No abstract

In SDN network based on OpenFlow a controller performs logically centralized control of enterprise network infrastructure, network policies, and data flows. At the same time the controller is a single point of failure which can cause a very serious problem (e.g. network outage) for network reliability and production use cases. To address this problem, we consider different active/standby strategies to provide a controller failover in case of controller failure. We propose a high-available controller (HAC) architecture, which allows to deploy a highavailability control plane for enterprise networks. We develop a HAC prototype to demonstrate the efficiency of our solution and also describe initial experimental results. A. Typical SDN controller architectureA typical SDN/OpenFlow controller [6, 9, 10 and 11] includes:Controller core which handles and supports connectivity with switches and translates control protocol messages (e.g. OpenFlow) into internal controller events and vice versa.Controller network services which control, form and monitor network view, states of network devices, provide an interface (Northbound API) for controller applications. Usually network services include event dispatching, device managing, topology managing and the others.Controller network applications which configure network infrastructure and manage data flows to solve some business use cases.

In this paper, the reliability allocation problem (RAP) for real-time avionics systems (RTAS) is considered. The proposed method for solving this problem consists of two steps: (i) creation of an RTAS simulation model at the necessary level of abstraction and (ii) application of metaheuristic algorithm to ¦nd an optimal solution (i. e., to choose an optimal set of fault tolerance techniques). When during the algorithm execution it is necessary to measure the execution time of some software components, the simulation modeling is applied. The procedure of simulation modeling also consists of the following steps: automatic construction of simulation model of the RTAS con¦guration and running this model in a simulation environment to measure the required time. This method was implemented as an experimental software tool. The tool works in cooperation with DYANA simulation environment. The results of experiments with the implemented method are presented. Finally, future plans for development of the presented method and tool are brie §y described.

No abstract