Specific features of the practical implementation of observation planning in systems for monitoring networks with dynamic topology

. The article is devoted to the issues of practical implementation of planning observations of information processes flows, the sources of which are dynamic network objects that can appear in the network at random times. A variant of the implementation of the suboptimal distribution of time resources of the monitoring system between the tasks of detecting and evaluating random processes that appear in accordance with the laws of the Poisson flow is proposed.


Introduction
In many branches of modern technology, information surveillance systems are of great importance [1][2][3][4][5][6][7]. Typical tasks that such systems solve are the tasks of monitoring communication networks [4], ensuring the security of information in computer systems and networks [6], monitoring the trajectories of air and ground transport [2,3,5,7], etc. Optimization of the operating modes of such systems can significantly increase the efficiency of their use.
Of practical interest are the issues of observation management, taking into account the probabilistic nature of the appearance of information processes [1][2][3][4][5][6][7], since this allows you to remove a number of severe limitations inherent in classical algorithms for planning an experiment. It is these operating conditions that correspond to monitoring systems for networks with a dynamic topology, in which network objects can appear at random times. Such objects are sources of information processes that can be described by dynamic models [4], and, therefore, the classical mathematical apparatus for controlling observations can be applied to such problems [8][9][10][11].
Of great importance is the model that describes the regularity of the appearance of the evaluated processes at the input of the information system. Within the framework of the article, it is assumed that information processes appear in accordance with the laws of the Poisson flow. In [1] the mathematical formulation of such a problem and the procedure for its solution are presented in detail. At the same time, it seems relevant to study the features of the practical application of observation plans obtained as a result of the implementation of the procedure given in [1].
Suppose that as a result of the implementation of the iterative procedure [1], an observation plan is synthesized (Fig. 1 a). This plan is a priori, and involves the observation of two processes in accordance with the given probabilistic characteristics of the moments of occurrence of such processes at the input of the information system. The probability of the appearance of the third process during the observation interval is so small that the program function associated with it does not affect the formation of the observation plan [1]. However, in real observation mode, the third and other information processes may appear, i.e. the obtained a priori plan can be used only as a reference plan, which must be corrected directly during observations.
It should be noted that the approach proposed in [1] can be used in information systems with time division of resources, ie. at the same moment in time, either one of the information processes can be evaluated, or new processes can be detected. The allocation of time resources in such an information system occurs by allocating quanta of machine time, and the transition from one task to another can be carried out only after the completion of the next quantum ( Fig. 1 b).
It is assumed that the minimum frequencies of using quanta, spent on detection min s f and on estimation of one information process min о f , are determined, at which a satisfactory solution of the problems of detection and estimation is provided. The surplus resource is determined by the expression *, where ex f is the repetition rate of quanta available for suboptimal distribution between tasks; max f -the maximum repetition rate of quanta, determined by the characteristics of the information system; (2) I -the number of detected processes at the current time (the meaning of parameter will be explained below). Let us consider in more detail the procedure for adjusting the a priori plan directly during the observation process. [ , ) t t t  the detection of the 2nd process occurred, then starting from this moment in time the resource of the information system is distributed between the tasks of evaluating the 1st and 2nd information processes and detecting the following processes. In this case, the number of quanta used for detection is determined by frequency min s f , the number of quanta used for evaluating the 2nd process is determined by min о f ("minimum" estimation mode), the rest are spent on evaluating the first process. This is due to the fact that according to the a priori plan (Fig. 1 a), the interval 01 [ , ) tt is defined as the interval of priority assessment of the first information process.
Let's call this mode "priority".
If one or more information processes are detected on interval 01 [ , ) tt , then all of them will be evaluated in the "minimum" mode, i.e. the frequency of the allocated quanta will be  (4) quanta, the estimation of all processes, except the first, will be spent 1 0 quanta. The evaluation of the first process in priority mode will use all the remaining к tt , according to the a priori plan, the second process is accompanied in the "priority" mode, the rest in the "minimum" mode. The distribution of resources in this case is similar to that considered above, and expressions (2) -(5), respectively, take the form where r is the number of quanta that fit in the interval 1 [ , ] к tt .

Fig. 2. Graph of abstract Mealy automaton.
For the practical implementation of correcting the observation plan, we use the theory of discrete automata [8]. Figure 2 shows the graph of an abstract Mealy automaton, which implements the adaptation of the a priori plan (Figure 1 a) to real conditions directly during observations. Figure 2 introduces the following notations: informational mode in the "priority" mode, and the 1st -in the "minimum" mode; ' n A -the state of the automaton corresponding to the evaluation of the 2nd information process in the "priority" mode, and the rest of the processes (1st, 3rd, ..., nth) -in the "minimum";