CN106209621A - The link failure recovery method of qos constraint - Google Patents

Abstract

nullThe present invention relates to the link failure recovery method of a kind of qos constraint，Initially set up link failure many backup path restorations mathematical model of a kind of qos constraint，Optimization aim is to minimize the heavy-route flow interrupt amount sum that in network, all links cause because of fault，Farthest heavy-route flow under conditions of guaranteeing QoS，Finally use heuristic iterative algorithm LFR QoS that the link failure many backup path restorations mathematical model set up is solved，Splice through wall scroll backup path、After backup path selects 2 sub-algorithm steps，Obtain both considering service quality QoS constraint it is further contemplated that the link of link failure recovery rate a plurality of backup path optimum results，Achieve the most quickly repairing of impaired link，And overcome the complexity height that SelectBP link failure recovery algorithm exists、The problems such as amendment topology information.

Description

The link failure recovery method of qos constraint

Technical field

The present invention relates to a kind of IP network fast rerouting method, particularly the link failure recovery side of qos constraint Method.

Background technology

Document " Zheng Q, Cao G H, Porta T F, et al.Cross-layer approach for Minimizing routing disruption in IP network. " disclose a kind of IP network fast rerouting method, I.e. SelectBP link failure recovery method, is a kind of fault recovery method based on many backup paths.The method is with IP network For the purpose of minimizing heavy-route interruption, use a plurality of backup path to promote the recovery rate of heavy-route flow, and backed up by consideration The reliability in path promotes the success rate of fault recovery, when network failure, is directly transformed into many by affected flow Backup path is transmitted, it is achieved quickly forwarding without interruption of flow.But, SelectBP link failure recovery algorithm remains In problems with:

(1) QoS demand is not considered.The reliability of backup path is considered by SelectBP algorithm as main Object, is associated the link failure of physical layer at IP layer, provides the fault correlation relation of logical links, on structure backup road Those high reliability links are preferentially chosen, it is ensured that the success rate of fault recovery during footpath.But do not consider the clothes in business recovery Business quality requirement so that poor for business suitabilitys such as the Internet video of delay sensitive, voice calls.

(2) physical layer state information is needed.The fault correlation model that SelectBP algorithm is given is firstly the need of there being physical layer Running state information, Computational Physics link failure rate, then it is mapped to IP layer, obtains the association event between IP layer logical links Barrier.Often physical layer running state information is difficult to obtain, and the application deployment of method have difficulties.

(3) routing table memory space requirements is big.On SelectBP algorithm is judged by the reliability of bottom physical link The fault correlation situation of layer logical links, needs detailed information of lower layer, and will through numerous and diverse derivation of equation find cross-layer it Between fault correlation relation, the amendment to former topology table is uncontrollable, it is likely that can cause and cause route because of too much amendment Table amount of storage excessively increases, and causes network to be difficult to undertake.

Summary of the invention

Solve the technical problem that

In order to be difficult to ensure that the service quality problem of business when solving existing link failure recovery algorithm heavy-route, this Bright link failure recovery method (the Link Failure Recovery algorithm proposing a kind of qos constraint With QoS constrain, LFR-QoS algorithm).

Technical scheme

A kind of link failure many backup path restorations optimization method based on LFR-QoS algorithm, it is characterised in that step is such as Under:

Step 1: set up probabilistic correlation fault model based on many backup paths strategy

Letting R be sharing memory SRLG event sets, when any one occurrence r ∈ R occurs, fault rate is not The link set of 0 constitutes the probability sharing memory PSRLG of event r, such as formula (1):

r_PSRLG={ e_i,j∈E:p_r(i,j)≠0} (1)

Wherein, p_r(i, link e when being j) generation of SRLG event r_i,jThe probability broken down；

Use p_rExpression event r probability of happening, link e_u,vAnd e_i,jWhen there is fault correlation, the probability each broken down divides Do not use p_u,vAnd p_i,jRepresent, such as formula (2) and formula (3):

p_u,v=p_r·p_r(u,v) (2)

p_i,j=p_r·p_r(i,j) (3)

According to (2) formula and (3) formula, only work as p_rP when ≠ 0_u,vAnd p_i,jIt is not 0, link e_u,vAnd e_i,jThere is fault correlation Time the probability that each breaks down affected by event r simultaneously；

Step 2: provide optimization object function TD

$TD={\mathrm{\Σ}}_{e_{i,j}\∈E}\left({\mathrm{\Σ}}_{k=1}^N\right(1-{\Π}_{e_{u,v}\∈S\left(B_{i,j}^k\right)}\left(1-p_r\·p_r\left(u,v\right)\right))r_{i,j}^k+p_r\·p_r(i,j\left)\right(l_{i,j}-{\mathrm{\Σ}}_{k=1}^N{r}_{i,j}^k\left)\right)$

Wherein, any link e_i,j∈ E,Represent link e_i,jKth bar backup path,Represent and constitute's All link set, link e_u,vRepresentMiddle arbitrary element,Represent kth bar backup pathFor link e_i,jRetain Bandwidth, l_i,jIts flow load when representing fault-free, N is the total quantity of backup path；

Step 3: set up link failure recovery model

1) variable

Represent kth bar backup pathFor link e_i,jThe bandwidth retained；

If e_u,vIt is to constituteA certain bar link be then set to 1；It is otherwise 0；

2) object function

Min TD

3) constraints

1. conservation constraints is flowed

${\mathrm{\Σ}}_{\∀u:e_{v,u}\∈E}{f}_{e_{v,u}}^h\left(e_{i,j}\right)-{\mathrm{\Σ}}_{\∀u:e_{u,v}\∈E}{f}_{e_{u,v}}^{h-1}\left(e_{i,j}\right)=0,\∀e_{i,j}\∈E,\∀v\∈V\backslash \{i,j\}---\left(10\right)$

Constraint (10) is that nodes stream conservation limits, and represents link e_i,jDuring fault, by h-1 jumping figure backup path Enter arbitrary node v ∈ V all flow sums of i, j} equal to flow out this node all backup path streams through h jumping figure Amount sum, link e_i,jEnd node be unsatisfactory for this constraint, whereinRepresent link e_i,jDuring fault, from node i to u warp Cross all backup path heavy-route flow sums of h jumping figure；

2. capacity-constrained

${\mathrm{\Σ}}_{k=1}^N{r}_{i,j}^k\≤l_{i,j}---\left(11\right)$

${\mathrm{\Σ}}_{e_{i,j}\∈E}{f}_{e_{u,v}}\left(e_{i,j}\right)\≤c_{u,v}-l_{u,v}---\left(12\right)$

Formula (11) attach most importance to routing traffic constraint, represent link e_i,jAll backup path heavy-route maximum flow be its bear Carry；Formula (12) is link bandwidth capacity constraint, represents that the heavy-route flow of each of the links carrying is not to be exceeded its available bandwidth；Its InRepresent link e_i,jLink e is distributed to during fault_u,vFlow, its value with formula (13) represent；

$f_{e_{u,v}}\left(e_{i,j}\right)={\mathrm{\Σ}}_{k=1}^N{x}_{u,v}^{i,j,k}{r}_{i,j}^k---\left(13\right)$

3. variable bound

1≤k≤N,0≤h≤H (15)

$r_{i,j}^k\≥0---\left(16\right)$

$x_{u,v}^{i,j,k}\∈\{0,1\}---\left(17\right)$

Formula (15) represents that each of the links is up to N bar backup path, and the time delay of every backup path is H to the maximum and jumps；Formula (16) represent that backup path reserves bandwidth non-negative；Meet Integer constrained characteristic conditional (17)；

Step 4: use the link failure recovery model that heuristic iterative algorithm LFR-QoS Algorithm for Solving step 3 is set up: LFR-QoS algorithm is selected 2 sub-Algorithm constitutions, the wherein splicing of wall scroll backup path by the splicing of wall scroll backup path and backup path Subalgorithm specifically includes following steps:

4a: use k critical path method (CPM) to calculate link e_i,jThe shortest backup path set omega (BP_i,j), andThere is available bandwidth and meet delay constraint；

4b: calculate Ω (BP_i,jThe heavy-route flow Δ TD of the shortest backup path BP of every in)_i,j；

4c: by Δ TD_i,jThe shortest backup path BP of value maximum is as link e_i,jKth bar backup path；

Step 5: use backup path to select subalgorithm to be each link construction most N bar backup path, specifically include Following steps:

5a: calculate each link e_i,jPriority LP_i,j=p_i,jl_i,j；

5b: according to LP_i,jValue descending order is to e_i,jSequence；

5c: for each link e_i,j, from the beginning of Article 1 backup path, call wall scroll backup path splicing subalgorithm Structure kth bar backup pathUntil k=N orConstruct unsuccessfully.

Beneficial effect

The link failure recovery method of a kind of qos constraint that the present invention proposes, is converted into fault recovery problem Optimization problem, eventually through the network optimum solving global extremum acquisition link failure recovery.The link event of qos constraint Relation between barrier restoration methods sweetly disposition QoS requirement and heavy-route flow, time delay is long, available bandwidth avoiding Link-recovery rate is taken into account again on the premise of deficiency.Comparing SelectBP link failure recovery algorithm, the present invention is extensive at link failure Ensure that QoS time multiple, computation complexity is low, and is not related to bottom physical link information.

Accompanying drawing explanation

Fig. 1 is the stratification LFR-QoS technical system schematic diagram that the present invention is set up

Fig. 2 is the wall scroll backup path available bandwidth deficiency schematic diagram given by the present invention

Fig. 3 is that 2 backup paths given by the present invention share heavy-route flow schematic diagram

Link failure recovery rate comparing result figure under quick business when Fig. 4 is low in the inventive method

Link failure recovery rate comparing result figure under quick business when Fig. 5 is high in the inventive method

Fig. 6 is heavy-route flow QoS Service Efficiency comparing result figure in the inventive method

Fig. 7 is the inventive method link Overflow RateHT comparing result figure

Detailed description of the invention

First this method gives backup path strategy, as the foundation of backup path structure in follow-up optimized algorithm, its A kind of probabilistic correlation fault model of secondary proposition, it is considered to the relatedness of link failure, defines probability sharing memory (Probabilistically Shared Risk Link Group, PSRLG) concept, establishes a kind of service quality based on this Link failure many backup path restorations mathematical model of constraint, optimization aim causes because of fault for minimizing in network all links Heavy-route flow interrupt amount sum, farthest heavy-route flow under conditions of guaranteeing QoS, finally adopt With heuristic iterative algorithm LFR-QoS, the link failure many backup path restorations mathematical model set up is solved, Jing Guodan The splicing of bar backup path, backup path select after 2 sub-algorithm steps, obtained both considering service quality QoS constraint it is further contemplated that The link a plurality of backup path optimum results of link failure recovery rate, it is achieved that the most quickly repairing of impaired link, and overcome The problems such as the complexity that SelectBP link failure recovery algorithm exists is high, amendment topology information.

Specifically include following step:

Step 1: set up probabilistic correlation fault model based on many backup paths strategy.

Current link failure recovery algorithm, mostly for independent failure, i.e. thinks the most only between link failure Vertical, mutually incoherent.And the most sometimes link failure is not completely self-contained, there is association each other, such as work as at the end During layer optical fiber link fault, a plurality of logical links being carried by may lose efficacy simultaneously.When a logical links lost efficacy, with Its other logical links being carried on same optical fiber link is likely to break down with a certain probability, i.e. between link failure There is probabilistic correlation.Sharing memory (Shared Risk Link Group, SRLG) model is used for representing that one group is shared The link set of same risk, when a link failure in SRLG, in this group, other link occurs losing efficacy with the probability of 1. But actually relevant fault not 100% definitely associates, have in the both links of relevant fault link and break down Time, another link simply breaks down with a certain probability, and relevant fault probability of happening is likely less than 1.To this end, the present invention is given Probability sharing memory concept, be add in traditional SRLG model fault correlation probability represent have certain general The fault model of rate association.

Definition 1: probability sharing memory (Probabilistically Shared Risk Link Group, PSRLG), letting R be SRLG event sets, when any one occurrence r ∈ R occurs, fault rate is not the link set composition of 0 The PSRLG of event r, such as formula (1).

r_PSRLG={ e_i,j∈E:p_r(i,j)≠0} (1)

Wherein, p_r(i, link e when being j) generation of SRLG event r_i,jThe probability broken down.

Link e_i,jWith link e_u,vBetween there is fault correlation and refer to when certain event r occurs, p_r(i, j) ≠ 0 and p_r(u, v)≠0.And in traditional SRLG model, link e_i,jWith link e_u,vBetween there is fault correlation and refer to when certain event r occurs, p_r(i, j)=1 and p_r(u, v)=1, traditional SRLG model is a special case of probabilistic correlation model.Utilize probabilistic correlation event Barrier model can portray the feature of relevant fault more truly.

Use p_rExpression event r probability of happening, link e_u,vAnd e_i,jWhen there is fault correlation, the probability each broken down divides Do not use p_u,vAnd p_i,jRepresent, such as formula (2) and formula (3)

p_u,v=p_r·p_r(u,v) (2)

p_i,j=p_r·p_r(i,j) (3)

According to (2) formula and (3) formula, only work as p_rP when ≠ 0_u,vAnd p_i,jIt is not 0, link e_u,vAnd e_i,jThere is fault correlation Time the probability that each breaks down affected by event r simultaneously.Probabilistic correlation fault model is utilized to calculate easily There is the probability that the link of relevant fault breaks down.

Step 2: provide optimization object function TD.

For any link e_i,j∈ E, present invention c_i,jRepresent its bandwidth capacity, use l_i,jIts flow when representing fault-free Load.So link e_i,jAvailable bandwidth be c_i,j-l_i,j, it is link e_i,jThe maximum heavy-route flow that can carry.Utilize Represent link e_i,jKth bar backup path, useRepresent backup pathFor link e_i,jThe bandwidth retained, standby by most N bars The heavy-route flow of part trail protection amounts toAssume that all N bar backup paths are the most available, heavy-route flow interrupt amount Available formula (4) represents.Heavy-route flow interrupt amount refers to that during link failure, former bearer traffic weighs road with through all backup paths By the difference of total flow, this partial discharge stops when fault to forward.

${\mathrm{TD}}_{i,j}=l_{i,j}-{\mathrm{\Σ}}_{k=1}^N{r}_{i,j}^k---\left(4\right)$

Formula (4) is assumed N bar backup path is the most available, according to above analyzing, exists probabilistic correlation fault between link, that Link e_i,jDuring faultIt is likely to break down simultaneously, ifBreaking down, on it, the heavy-route flow of carrying stops Forward.Utilize p_i,jRepresent link e_i,jProbability of malfunction, and set link e_i,jDuring faultProbability of malfunction beThen Consider link e during backup path reliability_i,jHeavy-route flow interrupt amount under fault can use formula (5) to represent.

${\mathrm{TD}}_{i,j,p}=p_{i,j}\left({\mathrm{\Σ}}_{k=1}^{N}P\right(B_{i,j}^k|e_{i,j})r_{i,j}^k+l_{i,j}-{\mathrm{\Σ}}_{k=1}^N{r}_{i,j}^k)---\left(5\right)$

For conditional probability, only as link e_i,jFault and its kth bar backup pathFault exist During association, its value is not 0.

Solve belowValue.WithRepresent and constituteAll link set, thenBreak down Probability can use formula (6) to represent

$P\left(B_{i,j}^k\right)=1-{\Π}_{e_{u,v}\∈S\left(B_{i,j}^k\right)}(1-p_{u,v})---\left(6\right)$

Its link e_u,vRepresentMiddle arbitrary element, p_u,vIt is e_u,vThe probability broken down.

Join probability relevant fault model,(7) formula can be further represented as

$P\left(B_{i,j}^k\right|e_{i,j})=(1-{\Π}_{e_{u,v}\∈S\left(B_{i,j}^k\right)}(1-p_r\·p_r\left(u,v\right)\left)\right)/p_r\·p_r(i,j)---\left(7\right)$

So, in network, all links can use formula (8) to represent because of the heavy-route flow interrupt amount sum that fault causes

$TD={\mathrm{\Σ}}_{e_{i,j}\∈E}\left({\mathrm{\Σ}}_{k=1}^N\right(1-{\Π}_{e_{u,v}\∈S\left(B_{i,j}^k\right)}\left(1-p_r\·p_r\left(u,v\right)\right))r_{i,j}^k+p_r\·p_r(i,j\left)\right(l_{i,j}-{\mathrm{\Σ}}_{k=1}^N{r}_{i,j}^k\left)\right)---\left(8\right)$

The design object of LFR-QoS algorithm of the present invention is exactly under QoS of survice constraint of demand so that formula (8) value is minimum, formula (8) optimization aim of the present invention it is.

Step 3: provide the mathematical model of link failure recovery.Specifically include stream conservation constraints, capacity-constrained and variable about Bundles etc., establish frame foundation for subsequent algorithm.

The present invention, to minimize heavy-route flow interrupt as target, with QoS of survice demand for constraint, utilizes many backup paths Technology, carries out MILP (Mixed Integer Linear Program, MILP) and builds fault recovery problem Mould.

1) variable

Represent kth bar backup pathFor link e_i,jThe bandwidth retained.

If e_u,vIt is to constituteA certain bar link be then set to 1；It is otherwise 0.

2) object function

Min TD (9)

The object function of the present invention is to minimize the heavy-route flow interrupt amount in network-wide basis.It comprises two parts, and one Being above all backup path heavy-route part ability, two is interrupt unit because of backup path faults itself.

3) constraints

1. conservation constraints is flowed

${\mathrm{\Σ}}_{\∀u:e_{v,u}\∈E}{f}_{e_{v,u}}^h\left(e_{i,j}\right)-{\mathrm{\Σ}}_{\∀u:e_{u,v}\∈E}{f}_{e_{u,v}}^{h-1}\left(e_{i,j}\right)=0,\∀e_{i,j}\∈E,\∀v\∈V\backslash \{i,j\}---\left(10\right)$

Constraint (10) is that nodes stream conservation limits.Represent link e_i,jDuring fault, by h-1 jumping figure backup path Enter arbitrary node v ∈ V all flow sums of i, j} equal to flow out this node all backup path streams through h jumping figure Amount sum, it needs to be noted link e_i,jEnd node be unsatisfactory for this constraint.WhereinRepresent link e_i,jFault Time, through all backup path heavy-route flow sums of h jumping figure from node i to u.

2. capacity-constrained

${\mathrm{\Σ}}_{k=1}^N{r}_{i,j}^k\≤l_{i,j}---\left(11\right)$

${\mathrm{\Σ}}_{e_{i,j}\∈E}{f}_{e_{u,v}}\left(e_{i,j}\right)\≤c_{u,v}-l_{u,v}---\left(12\right)$

Formula (11) attach most importance to routing traffic constraint, formula (12) be link bandwidth capacity constraint.Heavy-route flow constraint representation chain Road e_i,jAll backup path heavy-route maximum flow be its load.The carrying of link bandwidth capacity constraint representation each of the links Heavy-route flow is not to be exceeded its available bandwidth.WhereinRepresent link e_i,jLink e is distributed to during fault_u,vFlow, Its value can use formula (13) to represent.

$f_{e_{u,v}}\left(e_{i,j}\right)={\mathrm{\Σ}}_{k=1}^N{x}_{u,v}^{i,j,k}{r}_{i,j}^k---\left(13\right)$

It is made up of, such as formula all backup path heavy-route flow sums passing through different jumping figures from node i to u (14) shown in.

$f_{e_{u,v}}\left(e_{i,j}\right)={\mathrm{\Σ}}_{h=0}^H{f}_{e_{u,v}}^h\left(e_{i,j}\right)---\left(14\right)$

3. variable bound

1≤k≤N,0≤h≤H (15)

$r_{i,j}^k\≥0---\left(16\right)$

$x_{u,v}^{i,j,k}\∈\{0,1\}---\left(17\right)$

Formula (15) represents that each of the links is up to N bar backup path, and the time delay of every backup path is H to the maximum and jumps.Formula (16) represent that backup path reserves bandwidth non-negative.Meeting Integer constrained characteristic conditional (17), therefore this model belongs to non-determined Property polynomial time difficulty problem (NP-hard).

Step 4: use the link failure recovery model that LFR-QoS Algorithm for Solving step 2 is set up.LFR-QoS algorithm is by list The splicing of bar backup path and backup path select 2 sub-Algorithm constitutions, and wherein wall scroll backup path splicing subalgorithm specifically includes Following steps:

(1) k critical path method (CPM) is used to calculate link e_i,jThe shortest backup path set omega (BP_i,j), andThere is available bandwidth and meet delay constraint；

(2) Ω (BP is calculated_i,jThe heavy-route flow Δ TD of the shortest backup path BP of every in)_i,j；

(3) by Δ TD_i,jThe shortest backup path BP of value maximum is as link e_i,jKth bar backup path.

Step 5: use backup path to select subalgorithm to be each link construction most N bar backup path, specifically include Following steps:

(1) each link e is calculated_i,jPriority LP_i,j；

(2) according to LP_i,jValue descending order is to e_i,jSequence；

(3) for each link e_i,j, from the beginning of Article 1 backup path, call wall scroll backup path splicing subalgorithm Structure kth bar backup pathUntil k=N orConstruct unsuccessfully.

In conjunction with embodiment, accompanying drawing, the invention will be further described:

1, stratification LFR-QoS technical system is set up

LFR-QoS technical system has 4 layers of intension, as shown in Figure 1.

The bottom is LFR-QoS technology platform, illustrates the ultimate principle of this technology, implementation framework, and technology platform determines The advantage that this technology is simple to operation.

The second layer is many backup paths generating algorithm, for obtaining the link backup path allocation scheme determined, generates and calculates Method must be based on technology platform on, but there is again certain motility, backup path can be optimized from different perspectives raw Become result, such as can be for the purpose of maximizing heavy-route flow, it is also possible to protect high capacity or high probability of malfunction link to be Purpose.

Third layer is the ways of distribution of converting flow, be primarily referred to as by fault affected flow switch to after backup path be with What kind of mode is forwarded to destination, such as, single path can be selected as required to forward or multipath Parallel transmutation.

4th layer is the application model of LFR-QoS, is primarily referred to as in the agreement that this technology is applied to give, how with original Agreement carries out collocation and coordinates.Such as, when network failure be characterized as temporary transient multiple time, in order to avoid routing convergence frequently, permissible Extend the action time of LFR-QoS route stand-by；When fault is nonvolatile, performance can be switched in time in order to ensure flow Preferably original route, then, after normal routing convergence completes, be switched to the flow in backup path in initial topology immediately again.

2, backup path strategy

Fig. 2 and Fig. 3 is the necessity schematic diagram utilizing a plurality of backup path protection link that the present invention is given.Link capacity It is 1, the flow load under digitized representation normal operational condition on link.

V in Fig. 2₁Utilize wall scroll backup path { v₁→v₂→v₄Protection link e_1,4, this backup path available bandwidth is min { 1-0.6,1-0.5}=0.4.Work as e_1,4During fault, link e_1,2The total traffic load of carrying will exceed own bandwidth, and link is just There will be overload phenomenon.

For the problem that backup path bandwidth is not enough, resolving ideas of the present invention is to utilize a plurality of backup path.With Fig. 3 it is Example, utilizes backup path { v simultaneously₁→v₂→v₄And { v₁→v₃→v₄Undertake faulty link e_1,4Flow, wherein path1 road The load of footpath carrying 0.4, the load of path2 path carrying 0.2.Based on this many backup paths method, faulty link e_1,4Flow Can be recovered completely without causing link Overload.

Use more backup path can reduce the interruption of flow, but significantly increase if backup path quantity crosses conference Add configuration complexity and storage overhead, and node between the backup path quantity that can arrange have strict restriction, set Each of the links at most has N bar backup path.The queue time delay of congested network increases with jumping figure exponentially rule, and in optical-fiber network Signal quality decline rapidly along with the increase of path jumping figure, therefore for support QoS demand, to backup path apply jumping figure limit System is necessary, and limiting backup path maximum hop count is H.Network topology utilizes the non-directed graph G (V, E) that has the right to represent, its Middle V and E represents router node and link set respectively.

3, LFR-QoS algorithm design

The MILP model of link failure recovery problem is NP-hard problem, although can utilize the traditional wire such as simplex method Property planing method solve, but along with the increase of problem scale, calculate time complexity higher, be not particularly suited for large scale network therefore Solving of barrier recovery problem.Therefore, the many backup path restorations of fault that the present invention designs under QoS is retrained by LFR-QoS algorithm are asked Topic solves.This algorithm by wall scroll backup path splicing and backup path select 2 sub-Algorithm constitutions, its solve target be By selecting most N bar backup paths for each of the links in network under QoS constraint, and reasonable distribution resource makes the whole network heavy-route Flow interrupt amount is minimum.

1) wall scroll backup path splicing

The backup path of link selects to carry out one by one, and the selection of wall scroll backup path is similar to calculate shortest path K shortest path first, from the beginning of node i, expands backup path by adding link one by one.Assume link e_i,jExisting k-1 bar Backup path, the extra flow interrupt amount adding a backup path minimizing can use formula (18) to represent.

${\mathrm{\ΔTD}}_{i,j}=p_{i,j}{r}_{i,j}^k(1-P(B_{i,j}^k|e_{i,j}\left)\right)---\left(18\right)$

LFR-QoS algorithm is when carrying out the splicing of wall scroll backup path, it is therefore an objective to will construct one and make Δ TD_i,jMaximumWith link e_i,jKth bar backup path stitching algorithm as a example by provide the wall scroll backup path stitching algorithm shown in table 1. The splicing of wall scroll backup path uses the k critical path method (CPM) improved, be allowed to ensure each backup path exist available bandwidth and Meet service delay constraint, choose the shortest backup path set omega (BP_i,jΔ TD in)_i,jThe shortest maximum backup path is as chain Road e_i,jKth bar backup path.

Wall scroll backup path splicing subalgorithm (the link e of table 1 LFR-QoS_i,jKth bar backup path stitching algorithm)

2) backup path selects

It is more owing to carrying link heavy-route flow when fault of more flow load, it is likely that because network does not has Abundant bandwidth resources and cause the interruption of heavy-route flow, and high probability of malfunction link is more easy to break down, therefore this Bright utilize the flow load of link and amassing of probability of malfunction to determine its protected priority, define link e_i,jPriority such as Formula (19).

LP_i,j=p_i,jl_i,j (19)

The LFR-QoS algorithm algorithm in the backup path choice phase is that one is didactic takes turns iterative algorithm, such as table 2 institute more Show.This algorithm is first according to bearer traffic load and probability of malfunction and calculates priority LP of link, and big according to LP value Little it is followed successively by each link construction most N bar backup path, while building each backup path, updates the residue of link Heavy-route flow load and network bandwidth resources.

The backup path of table 2 LFR-QoS selects subalgorithm

The present invention compares failure recovering algorithm SelectBP, and advantage is, one is by determining preferentially for each of the links Level so that high priority link obtains and more protects resource, reduces the flow load that network-wide basis endogenous cause of ill fault causes Interrupt；Two be each take turns iterative construction wall scroll backup path time, not only consider available bandwidth resources, it is also contemplated that backup path Time delay, under QoS of survice retrains, it is ensured that each newly-increased backup path the most farthest reduces in heavy-route flow Disconnected.Finally, LFR-QoS algorithm had both met QoS of survice constraint, the most farthest ensure that the weight of faulty link flow load Route.

3) LFR-QoS correctness proof

Theorem 1. is under given network topology constraint, it is assumed that k shortest path first can generate k bar shortest path, then The wall scroll backup path stitching algorithm of LFR-QoS can generate and meet available bandwidth and the backup path set omega of service delay constraint (BP_i,j).It should be noted that Ω (BP_i,j) can be empty set, represent link e_i,jThere is not more available backup path, The flow part of carrying stops to forward.

Prove. investigate wall scroll backup path stitching algorithm knowable to arthmetic statement and the extension of k shortest path first is mainly existed Two aspects: 1) in k bar shortest path building process, increase jumping figure H restriction, change the judgement bar that k bar shortest path generates Part；2) after k bar shortest path generates, increase available bandwidth constraint, retain and meet available bandwidth backup path to Ω (BP_i,j).K shortest path first can comprise the k shortest path tree T of k shortest path by structure_k, set T_kRoot node be I, leaf node is k the backup of terminal note j.To this end, wall scroll backup path stitching algorithm is based on k shortest path tree T_kRealize.

Mathematical induction is used to prove below.

As k=1, spanning tree T_kOnly comprise 1 shortest path.If H time-out, obtains in this shortest path building process Ω(BP_i,j) it is empty；If H time delay meets but there is not available bandwidth, it is similarly obtained Ω (BP_i,j) it is empty；If 2 conditions are equal Meet, obtain Ω (BP_i,j) comprise an element.Proposition is set up.

Assume that proposition is set up as k=n. wall scroll backup path stitching algorithm can generate when meeting available bandwidth and business Prolong the backup path set omega (BP of constraint_i,j)。

As k=n+1, the situation that (n+1)th backup path generates is discussed, i.e. utilizes the p tried to achieve₁, p₂..., p_nAsk Take p_n+1。

We utilize deflection path concept to build T_k.Assume p=(m₁,m₂,...,m_r) and q=(n₁,n₂,...,n_s) respectively For two paths of i to j, meet the following x retrained if had,

(1) x < r, and x < s；

(2)m_t=n_t(0≤t≤x)；

(3)m_x+1≠n_x+1；

(4)(n_x+1,n_x+2,...,n_s) it is n_x+1Shortest path to j.

Then claim (n_x,n_x+1) it is the q deviation limit to p, (n_x+1,n_x+2,...,n_s) it is the q the shortest deflection path to p.Standard In solving the shortest path algorithm, at known front n bar solving the shortest path p_n+1Process as follows: first travel through p_nIn except node j All node m_t, to obtain deviateing node, obtain m_tShortest path to node j；Then by this path and p_nOn from i to m_t Path splice, obtain p_n+1Path candidate；Finally concentrate from path candidate and choose the shortest path p the most_n+1。

Wall scroll backup path stitching algorithm compares p based on the shortest deflection path_n+1Generating algorithm, main change occurs Traversal m_tFinding the process to node j shortest path, this process extends to bandwidth constraint when adding.Each node m_tTo j Short path can use dijkstra's algorithm to solve, the node m joining delay and obtaining after bandwidth constraint_tShortest path to j Set is the subset of non-addition of constraints, i.e. reduces deviation range of nodes, the m obtained_tShortest path to j is probably time shortest path Footpath；Owing to during k=n, proposition is set up, p_nOn from i to m_tPath extend to the shortest path of bandwidth constraint when being to meet, by i to m_t With m_tShortest path splicing to j i.e. can get p_n+1Path candidate；Finally choose the maximum path of available bandwidth as p_n+1。

Then as k=n+1, proposition is set up.

Prove that understanding theorem 1 must demonstrate,prove by above.

LFR-QoS backup path selects subalgorithm to be completed by many wheel interative computations.First each of the links priority is arranged Sequence, then according to priority is followed successively by link chooses backup path, and distributes heavy-route flow, until not available band in network Till wide or link does not has more backup path.Backup path selection algorithm mainly nested wall scroll backup path stitching algorithm Realize, be substantially interative computation process.Therefore, in the case of wall scroll backup path stitching algorithm can correctly perform, standby Part routing algorithm can correctly perform.It addition, LFR-QoS algorithm realizes based on k shortest path first, equally ensure The path produced is loop-free.Card is finished.

4) LFR-QoS time complexity is analyzed

The wall scroll backup path splicing of LFR-QoS algorithm is similar to k shortest path first, therefore has same meter Calculate complexity O (| E |+| V |) log (V).Due to | E | bar link total in network, each of the links is up to N bar backup path, then The computation complexity of LFR-QoS algorithm is O (| E |+| V |) log (V) N | E |.

5) LFR-QoS space complexity is analyzed

4 one-dimensional vector storage algorithms of LFR-QoS algorithm employing run the data produced: 1 vector stores each of the links The shortest backup path set omega (BP_i,j), path can be described by a series of nodes, the memory space of individual paths not over V, then this vector storage space is up to k | V | | E |；1 vector splices result as storage of linked list wall scroll backup path, required Memory space is up to N | V | | E |；1 vector storage link prioritization result, memory space is | E |；1 vector storage The backup path structure result of link, memory space is N | V | | E |.Therefore, the space complexity of LFR-QoS algorithm is O (N | V | |E|)。

4, Performance Evaluation and analysis

The present invention with NS2 as emulation platform, to LFR-QoS algorithm and link failure recovery algorithm R3, FR-TE, SelectBP emulates, and transports from link failure recovery ability, heavy-route flow QoS Service Efficiency, link Overload rate and algorithm The row time, 4 aspects verified the performance of LFR-QoS algorithm of the present invention.

Contrast arthmetic statement is as follows.SelectBP algorithm considers the available bandwidth constraint of backup path, does not consider to back up road Footpath time delay, to minimize heavy-route flow interrupt amount as target；R3 algorithm with the network under multiple failure without congested as target； FR-TE algorithm turns to target with load balancing optimum in the network-wide basis of failover procedure.

1) experimental situation is arranged

Experiment topology uses Tier-1 backbone network, has 50 nodes and 180 limits, makes bandwidth and its weight of link Being inversely proportional to, specifically, link bandwidth is set to (5/ weight) Mbps.Emulation generates CBR data stream, and random from network Selection source destination node is transmitted, and the data package size in data stream is 1Kbytes, and data stream transmission speed is 200Kbps, Traffic queue length is set to 50 data package sizes, bound threshold value T₁And T₂It is respectively set to 10% He of capacity of queue 100%.By change generate data stream quantity change link utilization, number of data streams with 20 for step-length from 20 to 160 Change, corresponding average link utilization changes between 5% to 40%, and simulation parameter is arranged such as table 3.Calculate link the shortest standby K value in the k shortest path first in part path takes 5.

Table 3 simulation parameter is arranged

Link failure scene is done arranged below.9 sharing memory (SRLG) events of configuration altogether, each event comprises altogether Enjoy risk link 2-5 bar, make SRLG event break down the probability (p in (2) formula_r) scope is [0.05%, 0.5%], respectively Under SRLG event fault, in its group, the conditional probability scope of link is [0.3,1].10% is taken as high in independent failure link Probability faulty link, its probability of malfunction scope is [0.1%, 0.5%], the probability of malfunction scope of remaining link be [0.01%, 0.1%].In simulation process, configure 50 groups of probability of malfunction scenes altogether, for often organize probability of malfunction scene different random several 50 fault recoveries emulation is carried out under son (seed).Emulation randomly chooses 1 link every time as faulty link, if this link It is contained in certain probability sharing memory, then selects concurrent relevant fault link according to conditional probability, and treating excess syndrome tests knot The meansigma methods of fruit is as final simulation result.

2) performance evaluation

The present invention is from fault recovery rate, heavy-route flow QoS Service Efficiency, link Overload rate and 4 sides of Riming time of algorithm In the face of algorithm carries out Performance comparision, comparing result is respectively such as Fig. 4 to Fig. 7 and table 4

Fig. 4 with Fig. 5 be respectively 2 kinds different time quick business under link failure recovery rate comparing result, fault recovery rate has been The full faulty link proportion recovered.As seen from the figure, LFR-QoS algorithm is along with the increase of backup path number N, fault recovery rate In being gradually increasing trend, and when high, during quick business, improve about 17.5%, therefore N value can arrange larger, but standby Part number of path too much can increase the workload of router so that algorithm practicality is poor, and network is for the backup that can arrange Number of paths has strict restriction, considers, and the N value of LFR-QoS is set to 4.When N is 4, other fault recovery that compares is calculated Method, time low, during quick business, LFR-QoS algorithm fault recovery rate at most improves 35.1%, LFR-QoS algorithm during quick business time high Fault recovery rate at most improves 48.0%.Still as a example by backup path takes 4, quick business demand when comparing low, quick business time high During demand, LFR-QoS algorithm fault recovery rate have dropped about 2.8%, and under other failure recovering algorithm fault recovery rate is most Drop about 19.9%, shown that during business, the fault recovery rate on LFR-QoS algorithm that strengthens of quick demand affects less.Owing to lacking Time delay limits, and SelectBP, FR-TE and R3 algorithm is when high under quick business, and fault recovery effect is poor, by business chronesthesy shadow Ringing big, FR-TE and R3 algorithm does not accounts for the integrity problem of backup path, and fault recovery rate is relatively low.

During different business, the simulation result of the heavy-route flow QoS Service Efficiency of the lower 4 kinds of algorithms of quick demand is as shown in Figure 6, from It can be seen that the heavy-route flow QoS Service Efficiency of LFR-QoS algorithm maintains more than 98%, all the time the most not by business in figure Time quick demand impact, and SelectBP, FR-TE and R3 algorithm is along with the enhancing of demand quick during business, and heavy-route flow QoS is full Foot rate declines rapidly, and when FR-TE algorithm compares 275ms when business demand 50ms, heavy-route flow QoS Service Efficiency slippage reaches 54.8%.

Fig. 7 be load low time quick business time link Overload rate simulation result, as can be seen from Figure, LFR-QoS and Within SelectBP algorithm link Overload rate remains at 5%, unrelated with link utilization；FR-TE and R3 algorithm is along with chain The rising of road utilization rate, there is overload on a large scale, when link utilization reaches 40%, the link Overload of 2 kinds of algorithms in link Rate, more than 45%, greatly reduces fault recovery performance.Main cause is LFR-QoS and SelectBP algorithm only with have can Recover fault, and the strict heavy-route flow controlling backup path with the link of bandwidth, farthest avoid chain and pass by The generation of load situation.

Table 4 Riming time of algorithm

Table 4 is the average operating time that the present invention and other 3 kinds of methods solve link failure recovery problem.From table permissible Finding out, compared to LFR-QoS and SelectBP algorithm, FR-TE and R3 algorithm time overhead is relatively big, greatly affects link failure Time needed for configuration protection resource during recovery.LFR-QoS and SelectBP algorithm have employed takes turns heuritic approach more, significantly contracts The short problem solving time, and FR-TE and R3 algorithm uses traditional linear programming for solution method to solve the pact in fault recovery Bundle optimization problem, the time that solves is along with the growth exponentially rule increase of network size.

Claims (1)

1. link failure many backup path restorations optimization method based on LFR-QoS algorithm, it is characterised in that step is as follows:

Step 1: set up probabilistic correlation fault model based on many backup paths strategy

Letting R be sharing memory SRLG event sets, when any one occurrence r ∈ R occurs, fault rate is not 0 Link set constitutes the probability sharing memory PSRLG of event r, such as formula (1):

r_PSRLG={ e_i,j∈E:p_r(i,j)≠0} (1)

Wherein, p_r(i, link e when being j) generation of SRLG event r_i,jThe probability broken down；

Use p_rExpression event r probability of happening, link e_u,vAnd e_i,jWhen there is fault correlation, the probability each broken down is used respectively p_u,vAnd p_i,jRepresent, such as formula (2) and formula (3):

p_u,v=p_r·p_r(u,v) (2)

p_i,j=p_r·p_r(i,j) (3)

According to (2) formula and (3) formula, only work as p_rP when ≠ 0_u,vAnd p_i,jIt is not 0, link e_u,vAnd e_i,jWhen there is fault correlation each The probability broken down is affected by event r simultaneously；

Step 2: provide optimization object function TD

$TD={\mathrm{\&Sigma;}}_{e_{i,j}\&Element;E}\left({\mathrm{\&Sigma;}}_{k=1}^N\right(1-{\mathrm{\&Pi;}}_{e_{u,v}\&Element;S\left(B_{i,j}^k\right)}\left(1-p_r\&CenterDot;p_r\left(u,v\right)\right))r_{i,j}^k+p_r\&CenterDot;p_r(i,j)\left(l_{i,j}-{\mathrm{\&Sigma;}}_{k=1}^N{r}_{i,j}^k\right))$

Wherein, any link e_i,j∈ E,Represent link e_i,jKth bar backup path,Represent and constituteAll Link set, link e_u,vRepresentMiddle arbitrary element,Represent kth bar backup pathFor link e_i,jThe band retained Width, l_i,jIts flow load when representing fault-free, N is the total quantity of backup path；

Step 3: set up link failure recovery model

1) variable

Represent kth bar backup pathFor link e_i,jThe bandwidth retained；

If e_u,vIt is to constituteA certain bar link be then set to 1；It is otherwise 0；

2) object function

Min TD

3) constraints

1. conservation constraints is flowed

$\begin{array}{cc}{\mathrm{\&Sigma;}}_{\&ForAll;u:e_{v,u}\&Element;E}{f}_{e_{v,u}}^h\left(e_{i,j}\right)-{\mathrm{\&Sigma;}}_{\&ForAll;u:e_{u,v}\&Element;E}{f}_{e_{u,v}}^{h-1}\left(e_{i,j}\right)=0& \&ForAll;e_{i,j}\&Element;E,\&ForAll;v\&Element;V\backslash \left\{i,j\right\}\end{array}---\left(10\right)$

Constraint (10) is that nodes stream conservation limits, and represents link e_i,jDuring fault, entered by h-1 jumping figure backup path Arbitrary node v ∈ V all flow sums of i, j} equal to flow out this node all backup path flows through h jumping figure it With, link e_i,jEnd node be unsatisfactory for this constraint, whereinRepresent link e_i,jDuring fault, jump through h from node i to u All backup path heavy-route flow sums of number；

2. capacity-constrained

${\mathrm{\&Sigma;}}_{k=1}^N{r}_{i,j}^k\&le;l_{i,j}---\left(11\right)$

${\mathrm{\&Sigma;}}_{e_{i,j}\&Element;E}{f}_{e_{u,v}}\left(e_{i,j}\right)\&le;c_{u,v}-l_{u,v}---\left(12\right)$

Formula (11) attach most importance to routing traffic constraint, represent link e_i,jAll backup path heavy-route maximum flow be its load；Formula (12) it is link bandwidth capacity constraint, represents that the heavy-route flow of each of the links carrying is not to be exceeded its available bandwidth；WhereinRepresent link e_i,jLink e is distributed to during fault_u,vFlow, its value with formula (13) represent；

$f_{e_{u,v}}\left(e_{i,j}\right)={\mathrm{\&Sigma;}}_{k=1}^N{x}_{u,v}^{i,j,k}{r}_{i,j}^k---\left(13\right)$

3. variable bound

1≤k≤N,0≤h≤H (15)

$r_{i,j}^k\&GreaterEqual;0---\left(16\right)$

$x_{u,v}^{i,j,k}\&Element;\{0,1\}---\left(17\right)$

Formula (15) represents that each of the links is up to N bar backup path, and the time delay of every backup path is H to the maximum and jumps；Formula (16) table Show that backup path reserves bandwidth non-negative；Meet Integer constrained characteristic conditional (17)；

Step 4: use the link failure recovery model that heuristic iterative algorithm LFR-QoS Algorithm for Solving step 3 is set up: LFR- QoS algorithm is selected 2 sub-Algorithm constitutions by the splicing of wall scroll backup path and backup path, and wherein wall scroll backup path splicing is calculated Method specifically includes following steps:

4a: use k critical path method (CPM) to calculate link e_i,jThe shortest backup path set omega (BP_i,j), andDeposit In available bandwidth and meet delay constraint；

4b: calculate Ω (BP_i,jThe heavy-route flow Δ TD of the shortest backup path BP of every in)_i,j；

4c: by Δ TD_i,jThe shortest backup path BP of value maximum is as link e_i,jKth bar backup path；

Step 5: use backup path to select subalgorithm to be each link construction most N bar backup path, specifically include following Step:

5a: calculate each link e_i,jPriority LP_i,j=p_i,jl_i,j；

5b: according to LP_i,jValue descending order is to e_i,jSequence；

5c: for each link e_i,j, from the beginning of Article 1 backup path, call wall scroll backup path splicing subalgorithm structure Kth bar backup pathUntil k=N orConstruct unsuccessfully.