CN103684864B - Communication network vulnerability analyzing system for large-scale area fault and working method of communication network vulnerability analyzing system - Google Patents

Abstract

The invention discloses a communication network vulnerability analysis system for large-scale regional faults and a working method thereof. Through the network vulnerability analysis system, it is possible to evaluate the impact of communication network topology design and network wiring on geographical location-related regional faults such as natural disasters (such as earthquakes, floods) and man-made damage (such as purposeful EMP attacks, fishing boat trawling), and found that Statistical behavioral characteristics of the network in the case of regional faults, including the average faulty link capacity, end-to-end traffic changes and other performance. On this basis, the network vulnerability analysis system uses the topology information of the physical network to locate the fault area that has the greatest impact on the network, and then guides the network protection design. The network vulnerability analysis system provides a visual interface environment, which can clearly indicate the vulnerability of the system, and help communication network designers and maintainers to plan ahead and build more robust network services.

Description

Communication Network Vulnerability Analysis System and Its Work for Large-Scale Area Failures method

technical field

The invention relates to network vulnerability analysis, especially analyzing the behavior characteristics of physical network topology under large-scale area faults, and simultaneously locating the potential fault area position that has the greatest influence on the network.

Background technique

Computer networks have become the main means of communication in today's society. As people's dependence on the network increases, users have higher and higher requirements for network reliability. On the other hand, due to the rapid development of the network itself, the impact of large-scale regional failures, such as natural disasters and man-made damage on the network is becoming increasingly prominent, and its frequency and degree of damage are on the rise, which has become a major problem that cannot be ignored affecting network reliability. one.

Network vulnerability analysis can help designers and maintenance personnel understand the degree of performance change of network systems under attack and failure conditions, thereby guiding us to design upgrades and operation and maintenance management of network systems. At present, the general network vulnerability analysis and protection work mainly involves the logical topology of the network, and only for a small number of network link and node failures. By observing the changes in data flow in the case of a limited number of network equipment failures, key equipment can be located.

How to effectively analyze the impact of large-scale area faults on network performance in the existing work under the physical network topology has not yet found a feasible solution. This is mainly due to the fact that large-scale regional faults have strong geographical correlation, suddenness and multiple fault characteristics. The location, shape and extent of failures are unknown, and the impact of network equipment on failures is difficult to predict. Simply following the traditional logical network analysis method and using a deterministic regional fault model will over-simplify the vulnerability analysis of the network and fail to reflect the important characteristics of regional faults, resulting in the failure of network recovery strategies or over-investment in network protection. Therefore, it is necessary to design an effective communication network vulnerability analysis system for large-scale regional failures such as natural disasters.

Contents of the invention

The present invention provides a communication network vulnerability analysis system and its working method for large-scale regional faults. Through the network vulnerability analysis system, it can evaluate the communication network topology design and network wiring related to geographical location such as natural disasters and man-made damage. The impact degree of regional faults, discover the statistical behavior characteristics of the network in the case of regional faults, and use the topology information of the physical network to locate the fault area that has the greatest impact on the network.

The present invention provides a communication network vulnerability analysis system for large-scale regional faults, which analyzes the impact of large-scale regional faults on the communication network and locates the fault location that has the greatest impact on the network, and is characterized in that: the system includes

Regional fault model, which simulates the real-world physical network topology. The topology includes link capacity, geographical location information of nodes and links, and simulates common regional faults, such as earthquakes and hurricanes;

The calculation and analysis module calculates and analyzes the number and capacity of network links being interrupted, the flow change between end-to-end node pairs, the probability and average value of the main path and backup path being cut off at the same time for end-to-end node pairs;

The GUI module receives the parameter input of the regional fault model, such as model selection and variable value setting, and sends it to the calculation and analysis module, and the calculation and analysis results, that is, the most vulnerable areas in the network, are marked in a visual way.

The present invention also provides a working method of a communication network vulnerability analysis system for large-scale regional faults, which includes the following steps:

1) First, simulate the physical network topology in the real world to build a regional fault model. In this fault model, the network facilities in the faulty area will be destroyed with a certain probability p. The probability of damage will vary with the distance from the disaster center and the size of the disaster area;

2) Divide the two-dimensional geographic plane in the fault model according to the physical network topology to form a series of grids, considering that the regional faults in each grid have the same impact on network equipment, and then use each grid as the fault center , taking the given index Δ as the standard to measure the network performance, first calculate the impact Δn on the network performance index when the regional fault occurs in each grid, and then use the knowledge of geometric probability to calculate the network performance index by the following formula,

Δ=∑ _n (area(n)/all area)Δ _n

Among them, area(n) represents the area of the nth grid, and all area represents the geographical area of the network deployment;

3) Form three standard quantities to measure network vulnerability:

(1) Disconnected link capacity, that is, DLC: the average value of the disconnected link capacity when a regional fault occurs. When the link capacity is a unit value, it indicates the average number of disconnected links;

(2) Reduced flow between point pairs, that is, PTR: the average value of data flow reduction between specified node pairs when a regional failure occurs;

(3) The probability of disconnection between point pairs, that is, PDP: when a regional fault occurs, the probability of simultaneous disconnection of the working path and the protection path between the specified node pairs.

4) The process of locating the fault interval that has the greatest impact on the network, specifically: based on the three standard quantities calculated above to measure the vulnerability of the network; sort these standard quantities from large to small, and rank the corresponding small squares in the front Collections constitute the most vulnerable area of the network.

The regional fault model described in step 1) includes a concentric circle probability regional fault model for simulating earthquakes, and a line segment fault probability model for simulating hurricanes and trawl damage to optical cables, wherein:

The structure of the concentric circle probability area fault model is:

(1) M concentric circles with radii of m r, m=1,..., M are divided into M rings, the central ring is also a circle, and each ring is a uniform disc-shaped probability Area faults, where r is the pitch;

(2) In the m-th disc-shaped concentric ring fault area, the probability of a link with an arbitrary short length δ to fail is q _m δ, where q _m represents the failure probability of the m-th area;

(3) Since the degree of damage decreases with the increase of the distance from the source point, the parameter q in the ring decreases monotonically from the inside to the outside, that is, q ₁ >q ₂ >…>q;

(4) The longer the length of a link passing through the fault in the concentric circle probability area, the greater the probability of being damaged. Assuming that the length of the link passing through the concentric circle probability area fault is l, its failure probability p is:

In particular, when q=0, p=0, which means no regional fault occurs; when q=+∞, p=1, which degenerates into a deterministic model;

The structure of the line segment probability area fault model is:

(1) The part of the network cut by the fault in the linear area will be completely destroyed;

(2) The length 2r of the fault in the linear region is fixed, but the direction is random;

(3) Note that the length of the linear fault is 2r, the distance between its center point and the link is x, and the angle formed by the normal of the link and the linear fault is α. Assuming that the direction of the fault in the line segment probability area is uniformly distributed, then The probability p that the link is interrupted is

p=2α/π=(2/π)arccos(x/r)

The specific process of step 2) is:

Assuming that the network with E edges is divided into N grids, the calculation process is as follows:

Step1: Take a certain grid as the regional fault center, assuming that the fault model and its parameters have been set by the user, calculate the probability that each edge is affected by the fault. After all the probability calculations are completed, an N*E fault probability matrix is formed. , set to failure_prob[][], failure_prob[i][j] indicates the probability that the jth edge is destroyed with the i-th grid as the failure center;

Step2: Taking each given network performance index Δ as the evaluation standard, calculate the impact on the network with any grid as the fault center.

The specific process of step 4) is:

(1) Calculation method of DLC: first calculate the probability that the link in the fault area is interrupted, and then use this probability as the weight to calculate the weighted sum of the link capacity.

(2) Calculation method of PTR: use Suurballe algorithm to find all non-repetitive path sets between point pairs, and calculate the weighted sum of link capacities of these path sets.

(3) Calculation method of PDP: use Suurballe algorithm to find two shortest non-repetitive paths between point pairs, and calculate the weighted sum of the link capacities of these two paths.

The Suurballe algorithm is a classical algorithm for finding a path set with K non-repetitive edges between specified two points and a minimum total length.

For all the grids, they are sorted by DLC, PTR, and PDP from large to small, and the top 0.1% grids are selected as the final result, that is, the most vulnerable area in the network.

The present invention has following beneficial effect:

Through the network vulnerability analysis system, the present invention can evaluate the influence degree of communication network topology design and network wiring by regional faults related to geographical location such as natural disasters and man-made damage, and find the statistical behavior characteristics of the network in the case of regional faults, including average fault Link capacity, end-to-end traffic changes and other performance. On this basis, the network vulnerability analysis system uses the topology information of the physical network to locate the fault area that has the greatest impact on the network, and then guides the network protection design. The network vulnerability analysis system provides a visual interface environment, which can clearly indicate the vulnerability of the system, and help communication network designers and maintainers to plan ahead and build more robust network services.

Compared with the theoretical analysis of network vulnerability based on graph theory, the mesh division calculation has wider adaptability, and is simpler and more effective in dealing with the complex topology of the actual network; compared with the deterministic regional fault model, the probabilistic fault model is better It perfectly simulates the characteristics of physical destruction and has higher practicability.

Description of drawings:

Figure 1 is a schematic diagram of a physical network topology.

Fig. 2 is a fault model diagram of concentric circle probability area when M=3.

Figure 3 is a diagram of the line segment probability area fault model.

Figure 4 is a schematic diagram of grid division.

Figure 5 is a schematic diagram of the grid and network vulnerability analysis system.

detailed description

1. System input

Figure 1 shows that the input of the network vulnerability analysis system is the real-world physical network topology, rather than the network logical topology. The topology includes link capacity, geographical location coordinate information of nodes and links. In addition, the analysis system also needs to input the size and probability parameters of the large-area regional fault under investigation. For example, for a concentric circle probabilistic area fault model that simulates natural disasters such as earthquakes, the data parameters include the area radius and the failure probability of a unit link within a specified fault range, and this information can be obtained through historical data. For the line-segment probability area fault model for simulating natural and man-made disasters such as hurricanes and trawls, the data parameter is the length of the line segment to be investigated.

There are three standard quantities for measuring network vulnerability in the present invention:

(1). Disrupted link capacity (Disrupted Link Capacity, DLC): When an area fault occurs, the average value of the disconnected link capacity. When the link capacity is a unit value, it indicates the average number of disconnected links.

(2). Pairwise Traffic Reduction (PTR): When an area failure occurs, the average value of the data traffic reduction between the specified node pair.

(3). Pairwise Disconnection Probability (PDP): When an area failure occurs, the probability that the working path and the protection path between the specified node pairs will be disconnected at the same time.

The present invention mainly comprises following three components:

a). Regional failure model

In order to more accurately simulate large-scale regional faults in the real world, the present invention designs two kinds of probabilistic regional fault models according to the above-mentioned examples--concentric circle model and line segment model, to simulate natural disasters such as earthquakes, floods, trawls and man-made damage respectively impact on the network. In the probabilistic failure model, the network facilities in the disaster area will be destroyed with a certain probability, and the probability of being destroyed will vary with the distance from the disaster center and the area of the disaster area.

The concentric circle probability area failure model is shown in Figure 2, and has the following characteristics:

(a). M concentric circles with radii of m r, m=1,..., M are divided into M rings (the central ring is also a circle), and each ring is a uniform disc Shaped probability area faults, where r is the pitch.

(b). In the m-th disc-shaped concentric ring fault area, the probability of failure of a link with length arbitrarily short δ is q _m · δ, where q _m represents the failure probability of the m-th area.

(c). Since the degree of damage decreases with the increase of the distance to the source point, the parameter qm in the ring decreases monotonically from the inside to the outside, that is, q1>q2>…>qM.

(d). The longer the fault length of a link passing through the concentric circle probability area, the greater the probability of being damaged. Assuming that the length of the link passing through the concentric circle probability area is l, the probability p of its failure is:

In particular, when q=0, p=0, which means no regional fault occurs; when q=+∞, p=1, which degenerates into a deterministic model.

The line segment probability regional fault model is shown in Figure 3 and has the following characteristics:

(a). The portion of the network cut by a fault in a linear region will be completely destroyed.

(b). The length 2r of the fault in the linear region is fixed, but the direction is random.

(c). Note that the length of the linear fault is 2r, the distance between its center point and the link is x, and the angle formed by the normal line of the link and the linear fault is , assuming that the direction of the fault in the line segment probability area is uniformly distributed, then The probability p that the link is interrupted is

p=2α/π=(2/π)arccos(x/r)

b). Analysis of network performance changes under a specific area failure model

In the process of analyzing network vulnerability, in addition to introducing geometric probability to analyze the impact of faults and equipment geographical location, grid subdivision technology is also used to locate the vulnerable areas that have the greatest impact on the network, which more realistically reflects the impact of disaster events on the network . The technical solution adopted is: first divide the plane where the network is located into a series of small squares. When the squares are small enough, it can be considered that any point in the square has the same impact on network performance.

The impact of regional faults on the network is calculated by meshing. For each given network performance index Δ, the system first calculates the impact Δn on the network performance index when regional faults occur in each grid, and then uses the knowledge of geometric probability to calculate the network performance index using the following formula,

Δ=∑ _n (area(n)/all area)Δ _n

Where area(n) represents the area of the nth grid, and all area represents the geographical area of the network deployment.

For the first indicator: Disconnected Link Capacity (DLC), the failure probability of links in the fault area is calculated first, and then the value of DLC is obtained by calculating the expected value of all possible disconnected link capacities. For the second metric: Reduced Traffic Between Point-Pairs (PTR), first compute the maximum number of edge-unduplicated paths between specified node pairs. The nodes and links on these paths constitute a subnetwork of the original network. When a probabilistic area failure occurs, the expected value of the reduced data flow between the original network node pair is the PTR caused by the link failure in the sub-network. Path protection is a commonly used strategy to improve network reliability at a relatively low cost. Its idea is to establish two links between a pair of nodes, one of which is the main path and the other is the backup path, which can be used The Suurballe algorithm finds primary and secondary links between pairs of nodes. By calculating the probability that the main path and the backup path fail at the same time when the area failure occurs, it is the probability of disconnection between point pairs (PDP). The Suurballe algorithm is a classical algorithm for finding a path set with K non-repetitive edges between specified two points and a minimum total length.

c). Locate the vulnerable fault area that has the greatest impact on the network

Assuming that the fault center appears in the center of each subdivided network, calculate the corresponding performance index to measure the vulnerability of the network. These standard quantities are sorted from large to small, and the corresponding set of small squares in the top (such as 0.1%) constitutes the most vulnerable area in the network.

When the step size of grid division is small and the scale of the network is large, the corresponding calculation amount will be very large. For this reason, the present invention adopts the distributed parallel computing technology, and speeds up the calculation process by calculating the influence on the network when regional faults occur in each grid subdivision in multiple computing units, and then calculating the average value.

2. Principle of Network Vulnerability Analysis System

FIG. 4 is a schematic diagram of network plane dissection. In the process of analyzing network vulnerability, in addition to introducing geometric probability to analyze the impact of faults and equipment geographical location, grid subdivision technology is also used to locate the vulnerable areas that have the greatest impact on the network, which more realistically reflects the impact of disaster events on the network . The technical solution adopted is: first divide the plane where the network is located into a series of small squares. When the squares are small enough, it can be considered that any point in the square has the same impact on network performance. Then calculate the impact on the network with each grid as the fault center and the given index Δ as the standard to measure the network performance. Assuming that the network with E edges is divided into N grids, the calculation process is as follows:

Step1: Take a certain grid as the regional fault center, assuming that the fault model and its parameters have been set by the user. Calculate the probability that each edge is affected by the fault. After all the probability calculations are completed, an N*E failure probability matrix is formed, which is set to failure_prob[][], and failure_prob[i][j] means that the i-th grid is Fault center, the probability that the jth edge is destroyed.

Step2: Taking each given network performance index Δ as the evaluation standard, calculate the impact on the network with the index grid as the fault center. The calculation methods of the three indicators to measure the vulnerability of the network are:

(1). Calculation method of DLC: first calculate the probability that the link in the fault area is interrupted, and then use the probability as the weight to calculate the weighted sum of the link capacity;

(2). Calculation method of PTR: use the Suurballe algorithm to find all non-repeated path sets between point pairs, and calculate the weighted sum of the link capacities of these path sets;

(3). Calculation method of PDP: use the Suurballe algorithm to find the two shortest non-repetitive paths between point pairs, and calculate the weighted sum of the link capacities of these two paths.

Step3: For all grids, sort them according to DLC, PTR, and PDP respectively from large to small. The top 0.1% of the grid is selected as the final result, which is the most vulnerable area in the network.

When the step size of grid division is small and the scale of the network is large, the corresponding calculation amount will be very large. However, except that the failure probability matrix failure_prob[][] calculated in the first step needs to be used in the calculation process of each grid, the calculations of the second and third steps are independent of each other, so the parallel calculation method is used to speed up the calculation speed. Step2 is a simple accumulation process, and Step3 is a sorting process. The two steps are combined into a one-step parallel calculation of Mapreduce. The input of the Mapreduce calculation process is the failure probability matrix failure_prob of size N*E, and the output is the sorted key-value pair< dlc,Point>,<ptr,Point>,<pdp,Point>, where Point represents the center of the grid.

3. Composition of network vulnerability analysis system

The system is divided into two parts, the calculation module and the GUI module. The GUI module is responsible for receiving user input (including model selection, parameter setting, etc.) and presenting the results in a visualized manner. The calculation module calculates the regional failure probability in parallel according to the model and parameters, and the calculation process is as described in 2.

Figure 5 is a schematic diagram of the network vulnerability analysis system, which contains the following elements:

(1) Select the drop-down box of the physical network;

(2) Select the radio button of the probabilistic area fault model (concentric circle model/line segment model);

(3) Text fields displaying network and probabilistic area fault parameters;

(4) Execute button (executes the analysis of the given parameters);

(5) Select the radio button of the evaluation standard (DLC/PTR/PDP);

(6) A progress bar showing the change of network performance under the condition of given fault center, under the evaluation condition and under the worst condition;

(7) Empty button (prepare for the next analysis);

(8) Display all nodes and edges in the network, and use thicker dots to indicate the source and destination;

(9) The found path sets with non-repeating edges are marked with different colors to distinguish them from other paths.

Finally, the shaded portion of the graph indicates the location of potential fault areas that have the greatest impact on a particular network metric.

The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, some improvements can also be made without departing from the principle of the present invention, and these improvements should also be regarded as the invention. protected range.

Claims (6)

1. a kind of method of work of the communication network vulnerability analysis system for large-scale area fault it is characterised in that：Its Comprise the steps：

1) simulate the physical network topology of real world first and build area fault model, in this fault model, faulty section The network facilities in domain can be destroyed with certain Probability p, and p is bigger, and the network representing in this region is more fragile, and destroyed Probability can change with the difference of the area in and place disaster region far and near apart from disaster center；

2) subdivision is carried out according to physical network topology to the two-dimentional geographic plane in fault model, form a series of grids, depending on every In individual grid, the impact to the network equipment for the regional faults is identical, with each grid as defect center, with given index Δ for weighing apparatus The standard of amount network performance, the impact Δ n, Ran Houtong to network performance index when zoning fault occurs in each grid Cross the knowledge of geometric probability, using equation below calculating network performance indications,

Δ=∑_n(area(n)/all area)Δ_n

Wherein area (n) represents the area of n-th grid, and all area represent the geographic area area of network design；

3) form the standard volume of three measurement network vulnerabilities：

(1) link capacity of disconnection, i.e. DLC：When area fault occurs, the link capacity meansigma methodss of disconnection；When link capacity is During unit value, represent the number of links of average disconnection；

(2) point between reduce flow, i.e. PTR：Area fault occur when it is intended that node between data traffic reduce Meansigma methodss；

(3) point between disconnection probability, i.e. PDP：It is intended that operating path and Protection path between node pair when area fault occurs There is the probability of disconnection simultaneously.

2. it is directed to the method for work of the communication network vulnerability analysis system of large-scale area fault as claimed in claim 1, It is characterized in that：The method also includes step 4) process to the maximum fault section of web influence for the positioning, specially：According to upper State the standard volume of the three measurement network vulnerabilities calculating；These standard volumes are sorted from big to small, comes correspondence above Lattice set constitute the region of most fragile in network.

3. it is directed to the work side of the communication network vulnerability analysis system of large-scale area fault as claimed in claim 1 or 2 Method it is characterised in that：Step 1) described in area fault model include simulate earthquake concentric circular probability region fault model, with And simulation hurricane, the line segment failure probability model of trawlnet infringement optical cable, wherein：The knot of described concentric circular probability region fault model Structure is：

(1) m r is followed successively by by M radius, the concentric circular of m=1 ..., M is divided into M annulus, and the annulus at center is also simultaneously Circle, each annulus is a uniform disk like probability region fault, and wherein r is pitch；

(2) in m-th disk like concentric ring fault zone, length is the probability that breaks down of link of arbitrarily short δ is q_mδ, its Middle q_mRepresent the probability of malfunction in m-th region；

(3) because destructiveness reduces, parameter q in annulus with the increase to source point distance_mBy inner monotone decreasing outward, I.e. q₁>q₂>…>q_m；

Article (4) one, link is longer through the length of concentric circular probability region fault, and the probability being damaged is also bigger, if through same The linkage length of heart circular probability area fault is l, and the Probability p that it breaks down is：

$p=1-e^{-{\mathrm{\&Sigma;}}_{m=1}^M{q}_m{l}_m}$

Especially, as q=0, p=0, representing does not have generation area fault；As q=+ ∞, p=1, it is degenerated to definitiveness mould Type；

The structure of described line segment failure probability model is：

(1) network portion cut by line-like area fault will be completely destroyed；

(2) length 2r of line-like area fault is fixing, but direction has randomness；

(3) length of note linearity failure is 2r, and the distance away from link for its central point is x, and link normal is formed with linearity failure Angle is α it is assumed that the direction of line segment probability region fault is equally distributed, then the Probability p that link is interrupted is

P=2 α/π=(2/ π) arccos (x/r).

4. it is directed to the work side of the communication network vulnerability analysis system of large-scale area fault as claimed in claim 1 or 2 Method is it is characterised in that step 2) detailed process be：

Assume that the network having E bar side is divided into N number of grid, its calculating process is as follows：

Step 1：Using some grid as area fault center it is assumed that fault model and its parameter are good by user setup, Calculate the probability that each edge is affected by this fault, after the completion of all probability calculations, define the probability of malfunction matrix of N*E, if For failure_prob [] [], with i-th grid as defect center, j-th strip side is destroyed for failure_prob [i] [j] expression Probability；

Step 2：With each given network performance index Δ as evaluation criteria, calculate with any one grid as defect center The impact that network is caused.

5. it is directed to the method for work of the communication network vulnerability analysis system of large-scale area fault as claimed in claim 2, It is characterized in that step 3) detailed process be：

(1) computational methods of DLC：Calculate the probability that the link in fault zone is interrupted first, then with this probability as weight, Calculate the weighted sum of link capacity；

(2) computational methods of PTR：Using Suurballe algorithm find point between all sides not repeatedly path set, calculate The weighted sum of the link capacity of these path sets；

(3) computational methods of PDP：Using Suurballe algorithm find point between two sides the shortest not repeatedly path, Calculate the weighted sum of the link capacity of this two paths；

K bar side between described Suurballe algorithm is look for specifying at 2 points is not repeated and overall length minimum path set Classic algorithm.

6. it is directed to the method for work of the communication network vulnerability analysis system of large-scale area fault as claimed in claim 5, It is characterized in that step 4) detailed process be：To all grids, sort from big to small by DLC, PTR, PDP respectively, before selecting 0.1% grid is as final result, i.e. most fragile region in network.