WO2011080228A1

WO2011080228A1 - Method for determining or optimizing the availability of a communication network

Info

Publication number: WO2011080228A1
Application number: PCT/EP2010/070705
Authority: WO
Inventors: Philippe Loisele; Jacques Chommeloux; Philippe Berenguer
Original assignee: Thales
Priority date: 2009-12-30
Filing date: 2010-12-23
Publication date: 2011-07-07
Also published as: FR2954876B1; FR2954876A1

Abstract

The method according to the invention relates to calculating the availability of the nodes Ni and/or links Nj between a first node of the network and a plurality of nodes of the network (point to multipoint context). Said method next compares the calculated availability (probability) of the nodes and links by implementing the specific steps of the method according to the invention, using the value of the randomly generated numbers. If the randomly generated number is lower than the availability of the link or node, it is tagged as OK, and therefore in good operating condition (enabling the information to flow), and in the opposite case, it is considered faulty, NOK (not allowing the information to flow). Next, based on the open links and nodes (in the operating state), the method verifies if there is a path leading from the point of origin to all the user-defined end points without exception.

Description

METHOD FOR DETERMINING AND OPTIMIZING THE AVAILABILITY OF A COMMUNICATION NETWORK

The object of the present invention relates to a method and a system for determining the probability of being able to communicate (availability) between nodes in a telecommunications network whose initial architecture is known.

The results of the availability are, for example, stored in a file connected to a device for adapting the architecture of the network, that is to say the number and quality of the nodes and associated links, as needed. a user or a human machine interface according to the network operation requirements previously required by a user.

Definitions used in the description of the invention

The dependability of a communication network is generally based on the following notions:

• Availability: The probability that a component or system is in working order at a given moment t.

· Reliability: This is the probability that a component or system will operate on the time interval [0, t].

• Maintainability: The ability of an entity to be maintained or rehabilitated (consideration of the ability of the support system to maintain or restore the entity is in the area of support integrated logistics).

• Safety: This is the ability of a product to meet, during all phases of life, an acceptable level of risk of accident likely to cause aggression of personnel or a major degradation of the product or its environment. In the field of telecommunications networks, it is essential to know at best the availability or the operating status of components involved in the transmission of data or information. Thus, it is desirable to know the operating status or availability of nodes and links that will participate in relaying information.

Different methods are described in the prior art. Most of these methods use complex tools that require significant computing power.

FIG. 1 represents an example of a network consisting of several nodes Ni interconnected by communication links li, which may be a physical medium or any other medium known and used in the field of wireless communications.

Nodes are defined by:

a unique identifier, represented by a number between 0 and 999 for the nodes.

- an indicator of good functioning, which makes it possible to define whether the element is in a state of good functioning or if on the contrary it is out of order.

- a value of availability in the general sense, ie a value between 0 and 1 which can represent an intrinsic availability, an operational availability, a probability of good functioning over a given time, etc.

- a type (subscriber or transit). Some methods only deal with subscriber nodes (the analytic calculation for example).

- Since the network consists of a logical layer and a physical layer, there is an indicator defining the membership of this or that layer. The links between nodes are defined for example with:

a unique identifier whose minimum number is, for example, 1000.

- availability.

- a weight that actually represents the technical cost of a link.

an indicator that defines the type of the link (physical or logical link). It is known from the prior art to analytically determine the availability of nodes and links by an analytical calculation. It involves analytically calculating the probability of having at least k roads available between two points among n roads. Since the probabilities for the routes to be available are not independent of each other, the corresponding probability formulas must be applied. The probability of a route being available is equal to the product availability of the elements that make it up.

On the other hand, for 2 roads A and B which connect two points Rd and Ra, the probability of having at least one (k = 1) available road between Rd and Ra is noted:

P (AuB) = P (A) + P (B) - P (AnB)

or the sum of the operating probabilities of routes A and B, to which is subtracted the probability that the two roads are running simultaneously.

The generalization is given by the formula of Poincaré (at least 1 among n):

m m j-l m j-l i-l

A (t) = Σp (Pj) -ΣΣ p (pipj) + ΣΣΣ p (PkPiPj) + ... + (-l) ^m p (PLP2 ... Pm) = j = 2 i = lj lj = 3 i = 2 k = 1 where p (Pj) represents the probability that at time t the elements x of the set of routes Pj, of probability q _x (t) are running, therefore p ( Pj) = Y [q _x (t) .Poincaré formula.

XEJ

Note that when calculating the probability of the intersection of the events Ai ("the road i is available" or "all the elements that make up the road i are available") and Aj ("the road j is available" or still "all the elements that make up the road j are available"), it must be taken into account that Ai and Aj are not necessarily disjoint ("roads i and j are not necessarily totally disjointed" or "it can exist elements that belong at the same time to i and j "). The probability of the intersection of the events Ai and Aj is thus calculated ("all the elements of the road i and all the elements of the road j are available" without counting twice the elements which are in i and in j) and not not the simple product of road probability.

To make calculation possible in computing time, it is known to use route filtering. Thus, the selection of the roads is a determining criterion in the results obtained by the analytical calculation. For each of these criteria, the user is asked how many routes he wants to calculate:

· Selection of the most disjointed routes plus the shortest in the sense of the network routing algorithm (RIP)

• Selection of the most disjointed roads

• Selection of the most disjointed routes plus the least heavy ones in the sense of the network routing algorithm (OSPF) · Route selection based on a logical layer

• Selection of a forced road.

The analytical calculation used according to the prior art has the disadvantage of being expensive in terms of calculation time.

Monte Carlo simulation availability calculation for end-to-end calculation

It is also known from the prior art that performing an availability calculation between two nodes of the network by Monte Carlo simulation is a method using the entire network.

To do this, one must first generate a series of shots or simulations. During each shot or simulation, random numbers between 0 and 1 are generated, and a probability law (uniform law) is used. Thus for each simulation step, each node and each link is assigned a randomly generated number. The availability of the nodes Ni and links li is compared with the numbers randomly generated, if these are inferior to the availability we consider that the links or the nodes function correctly. The availability for a computation on a pair of the network is thus given by the relation between the number of successes and the total number of simulation. To check the availability of a path between 2 nodes, we run a deep-path algorithm that will be explained later in the description.

The invention is based on a new approach for calculating availability from one end of the network to several other ends, which therefore takes into account a starting node and one or more destination nodes of the communications network, as well as the links existing between said nodes. .

The invention relates to a method for determining and / or optimizing the availability of a communications network, that is to say the probability of being able to communicate between nodes of said communications network, characterized in that it executes at least the steps following:

Define in Input: a Graph g = G (N, L, p, q), corresponding to the Number of simulations nbSimulations, array of pairs, with N the number of nodes, L the number of links, p the probability of the links p = (pi, p ₂ , P3, ■■■, PN), q the probability of nodes q = (q-ι, q ₂ , q3, ■■■, qi_), and two functions

Use a RandUniformOI () function to generate random numbers, with a random generator.

Use a DFS_lsRouteExist () function that traverses the graph in depth to check the existence of a path between 2 nodes of the graph. Determine Output: An approximation of D availability by performing the following steps

For i ranging from 1 to nbSimulations

// Initialize

X "0

For j ranging from 1 to N

We generate a random number with RandUniformOI () for each Node of the network and we compare it with its probability P _j End for

For k ranging from 1 to L

We generate a random number with RandUniformOI () for each link of the network and we compare it with its probability q _k , initial value of the node

End for

Make

We then look for the existence of a path communicating between the 2 nodes that we want to study with DFS_lsRouteExiste ().

As long as (there are still pairs to process)

If (all pairs of nodes could communicate) then

X - X + 1

End if

End For

Return the availability value D = X / nbSimulations,

Store said availability value for one node to several other nodes in a file.

The method may include a step of determining the overall network availability, defined as the fact that a source node S can communicate with all the other nodes of the network, said step comprising an availability calculation step that all the pairs of network nodes are connected to each other, performing the preceding steps from one source node to all other nodes.

The method may comprise an additional convergence study step using two stop criteria, the first being set according to a number of iterations so that one is not in an infinite loop configuration, the second a criterion taking into account the accuracy of the result, these criteria being parameterized in the code of the software according to the invention, Let X be the value of the number of simulations in the code and Y the interval of the required precision,

The method executes at least 2 times the availability calculation of One to several for the same number of simulations according to the steps of one of claims 1 or 2,

Once the 2 computations carried out one compares the results, if the subtraction of these 2 numbers in absolute value is lower than the precision chosen then the result is valid, one thus stops the computation,

If the precision is not reached, however, then we multiply the number of simulations by a factor a and we start again the calculations to carry out again the comparison,

We stop when one of the stop criteria is reached,

Store said availability value in a file. Other features and advantages of the present invention will emerge more clearly on reading the following description of an exemplary embodiment of the method and of the system according to the invention, given by way of illustration and in no way limiting attached, of the figures which represent:

FIG. 1, an example of networks comprising several connected nodes, 1 T,

FIG. 2, a machine interface developed for the implementation of the method,

• Figure 3, a possible interface integrating a convergence feature,

FIG. 4, an explanatory graph in the case of a global availability calculation, and

FIG. 5, a human machine interface for calculating global availability, and

• Figure 6, a diagram of the steps performed by the implementation of the software according to the invention driven by a user. The method for determining the availability of one end or node of the network to several other nodes will notably use a given number of simulations. For each simulation at each node Ni and at each link lij, a randomly generated number will be assigned.

Hypothesis

For the availability calculation according to the invention "UnVersPlusieurs", no route is selected beforehand, but the method relies on the entire network.

The method according to the invention will be concerned with a calculation of availability of the nodes Ni and / or links lij between a first node of the network and several nodes of the network (point-to-multipoint context).

It will then compare the availability (probability) of the nodes and links calculated by implementing the specific steps of the method according to the invention with the value of the numbers generated randomly. If the randomly generated number is less than the availability of the link or node then it is marked as OK, therefore in a state of good operation (allowing the passage of information), otherwise it is considered as failing, NOK (not letting the information pass).

The method then checks for open links and nodes (in working order) if there is a path from the origin to all user-defined endpoints, without exception.

To verify the existence of a path between two nodes, the method uses, for example, the algorithm for in-depth traversing of a graph known to those skilled in the art, DFS (Depth First Search). This method is based on the marking of neighboring nodes of a given node and thus propagate in the graph. In the exploration, the algorithm tries to go very fast "deeply" in the graph, moving away from the starting vertex.

The search selects at each step a vertex next to the vertex marked in the previous step. If there is a path between the starting point and all the points of arrival the process considers this as a success. Otherwise it is a failure.

Availability is then given by the ratio of the number of hits to the total number of simulations. This gives the availability for one node to several other nodes.

The availability calculation according to the invention is based in particular on the generation of random numbers. It is possible to use as a random number generator the generator (Pseudo Random Number Generator) which is called the 'Marsenne Twister' known to those skilled in the art. This random generator was developed in 1996/97, with an update in 2002 by Mr. Makoto Matsumoto and Takuji Nishimura (Hiroshima University).

Description of the availability calculation step according to the invention of a first node to several other nodes of the network

Useful functions:

o The RandUniformOI () function that generates numbers

thanks to the "Marsenne Twister" generator.

o The DFS_lsRouteExist () function that traverses the graph in depth

to verify the existence of a path between 2 nodes of the graph.

Let G = G (N, L, p, q) be a graph

N: number of nodes

L: number of links

p: probability of the connections p = (pi, P2, P3, ■■■, PN)

q: probability of the nodes q = (q-ι, q ₂ , q3, ■■■, qi_)

Input: A Graph g = G (N, L, p, q), Number of simulations nbSimulations, table pairs.

Output: An approximation of D availability

For i ranging from 1 to nbSimulations

// Initialize

X "0

For j ranging from 1 to N

We generate a random number with RandllniformOI () for each Node of the network and we compare it with its probability P _j End for

For k ranging from 1 to L

We generate a random number with RandllniformOI () for each link of the network and we compare it with its probability q _k , initial value of the node

End for

Make

We then look for the existence of a communicating path between the 2 nodes that we want to study with DFSJsRouteExiste ().

As long as (there are still pairs to process)

If (all pairs of nodes could communicate) then

X - X + 1

End if

End For

Return the availability value D = X / nbSimulations

The found value of the availability D for one node to several other nodes of the network is, for example, stored in a file which is linked to a human machine interface and which makes it possible to display the data found on a console or in a computer. file related to a processor that will allow to interact and modify one or more elements (node, link) from the initial network on which the steps were executed in the case where the availability that has been found is not suitable in view of the requirements given by the network user.

A man-machine interface developed for the implementation of the method according to the invention is shown in FIG.

It includes buttons for launching the method from one end to several others in rapid calculation and with convergence study as detailed below.

Example of the application of the method in relation with FIG.

With different values of nodes and links.

According to an alternative embodiment of the invention, the method may comprise an additional step that allows convergence in the availability calculation results described above.

Figure 3 shows a possible interface integrating this functionality. At the level of this interface we find two functionalities allowing the computation of one towards several with or without convergence, as well as as the ability to do end-to-end availability calculation with convergence study.

Convergence is defined by stopping criteria that can be: the number of simulations and the precision that a user wants.

According to an implementation example, the method comprises a step of studying the convergence of the results obtained in order to refine the results.

To calculate the availability of a communication network with convergence study, it is possible to set, for example, two criteria for stopping the method. The first criterion can be chosen according to a number of iterations in order to avoid being in an infinite loop configuration, a second criterion can take into account the precision of the result. This allows to give a valid result to several digits after the decimal point (the number of digits after the decimal point depends on the accuracy that the user has entered in the field provided for this purpose in the interface- Figure 5).

Since the fields filled by a user at the interface are stop criteria in the code, the initialization values of the algorithm are hard-coded prior to its implementation. We therefore start the calculation, for example, with 1000 simulations, the value 1000 is modifiable only from the source code.

To start the calculation the user will enter a number of simulations greater than 1000 and a precision included in] 0; 0.1].

To ensure accuracy, the availability calculation of "One to many" will be run twice, for the same number of simulations. During the first iteration, the number of simulations and therefore of 1000. Once the 2 calculations are performed, the results are compared. If the subtraction of these 2 numbers in absolute value is lower than the precision chosen then the result is valid, one thus stops the computation. If the precision is not reached, however, then the number of simulations is multiplied by 10 and the calculations are repeated to perform the comparison again.

One stops when one of the stopping criteria is reached (the two can be reached at the same time).

A numerical example of the implementation of the method according to the invention with convergence is given below.

Using the method of calculation of CMCS_Un_Vers_Plusieurs () 10 with study of the convergence according to the method according to the invention the following results are obtained:

Stopping criterion: number of simulations 1,000,000,000 (10 ⁹ ), accuracy 0,00001

^■ Taking into account departure and arrival nodes.

15

Another implementation variant of the method according to the invention concerns the case of availability calculation between all the points or nodes of the network in order to be able to express a global availability of the network. From this value, it will then be possible to check whether this availability corresponds to the requirements required by a user.

Global availability is defined as follows:

Global availability is defined as the fact that a starting node can communicate with all other nodes in the network. 25 s communicates with t ₁ .... t _n -i (for a network with n nodes),

o A node is able to communicate with everyone else. I for all j such that j ≠ i, the pair (η ,, n,) is connected

it is possible to express this more generally by saying that global availability is the probability that any node n _j; n _j communicate with all other nodes. o all the nodes are able to communicate with each other.

For all i, j, such that i ≠ j the pair (n _i; nj) communicates!

These two notions are equivalent in fact:

For all i, j such that i ≠ j, for every j such that j ≠ i, the couple the couple (ni, nj) communicates (n ,, n _j ) communicates

<=>

=> for all i, j, i ≠ j the pair (n ,, ^) thus communicates for i fixed, for all j ≠ i we have well (η ,, η ^ connected.

Let us consider now that for all j, i, j, i fixed (n ,, ^) is connected.

Consider the pair (n _j , n _k ) with k ≠ i, since the pair (n ,, ^) communicates, so there is a path from n _j to n ,, but as n, is connected to all the others nodes of the network then n, is connected to n _k .

Therefore the pair (n _j , n _k ) communicates.

Explanatory Graph Figure 4:

Assuming that node 1 is connected to all other nodes (2, 3, 4). If we want to start from node 2 to reach node 4 then we can easily go through node 1 and then take the same path that node 1 took to go to node 4. It is therefore always possible starting from any other node that 1 to be able to go to all the others to see the different paths traced on figure 4.

The principle of the global availability calculation method is therefore based on the probability calculation that all the pairs of the network are connected to each other.

We can therefore apply the availability calculation method for one end to several others described above, this time from one end to all the others. Indeed, according to the definition of the global availability it is enough to look at that a given node is connected to all the other nodes of the network.

For this variant of overall availability, a man-machine interface possibility is shown in FIG.

This window allows the global computation fast or with study of convergence. When calculating with convergence, it also displays the number of replay being calculated.

FIG. 6 represents the various steps for implementing the method according to the invention.

The steps implemented in the various embodiments described above are for example implemented using the C ++ language or any other programming language that allows manipulation of the different objects; nodes, links, network.

The Human Machine Interface can be an interface with Windows input windows, action buttons, drop-down menus, and pointing devices.

The method according to the invention offers a tool for determining different types of availability in a communication network, easy to use, having a graphical interface. in particular to meet the following needs:

• Participate in design efforts for fixed or low-mobility networks, thanks to announced operational safety magnitudes,

· Justify the choice of a network architecture in the face of operational safety requirements required by users,

Claims

1 - Method for determining and / or optimizing the availability of a communications network, that is to say the probability of being able to communicate between nodes of said communications network, characterized in that it performs at least the following steps:

Define in Input: a Graph g = G (N, L, p, q), corresponding to the Number of simulations nbSimulations, array of pairs, with N the number of nodes, L the number of links, p the probability of the links p = (pi, p ₂ , P3, ■■■, PN), q the probability of the nodes q = (q-ι, q ₂ , q ₃ , ■■■, qi_), and two functions

Use a RandllniformOI () function to generate random numbers. "

For i ranging from 1 to nbSimulations

// Initialize

o

For j ranging from 1 to N

We generate a random number with RandUniformOI () for each node of the network and we compare it with its probability Pj

End for

For k ranging from 1 to L

We generate a random number with

RandUniformOI () for each link in the network and compare it with its probability qk, the initial value of the node

End for

Make We then look for the existence of a path communicating between the 2 nodes that we want to study with DFS_lsRouteExiste ().

As long as (there are still pairs to process)

If (all pairs of nodes could communicate) then

X - X + 1

End if

End For

Return the availability value D = X / nbSimulations,

Store said availability value for one node to several other nodes in a file.

2 - Method according to claim 1 characterized in that it comprises a step of determining the overall availability of the network, defined as the fact that a source node S can communicate with all the other nodes of the network, said step comprising a an availability computation step that all node pairs of the network are connected to each other, performing the steps of claim 1 starting from a source node to all other nodes.

3 - Method according to one of claims 1 or 2 characterized in that it comprises a further step of convergence study using two criteria stop, the first being fixed according to a number of iterations for the l one is not in an infinite loop configuration, the second a criterion taking into account the precision of the result, these criteria being parameterized in the code of the software according to the invention,

Let X be the value of the number of simulations in the code and Y the interval of the required precision,

The method executes at least 2 times the availability calculation of One to several for the same number of simulations according to the steps of one of claims 1 or 2, Once the 2 computations carried out one compares the results, if the subtraction of these 2 numbers in absolute value is lower than the precision chosen then the result is valid, one thus stops the computation,

One stops when one of the stop criteria is reached.