Machine learning methods in embedded systems

Within this direction, the following methods are being developed:

• methods for constructing feedforward neural networks for embedded systems
• methods of “separate training” of feedforward neural networks for working with large databases
• an axiomatic approach for detecting pre-emergency operating modes of complex technical systems

Network Processor Architecture

Research directions in this area are as follows:

• Optimization of network processor software architecture
• Research on methods for processing packet headers in the stages of a network processor pipeline
• Research on methods for representing flow tables in a software-defined network switch
• Research on approaches to organizing search in a network processor architecture without a dedicated associative device
• Research on stateful packet processing algorithms

SDN Security

Research directions in this area are as follows:

• Development of system for controlling applications access to controller resources
• The analysis of SDN protocols security
• Development of intrusion detection systems
• Detection of compromised switches in SDN

Distributed Controlplane in SDN

Research directions in this area are as follows:

• Development of algorithms and tools for reliability, fault tolerance and high availability of SDN distributed control platform; development of recovery algorithms for controller failure, control channel failure and inter-controller connection loss.
• Analysis of distributed transactions algorithms, development of controller state synchronization algorithms, voter and leader election algorithms.
• Development of methods for predicting load and ensuring load balancing between SDN controllers
• Development and research on the application of artificial intelligence methods for the management and configuration of the distributed control plane for SDN networks.
• Development of methods for detecting and countering DDoS attacks on the distributed control plane.

Additional Functionality at the Data Transmission Level for Programmable Network Devices (Programmable Dataplane)

Modern switching equipment provides opportunities for programming additional logic for packet processing. This allows a switch to perform some tasks without the controller’s involvement, thereby ensuring a higher degree of responsiveness to changes in the network.

Within this direction, the following works are being carried out:

• Monitoring channel activity using the BFD protocol
• Research on ways to improve the performance of a software-defined switch

Development of Network Applications for the SDN Controller (SDN Applications)

SDN offers extensive opportunities for creating innovative applications that implement previously inaccessible logic in traditional networks.

Within this direction, the following works are being carried out for the RUNOS 2.0 controller:

• Development of applications for managing IoT device traffic in SDN/OpenFlow networks.
• Development of routing applications in SDN/OpenFlow networks.
• Development of applications for monitoring the state of the data transmission contour and control contour in SDN/OpenFlow networks.
• Development of core controller applications to support the PCEP protocol for interaction with the controller.
• Development of core controller applications to support the NETCONF protocol for interaction with the controller.
• Development of load balancing applications in the data transmission contour of SDN networks.
• Development of applications for ensuring fault-tolerant switch connection to the controller through the data transmission contour.
• Development of applications for detecting and countering DDoS attacks on the controller.

Development of New Programming Languages for SDN (SDN Programming)

Within this direction, the following works are being carried out:

• Development of new abstractions for programming network applications in SDN
• Development of an automatic rule generation system in SDN
• Development of a conflict resolution system between network applications in SDN
• Работа с гетерогенным коммутационным оборудованием (не полная поддержка OpenFlow, различный конвейер обработки пакетов, различающие таблицы правил и максимальное число правил Work with heterogeneous switching equipment (incomplete OpenFlow support, different packet processing pipelines, different rule tables, and the maximum number of rules)

Intelligent Network Interaction Systems in Heterogeneous Networks

Today, the Internet of Things (IoT) consists of loosely connected disjointed networks. For example, modern cars operate several networks at once: one manages engine operation, another — safety systems, a third supports communication, etc. In office and residential buildings, many networks are also installed for heating, ventilation, air conditioning, telephone communication, security, lighting management. As the IoT evolves, these and many other networks will be connected to each other and gain broader capabilities in security, analytics, and management.

Network Modeling and Prototyping Systems

When researching various properties of computer networks, simulation modeling tools are often used. The necessary detail of the simulation model of a computer network depends on the goals of modeling and is determined by the researcher when preparing the simulation experiment. The detail and accuracy of the simulation model depend on the choice of the level of abstraction of the object of modeling, as well as the choice of the mathematical apparatus in terms of which the model is constructed.

The laboratory is developing an approach to building simulation models based on lightweight virtualization technology, which allows for effective scaling of the computer network model, as well as reducing labor costs for its calibration and identification.

Technologies for Organizing and Managing Computations in a Cloud Environment

Network Functions Virtualization (NFV) is a concept of separating network functionality and the equipment that implements it through virtualization technology of physical resources.

NFV technology allows, through the virtualization of physical resources (computational, network, and storage data), to programmatically implement the necessary functionality on standard equipment. Thus, the logic of the service is made independent of the hardware on which it is executed. The engineering of “virtual network function” (VNF) depends on the goals for which the network function virtualization infrastructure is built, who builds this infrastructure, and for what purposes.

Examples of network functions virtualization include services for analyzing, managing, and engineering network traffic. For example, for telecom operators, a virtual network function is an entity that implements the functionality of specialized software and hardware network devices (so-called appliances) for switching, routing, filtering, balancing, and processing traffic. Other examples include IP telephony, video conferencing, EPC, billing, DPI (Deep Packet Inspection), traffic engineering, and monitoring, etc.

Real-Time Information and Control Systems

Real-time information and control systems (RTIC) have the following specifics:

• Integration of RTIC with the controlled object,
• Execution of application programs in real time,
• High reliability requirements,
• Strict limitations on the mass and size characteristics for onboard RTIC,
• Limited operator involvement in the operation of RTIC.

Data Center Resource Allocation Algorithms

The efficiency of cloud platforms and the use of physical resources of data centers largely depends on the algorithm for mapping requests to physical resources of the data center. The proposed approaches to improving the efficiency of using physical resources of the data center are based on expanding the functionality and increasing the accuracy of the algorithm for mapping requests to physical resources, used in the cloud platform scheduler. Various classes of algorithms are considered for mapping virtual resources to physical resources of the data center.

Adaptive Communication

One of the most relevant research directions in the field of computer networks is the development of intelligent network management methods that would improve its performance by more rational use of available resources and optimization of network operation for specific application tasks. Such optimizations often have a fundamental importance, as they can create a competitive advantage, for example, in organizing cloud computing, streaming broadcasting using content delivery networks, building interactive online services and games, consolidating sensors and actuators that form the basis of the Internet of Things technology.