MSN: A Playground Framework for Design and Evaluation of MicroServices-Based sdN Controller

Abstract

Software-defined networking decouples control and data plane in softwarized networks. This allows for centralized management of the network, but complete centralization of the controller functions raises potential issues related to failure, latency, and scalability. Distributed controller deployment is adopted to optimize scalability and latency problems. However, existing controllers are monolithic, resulting in code inefficiency for distributed deployment. Some seminal ongoing efforts have been proposed with the idea of disaggregating the SDN controller architecture into an assembly of various subsystems, each of which can be responsible for a certain controller task. These subsystems are typically implemented as microservices and deployed as virtual network functions, in particular as Docker Containers. This enables flexible deployment of controller functions. However, these proposals (e.g., $$\mu$$ μ ONOS) are still in their early stage of design and development, so that a full decomposition of the SDN controller is not been available yet. To fill that gap, this article derives some important design guidelines to decompose an SDN controller into a set of microservices. Next, it also proposes a microservices-based decomposed controller architecture, foreseeing communications issues between the controller sub-functions. These design and performance considerations are also proven via the implementation of the proposed architecture as a solution, called Micro-Services based SDN controller (MSN), based on the Ryu SDN controller. Moreover, MSN includes different network communication protocols, such as gRPC, WebSocket, and REST-API. Finally, we show experimental results that highlight the robustness and latency of the system on a networking testbed. Collected results prove the main pros and cons of each network communication protocol and an evaluation of our proposal in terms of system resilience, scalability and latency.

Design and Implementation of Print Advertisement Based on Computer Image Interactive Virtualization

“…Because the high demand for accuracy means that more sophisticated processing equipment needs to be introduced, the transmission and storage of image data and the corresponding postprocessing are becoming increasingly difficult. In addition, the sharp increase in the number of slices and the more complicated cutting process will also greatly increase the cutting error rate [11]. erefore, researchers often choose a reasonable compromise scheme to balance the accuracy of slice image and the requirements of current research on image accuracy [12].…”

Methods to Synchronize Data in a Microservice Architecture

“…Each message from the map is passed to the Message Processor, where it is processed, in parallel with the processing of other messages. The size of the Thread Pool Task Executor is set via the settings file [32][33][34].…”

A New Agent-Based Intelligent Network Architecture

“…Moreover, monolithic SDN deployment does not allow dynamic management of SDN components and/or functionalities to (de-)activate according to the scenario. To alleviate this, the work in [5] and [6] decomposed the SDN controller and implemented it as a set of microservices components that can run in a distributed fashion with flexibility in recomposition. In particular, the SDN controller is designed as a composition of logical sub-functions, i.e.…”

“…3GPP specifies REST APIs as the standard-de-facto for services intercommunication. In this regard, we have implemented a microservices-based SDN controller deployment that complies with the 5G standard, see Figure 1 (top left) [5]. Our implementation is based on Ryu SDN Framework and a first release is available to the community at the link: https: //gitlab.com/dscotece/ryu sdn decomposition/.…”

“…As a guideline in designing the agent-based system, we show a decomposed SDN controller [5], designing the components as autonomic agents, and recomposing them to create the intelligent controller system. The main principle to retain in decomposing an SDN controller is that the network information and state should be synchronized and selfconsistent.…”