Modern Network technologies 

Abstract: The architecture and organization of the high availability distributed control plane for SDN/OpenFlow networks are considered. High availability is achieved by redundancy of controller instances, active switch-controller communications, computing resources and tools for a controller instance failure and overloading detection and recovery. The proactive backup controller allocation algorithm which allows to minimize the time to repair in the case of a single controller instance failure is discussed. The algorithm for controllers' load-balancing allows dynamically reconfigure the control plane with a minimum number of switch control transfer operations to avoid controller instance overloading. The initial experimental results of the proposed algorithms for the HA distributed SDN control plane are described.

Abstract: Network function virtualization (NFV) is a promising technique for high quality, flexible and scalable service for telecommunication companies' clients and for enterprise data center clients. One of the important capabilities of this technique is providing a virtual service as a combination of multiple virtual functions. There are two types of virtual functions: those intended for a single customer (su -VF) and those that can serve multiple users (mu-VF). In case when output of mu-VF is chained with inputs of several different su -VFs, there is a need for a mechanism of identification and separation of users' network flows passing through mu -VF to allocate them correctly between inputs of su-VFs in the NFV infrastructure. In a cloud environment, it is not always possible to use VLAN tags, IP and MAC addresses for that. In this paper, we consider the problem of identification of network traffic coming from a certain user inside an NFV platform and present a solution implemented in C2 MANO-platform.

Abstract: The optimal control problem of traffic flows in the city road network is considered. The control is performed by means of traffic lights at regulated intersections. The road network is presented as a directed graph of variable structure. Nodes of the graph are road sections and edges of the graph are maneuvers from one section to another. The maneuvers of traffic are performed by certain traffic light phases. The statement of the optimal control problem for traffic flows is nontrivial due to the complexity of the object and a large number of uncertainties that influences its behavior. The complexity of an object is caused by the absence of a mathematical model that includes the control and is capable of representing the dynamics of an object. The object is exposed to uncertainties of the data collected from the road infrastructure and vehicles, influencing each other, and has a few evident functional dependencies. In this study for a generalized description of traffic flows, a universal recurrent model based on the controlled networks is considered. The developed model made it possible to formulate the optimal control problem in the classical statement. Evolutionary algorithms are used to solve the optimal control problem, in particular, to determine the duration of the traffic light phases in accordance with the control quality criteria. An example of solving the optimal control problem for a group of intersections in Moscow is given.

Abstract: In software-defined networks key network management functions are concentrated in the network operating system (or SDN controller) with a set of network applications, which is running on a dedicated server. Controller supports and monitors an actual global network view that includes the state of network devices. Based on global network view, controller is carrying out the logically centralized management of network devices and data flows in the network. In this paper we are considering the problem of controller instance protection from DDoS attacks in distributed control plane of software-defined networks. We propose a protection method based on monitoring and analyzing host behavior in the network. Method controls the dynamics of the malicious activity of infected hosts and changes in the network topology. The proposed method allows detecting a DDoS attack, suspending the attack and eliminating the consequences of the attack in the network. We have implemented the proposed method as an application for the RUNOS SDN/OpenFlow controller.