CN102325052A - Optical network multi-fault tolerant method - Google Patents

Abstract

The invention discloses an optical network multi-fault tolerant method which relates to the technical field of optical network survivability. Each optical network node i maintains an information prime matrix Ti and a local node model Mi. The method comprises the following steps of: S1. a positive ant in each node of an optical network selects a target node d according to flow distribution at a preset time interval, asynchronously moves to the target node d and records the physical connectivity of the passed path and a path node identifier; S2. after the positive ant arrives at the target node, a reverse ant is generated at the target node, and the positive ant transfers all memories to the reverse ant; and S3. the reverse ant returns along the passed path of the positive ant and adjusts the local node model in a network state and the information prime matrix of each accessed node on the way according to the final connectivity of the path. Through the invention, the damaged business can be rapidly turned to a normal network resource and the arrival business can be selected to a path with higher connectivity.

Description

Optical-fiber network multiple faults fault-tolerance approach

Technical field

The present invention relates to optical-fiber network survivability technical field, particularly a kind of optical-fiber network multiple faults fault-tolerance approach.

Background technology

Existent environment of people constantly worsens, and natural calamity takes place frequently, and adds the widespread characteristics of communication network, causes communication network to be easy to be damaged, and accident often takes place, and consequence is very serious.Human factor takes place the destruction of network especially again and again, and in China, backbone network hundreds of interruption of optical cables all occurs every year, and the loss that is caused can't be estimated especially.

Fault-tolerant exactly when path, node hardware fault or software error in optical-fiber network, having occurred for various reasons; System can return to the former state that has an accident with these corrupted or lost business and data automatically, a kind of technology that system can normally be moved continuously.Fault-tolerant implementation: protection and recovery all belong to fault-tolerant scope, and the main problem that exists is that protection needs reserved resource in advance, and recovery needs fault localization mechanism.

The number of the fault that protection strategy need occur according to the network maximum possible carries out the reservation of resource conservation according to the resource of network, sets up the protection passage, the utilance of maximization resource conservation, and the structure of protecting this moment is a polyhedral structure normally.After breaking down, recovery policy needs the operational failure location mechanism, confirms the number and the location of fault of fault, residual resource Network Based then, and the foundation that recovers passage realizes professional unbroken transmission.

In the network of reality; The number of defects of network is at random; Protection can only realize the prevention to fault to a certain extent, and network manager need be known the number of the fault that needs protection in advance, just can obtain the protection structure that an optimum under the maximization condition is shared in resource conservation.In optical-fiber network because fault propagation to downstream node; The multiple faults location belongs to the NP-Hard problem; It is good more that the common number that all is the hypothesis network breaks down lacks more; Under this supposed situation, try to achieve the set of the possible fault of network, the set of trying to achieve not necessarily is exactly the situation of real fault in the real network, has very big uncertainty.

Protection and reset mode all can not well be accomplished the reply of optical-fiber network multiple faults, consider afore-mentioned, exist to overcome needs not enough in the correlation technique.

Summary of the invention

The technical problem that (one) will solve

The technical problem that the present invention will solve is: how to avoid the fault-tolerant uncertainty of multiple faults in the optical-fiber network.

(2) technical scheme

For solving the problems of the technologies described above, the invention provides a kind of optical-fiber network multiple faults fault-tolerance approach, each optical network node i safeguards a pheromones matrix T _iWith the local node model M _i, said M _iBe used to estimate the path of forward ant structure, model M _i(μ _Id, δ ² _Id, W _Id) be adaptive, μ _IdFor ant successfully arrives the connective average of physical pathway of destination node d, δ ² _IdBe the connective sample variance of physical pathway, W _IdFor moving watch window, be used to write down the highest W as a result of ant traversal physical pathway connectedness _{Best_id}For each the destination node d in the network, Estimation of Mean μ _IdWith variance δ ² _IdThe connectedness of sign from node i to node d, the method comprising the steps of:

S1: every the forward ant of each node is selected destination node d according to flow distribution in the optical-fiber network, and moves to said destination node d asynchronously at a distance from preset time at interval, simultaneously the physical connection property and the path node identifier of record path;

S2: after arriving destination node, generate reverse ant in destination node, said forward direction ant passes to reverse ant with physical connection property, the path node identifier in said path;

S3: said reverse ant returns along the path of forward direction ant institute approach, and according to the physical connection property in said path, upgrades the local node model M of network state _iAnd the pheromones matrix T of each node of visiting _i

Wherein, the forward ant selects the mode of destination node to be according to flow distribution among the said step S1: f _SdBe the measurement functions of data flow s-＞d, be at the definition of probability of the node s forward ant that to create a destination node be d:

$P_{\mathrm{sd}}=\frac{f_{\mathrm{sd}}}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{f}_{\mathrm{si}}}$

Wherein, n is a node sum in the optical-fiber network, P _SdBig more, the probability that produces the forward ant that arrives destination node d is big more.

Wherein, among the said step S1, the forward ant selects the mode of next node to be in the destination node moving process:

On each node i; Destination node is that the forward ant of d is selected the next node j that will visit in the adjacent node of not visiting; If all adjacent nodes were all visited, then select at random therein, select the probability P of adjacent node j as next one traversal node _IjdBe pheromones τ _IjdWith heuristic value η _IjNormalization with:

$P_{\mathrm{ijd}}=\frac{{\mathrm{\&tau;}}_{\mathrm{ijd}}+\mathrm{\&alpha;}{\mathrm{\&eta;}}_{\mathrm{<figure-callout\; id="1"\; label="ij"\; filenames="HDA0000093161320000011.png"\; state="\{\{state\}\}">ij</figure-callout>}}}{1+\mathrm{\&alpha;}\left(\right|N_i|-1)}$

Wherein, | N _i| the adjacent segments that is i node is counted, and the value of α is used for weighing the importance of heuristic value with respect to the pheromones size of pheromones matrix T.

Wherein, among the said step S3, the connectedness value according in the reverse ant memory adopts following formula to upgrade the local node model M _i:

μ _id←μ _id+ζ(o _i→d-μ _id)

σ ² _id←σ ² _id+ζ((o _i→d) ²-σ ² _id)

Wherein, μ _IdBe the path connectivity average that ant successfully arrives destination node d, σ ² _IdBe sample variance, o _{I → d}Be that to observe the physical pathway of forward direction ant from node i to node j recently connective, weigh the proportion that factor ζ is used to influence the nearest number of samples of mean value.

Wherein, said pheromones satisfies condition:

$\underset{j\&Element;N_i}{\mathrm{\&Sigma;}}{\mathrm{\&tau;}}_{\mathrm{ijd}}=1,$ d∈[1，n] $\&ForAll;i$

Wherein, | N _i| the adjacent segments that is i node is counted.

Wherein, the pheromones matrix T that constitutes by pheromones _i=[τ _Ijd] the routing table isomorphism used with the forward ant.

(3) beneficial effect

The present invention is through being introduced into the AntNet method in the fault-tolerant field of multiple faults, and the structure node model M _iRealized the active of all path connectivities in the optical-fiber network is explored, make optical-fiber network can be automatically with impaired switching services in normal Internet resources, and the higher path of service selection connectedness of arrival got on; Connectedness is used for characterizing the feasibility of every light path; Connective high more link is safe more, and the possibility that produces fault is just more little, thereby can avoid the fault-tolerant uncertainty of multiple faults in the optical-fiber network.

Description of drawings,

Fig. 1 is the data structure sketch map that ant is used in the optical-fiber network multiple faults fault-tolerance approach of the present invention;

Fig. 2 is a kind of optical-fiber network multiple faults fault-tolerance approach flow chart of the embodiment of the invention.

Embodiment

Below in conjunction with accompanying drawing and embodiment, specific embodiments of the invention describes in further detail.Following examples are used to explain the present invention, but are not used for limiting scope of the present invention.

Optical-fiber network multiple faults fault-tolerance approach of the present invention realizes based on the AntNet algorithm, explores the connectedness in all paths on one's own initiative.For the tolerance of connectedness, can be divided into photosphere and electricity layer: can estimate according to the physical damnification of light signal at photosphere, adopt the error rate to estimate usually at the electricity layer.Two kinds of tolerance are applied in respectively in the various network type, in all optical network, use the physical damnification of light signal to estimate, and use the error rate to estimate at the network that opto-electronic conversion is arranged.Connective value is between (0～1).If one disconnected fine fault appears in link, this moment, signal was interrupted fully, and link is connected state not, and this moment, connective value was 0, and when having no the physical damnification and the error rate, connectedness is 1 at this moment.The connectedness of the link in the reality is between 0～1.As shown in Figure 1, each optical network node i safeguards a pheromones matrix T _iWith the local node model M _iThe pheromones matrix T _i=[τ _Ijd] with the routing table isomorphism that the forward ant is used, be convenient statistics and calculate that pheromones is turned to 1 by specification in the present embodiment, that is:

$\underset{j\&Element;N_i}{\mathrm{\&Sigma;}}{\mathrm{\&tau;}}_{\mathrm{ijd}}=1,$ d∈[1，n] $\&ForAll;i$

Wherein, | N _i| the adjacent segments that is i node is counted.Ant group algorithm is a kind of algorithm of positive feedback.Ant is at every turn through behind the paths; According to the overall connectivity of whole piece link, this link is estimated, if this link is connective good; Just, make later ant tend to use this more excellent path in the more pheromones of the link release ratio of process.In Fig. 1, each node has been preserved next that arrived certain destination node and has been jumped the value of the pheromones of link, the value of every corresponding pheromones of link just, and this value has characterized the tendentiousness of selecting this link and reaching this destination node.Every ant has arrived destination node behind next the jumping link through this destination node; Connectedness according to the whole piece link is upgraded this value with regard to needs; The plain size of updated information is connective good and bad according to the whole piece link, all upgrades with the information corresponding element for the link of every ant process.The value of pheromones is an input results in the continuous learning process, and it has held the current of the being seen whole network of local node and state in the past.

M _iBe used to estimate the path of forward ant structure, model M _i(μ _Id, δ ² _Id, W _Id) be adaptive, μ _IdBe the average that ant travels through the physical connection property that successfully arrives destination node d, δ ² _IdBe the sample variance of physical connection property, W _IdBe a mobile watch window, be used for writing down the highest W as a result of ant traversal physical connection property _{Best_id}For each the destination node d in the network, Estimation of Mean μ _IdWith variance δ ² _IdWhat show is the expectation connectedness from node i to node d.

As shown in Figure 2, this method comprises:

Step S201, every the forward ant of each node is selected destination node d according to flow distribution in the optical-fiber network, and moves to said destination node d asynchronously at a distance from preset time at interval, simultaneously the physical connection property and the path node identifier of record path.

The forward ant selects the mode of destination node to be according to flow distribution: f _SdBe the measurement functions of data flow s-＞d, the definition of probability the node s normal ant that to create a destination node be d is so:

$P_{\mathrm{sd}}=\frac{f_{\mathrm{sd}}}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{f}_{\mathrm{si}}}$

Wherein, n is a node sum in the optical-fiber network, P _SdBig more, the probability that produces the forward ant that arrives destination node d is big more.

After selecting destination node d, just move, in moving process, select next node in the following way, to set up the path of node i to node d to this destination node d.

On each node i; Destination node is that the forward ant of d is selected the next node j that will visit in the adjacent node of not visiting; If all adjacent nodes were all visited, then select at random therein, select the probability P of adjacent node j as next one traversal node _IjdBe pheromones τ _IjdWith heuristic value η _IjNormalization with:

$P_{\mathrm{ijd}}=\frac{{\mathrm{\&tau;}}_{\mathrm{ijd}}+\mathrm{\&alpha;}{\mathrm{\&eta;}}_{\mathrm{<figure-callout\; id="1"\; label="ij"\; filenames="HDA0000093161320000011.png"\; state="\{\{state\}\}">ij</figure-callout>}}}{1+\mathrm{\&alpha;}\left(\right|N_i|-1)}$

Usually select probability P _IjdMaximum corresponding nodes is a next node.Wherein, | N _i| the adjacent segments that is i node is counted.The value of α is used for weighing the importance of heuristic value with respect to the pheromones size of pheromones matrix T.

Because the shared identical physical pathway of the packet in forward ant and the optical-fiber network, so the network failure situation that they run into is identical, therefore, avoids the fault-tolerant uncertain problem of multiple faults in the optical-fiber network.

Step S202 after the arrival destination node, generates reverse ant in destination node, and the forward direction ant is with all memories, and promptly the physical connection property in path and path node identifier pass to reverse ant, delete the forward direction ant simultaneously.The forward direction ant that has not arrived destination node d possibly come into path or the node that breaks down, and it is dead in advance to be equivalent to the forward direction ant.

Step S203, reverse ant returns along the path of forward direction ant institute approach, and is the function of variable according to this path connectivity, the local node model of adjustment network state and the pheromones of the node that each was visited.Pheromones on each limit of network is all constantly being volatilized (for all ant group algorithms; Pheromones on every limit all has volatilization mechanism; The pheromones of every link all reduces according to the formula of a minimizing exactly; The mode of this minimizing just is called volatilization), only reach the forward direction ant of destination node and generated reverse ant just upgrades this node to all elements relevant with destination node d pheromones matrix T _iWith the local node model M _i

Pheromones is upgraded: the pheromones τ that sets every paths at first _IjdFor identical value then, during the k+1 time iteration, for the path that does not have the ant process: τ _Ijd(k+1)=(1-ρ (k)) τ _Ijd(k), obviously pheromones has reduced.The route that the ant process is arranged: τ _Ijd(k+1)=(1-ρ (k)) τ _Ijd(k)+ρ (k)/| W|.The parameter of ρ (k) for setting, ρ (k)=1-lnk/ln (k+1), W are the set of all nodes in the optical-fiber network, | W| is the node number.

In the return course, the connectedness in the reverse ant memory is worth upgrades M _i, the formula below adopting upgrades the local node model:

μ _id←μ _id+ζ(o _i→d-μ _id)

σ ² _id←σ ² _id+ζ((o _i→d) ²-σ ² _id)

Wherein, o _{I → d}Be to observe the connectedness of forward direction ant from node i to node j recently, factor ζ weighs the proportion of the nearest number of samples that can influence mean value.

Above execution mode only is used to explain the present invention; And be not limitation of the present invention; The those of ordinary skill in relevant technologies field under the situation that does not break away from the spirit and scope of the present invention, can also be made various variations and modification; Therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.

Claims (6)

1. an optical-fiber network multiple faults fault-tolerance approach is characterized in that, each optical network node i safeguards a pheromones matrix T _iWith the local node model M _i, said M _iBe used to estimate the path of forward ant structure, model M _i(μ _Id, δ ² _Id, W _Id) be adaptive, μ _IdFor ant successfully arrives the connective average of physical pathway of destination node d, δ ² _IdBe the connective sample variance of physical pathway, W _IdFor moving watch window, be used to write down the highest W as a result of ant traversal physical pathway connectedness _{Best_id}, for each the destination node d in the network, Estimation of Mean μ _IdWith variance δ ² _IdThe connectedness of sign from node i to node d, the method comprising the steps of:

S1: every the forward ant of each node is selected destination node d according to flow distribution in the optical-fiber network, and moves to said destination node d asynchronously at a distance from preset time at interval, simultaneously the physical connection property and the path node identifier of record path;

S2: after arriving destination node, generate reverse ant in destination node, said forward direction ant passes to reverse ant with physical connection property, the path node identifier in said path;

S3: said reverse ant returns along the path of forward direction ant institute approach, and according to the physical connection property in said path, upgrades the local node model M of network state _iAnd the pheromones matrix T of each node of visiting _i

2. optical-fiber network multiple faults fault-tolerance approach as claimed in claim 1 is characterized in that, the forward ant selects the mode of destination node to be according to flow distribution among the said step S1: f _SdBe the measurement functions of data flow s-＞d, be at the definition of probability of the node s forward ant that to create a destination node be d:

$P_{\mathrm{sd}}=\frac{f_{\mathrm{sd}}}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{f}_{\mathrm{si}}}$

Wherein, n is a node sum in the optical-fiber network, P _SdBig more, the probability that produces the forward ant that arrives destination node d is big more.

3. optical-fiber network multiple faults fault-tolerance approach as claimed in claim 1 is characterized in that, among the said step S1, the forward ant selects the mode of next node to be in the destination node moving process:

On each node i; Destination node is that the forward ant of d is selected the next node j that will visit in the adjacent node of not visiting; If all adjacent nodes were all visited, then select at random therein, select the probability P of adjacent node j as next one traversal node _IjdBe pheromones τ _IjdWith heuristic value η _IjNormalization with:

$P_{\mathrm{ijd}}=\frac{{\mathrm{\&tau;}}_{\mathrm{ijd}}+\mathrm{\&alpha;}{\mathrm{\&eta;}}_{\mathrm{ij}}}{1+\mathrm{\&alpha;}\left(\right|N_i|-1)}$

Wherein, | N _i| the adjacent segments that is i node is counted, and the value of α is used for weighing the importance of heuristic value with respect to the pheromones size of pheromones matrix T.

4. optical-fiber network multiple faults fault-tolerance approach as claimed in claim 1 is characterized in that, among the said step S3, the connectedness value according in the reverse ant memory adopts following formula to upgrade the local node model M _i:

μ _id←μ _id+ζ(o _i→d-μ _id)

σ ² _id←σ ² _id+ζ((o _i→d) ²-σ ² _id)

Wherein, μ _IdBe the path connectivity average that ant successfully arrives destination node d, σ ² _IdBe sample variance, o _{I → d}Be that to observe the physical pathway of forward direction ant from node i to node j recently connective, weigh the proportion that factor ζ is used to influence the nearest number of samples of mean value.

5. like each described optical-fiber network multiple faults fault-tolerance approach in the claim 1～4, it is characterized in that said pheromones satisfies condition:

$\underset{j\&Element;N_i}{\mathrm{\&Sigma;}}{\mathrm{\&tau;}}_{\mathrm{ijd}}=1,$ d∈[1，n] $\&ForAll;i$

Wherein, | N _i| the adjacent segments that is i node is counted.

6. optical-fiber network multiple faults fault-tolerance approach as claimed in claim 5 is characterized in that, the pheromones matrix T that is made up of pheromones _i=[τ _Ijd] the routing table isomorphism used with the forward ant.

Images