11.01.2014 Views

ANNUAL REPORT 2012

ANNUAL REPORT 2012

ANNUAL REPORT 2012

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

comparatively more frequent instances where<br />

there is a negative mismatch between demands<br />

for and supply of communication resources. It is<br />

expected that resources can be allocated in a<br />

manner where some users and services, that have<br />

been identified as particularly important,<br />

experience less service disruptions in these<br />

situations.<br />

Traditional network management systems are based on<br />

a manager-agent architecture where each network element<br />

contains an agent that provides software component<br />

interfaces, which allow for remote configuration and<br />

extraction of variables. A centralized manager<br />

periodically polls the agent of a network element to get or<br />

set a value of different variables’. An agent of a network<br />

element can also asynchronously send a message to the<br />

manager, e.g. reporting the occurrence of a fault.<br />

Traditional network management systems are managed<br />

manually by a network system operator via some<br />

graphical interface. The wireless networking<br />

environments that are of primary interest in this research<br />

project, characterized by the properties described earlier,<br />

bring several added challenges to traditional, centralized<br />

network management paradigms. Such traditional<br />

network management paradigms often rely on centralized<br />

and human-controlled management decisions propagated<br />

to network elements, which are clearly not adequate as a<br />

centralized entity can never be expected to have and<br />

analyze the network state information that is necessary to<br />

make informed network management decisions. This<br />

proposed research primarily addresses two management<br />

aspects, performance management and security<br />

management, coupled to a support tool that is used during<br />

the provisioning, operation and evaluation of a cognitive<br />

network on a per-mission basis. We will here and in the<br />

following research areas use the term Cognitive Network<br />

Management System as an abstraction for this software<br />

tool.<br />

3. Goal<br />

The goal is to give an architectural description of a<br />

cognitive radio network management system in the<br />

context of network and spectrum management of next<br />

generation military tactical communication system. The<br />

main capabilities that are focused on are: simplify and<br />

shorten the mission preparation and configuration phases<br />

for national and multi-national deployments of tactical<br />

networks, improve tactical network performance by<br />

enabling dynamic spectrum and network management,<br />

and improve spectrum utility to maximize the use of<br />

military spectrum allocations. Some identified important<br />

aspects are: security aspects of cognitive network<br />

management, capabilities for local processing of goals,<br />

monitoring of the local environment, reaction to<br />

contextual events by self-configuration and information<br />

propagation for interactions with and between CNMS<br />

components and network elements, objective functions<br />

and their interrelations to measuring and control<br />

parameters. As a reference point for describing the<br />

architecture a waveform is described and analyzed from<br />

the perspective of measures, control parameters, behavior,<br />

and their relation to objective functions and mission based<br />

utility. Identification of important parameters, their<br />

behavior and relations are conducted by smaller<br />

simulation (small world isolation) and implementation<br />

studies included in the project. The approach for<br />

designing these experiments is based on requirement<br />

extraction from available scenarios and user cases. The<br />

main focus is directed towards the goal of indentifying a<br />

feasible architecture for cognitive network management<br />

system, i.e. its structural design and its behavior..<br />

4. Accomplishments<br />

The proposed system architecture is a multi-tier structure,<br />

where individual tiers can operate autonomous without<br />

overlaying tiers. These tiers reflect the horizontal structure<br />

in a network of networks. It includes everything from<br />

individual parameters of a protocol executing on a<br />

platform to higher level policy respiratory. The<br />

management interfaces, figure 1, towards the waveform is<br />

inherited from architecture derived in the ARAGORN<br />

project [3]. The low level control and learning, blue box<br />

in figure 1, is representing all algorithms that control and<br />

optimize the operation of the system in short term. It is a<br />

very big challenge to truly understand all these control<br />

parameters in waveform, how they are dependent on each<br />

other and their time constant. In order to get a better<br />

understanding of such control parameters, mechanisms,<br />

and their behavior the project especially have looked at<br />

evolutionary algorithms for resource assignment [4]-[6]<br />

(typically NP hard problems), were algorithmic stability<br />

(robustness), algorithmic convergence and time<br />

complexity especially been studied for centralized<br />

algorithms. The next step is to look at the same algorithms<br />

with decentralized implementation.<br />

Figure 1. Functional modules and interfaces.<br />

Furthermore, to understand and to be able to model the<br />

behavior of the mechanisms a theoretical benchmark<br />

CERES Annual Report <strong>2012</strong><br />

29

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!