected inconsistency effects. We investigate update consistency for the case of multicast routing and show in an extensive analysis why simultaneous drop- and duplicate-freeness is not possible. We present an update procedure for multicast routing updates that identifies critical update steps, which are fed back into the reconfiguration process, along with a lightweight approach that allows for the selection of an update strategy, preventing either drops or duplicates. Furthermore, we present an optimization of an existing powerful, but resource-intensive update approach as well as an approach for in-network filtering of duplicates.