CN103281256A - Network tomography-based end-to-end path packet loss rate detection method - Google Patents

Abstract

The invention provides a network tomography-based end-to-end path packet loss rate detection method. The method comprises the following steps of: S1, modeling the relationship between path packet loss rate and link packet loss rate in a network path to obtain Gx=b; S2, screening by a routing matrix G to obtain an initial path set Ginde, wherein the path packet loss rate of the Ginde is detected during the screening process; S3, judging whether the Ginde is a maximum linearly independent path set, if not, continuing to execute S4, and if so, setting the Ginde as a set G1 and turning to S5; S4, continuing to screen the Ginde by utilizing a path selection algorithm; and S5, acquiring the path packet loss rate in the extreme set G1 and calculating remaining path packet loss rate. By the network tomography-based end-to-end path packet loss rate detection method disclosed by the invention, the path packet loss rate and the link packet loss rate are modeled by acquiring network topology, and a network environment supports a user datagram protocol (UDP), the load of a network is low, and the number of the obtained detection paths is reduced obviously.

Description

The end-to-end path packet loss detection method of chromatography Network Based

Technical field

The present invention relates to technical field of the computer network, particularly a kind of end-to-end path packet loss detection method of chromatography Network Based.

Background technology

Along with Internet fast development, people improve constantly the degree of dependence of network service, and the performance of network is also had higher requirement.Network performance measurement is basis and the checking means of carrying out network traffics, topology, behavior modeling analysis.How to use lower cost, the operation conditions that measures whole network exactly becomes the focus of more and more researcher's researchs.

Network tomography is to survey by the probe to end-to-end transmission, obtains structure and the performance information of network internal.The network tomography technology can be under the situation that does not have the internal network node cooperation, by initiatively sending probe and inferring large scale network link level qos parameter and topological structure according to statistical method.The packet loss Detection Techniques are the awareness network behaviors, carry out network control, eliminate network bottleneck, improve basis and the important step of network performance, so have very important meaning as one of them branch.

Transmit in the process of information at IP network, congested being difficult to avoided, and the congested network that may cause can't normally provide service, thereby causes economic loss and potential safety hazard.Network end-to-end path packet loss and network link packet loss can reflect operation conditions and the Congestion Level SPCC of network, are the important indicators of network performance, therefore how to obtain the focus that path and link packet drop rate become numerous researcher's researchs rapidly and accurately.The detection problem of primary study end-to-end path packet loss herein.

The path monitoring method of network can be divided into initiatively measurement and passive measurement according to the mode of measuring.The passive measurement technology is only obtained the network performance situation by the method that catches the available data bag that tested network transmits, and its performance index that can analyze out can be subjected to the influence of monitoring range.Initiatively measuring is present emerging network measure mode, and it is measured the transmission situation that wraps in the network by the particular data packet, the record that initiatively send and obtains network performance data, have higher flexibility, but this method need take certain Internet resources.In order to save the Internet resources expense, existing network monitoring method based on initiatively measurement has mainly adopted following measure: 1) optimize and dispose watch-dog, to reduce hardware spending; 2) select the to try one's best quantity of few measuring route; 3) reduce overlapping measurement, to reduce link load as far as possible.

Existing network measuring system is divided into two classes: a class is general measuring system, and a class is only to measure the system of time delay.The former can measure a lot of performances of network, but when n terminal, needs O (n ²) the measurement complexity.And latter's one side measurement result is accurate inadequately; On the other hand, even this latency measurement system can measure the result, still can not provide the guess value of accurate percent of pass and mortality ratio.

In the network with n terminal, the definition path is the connection of terminal room, and link is the connection between the point-to-point, and the path is exactly a series of link set.Just exist O (n at n terminal room so ²) path of number.And the existing technical problem that will solve is: wish path subclass of trying one's best little of monitoring, thereby can infer packet loss and the time delay that all paths of residue.

The method of available technology adopting has: adopt multicast detection or uni-cast probe, in order to solve the limited problem of multicast detection application environment, can adopt uni-cast probe; Perhaps place monitor in initial logic topological tree inside, and collect all observed data of initial logic tree, set up the packet loss sequence of the leaf node of every subtree correspondence then, thereby infer the packet loss that its inner every link; Perhaps use topology and the packet loss of every detective path of unicast packet detection network, and carry out modeling according to the packet loss in the topological sum path of described network, utilize Non-Linear Programming to calculate link packet drop rate according to the model of building.

Yet prior art has the shortcoming that detectivity is low, complexity is high and cost on network communication is big.

Summary of the invention

(1) technical problem of Xie Jueing

The technical problem that the present invention solves is that the more excellent path of few distribution of how selecting to try one's best is surveyed, and obtains the calculated value of network packet loss rate accurately.

(2) technical scheme

The invention provides a kind of end-to-end path packet loss detection method of chromatography Network Based, described method comprises:

S1: the relation to path packet loss and link packet drop rate in the network is carried out modeling, obtains Gx=b, and wherein, G is route matrix, and x is the logarithm of every link percent of pass, and b is the logarithm of every paths percent of pass;

S2: G screens to the route matrix, obtains initial path set G _Inde, in screening process, detect described initial path set G _IndeThe path packet loss;

S3: judge described initial path set G _IndeWhether be very big linear independence set of paths, if not, execution in step S4 then continued; If then establish described initial path set G _IndeBe set G ₁, and jump procedure S5;

S4: utilize routing algorithm that initial path is gathered G _IndeProceed screening;

S5: in step S2, survey the described initial path set G that obtains _IndeThe path packet loss in obtain described set G ₁In the path packet loss, and calculate the residual paths packet loss; Described residual paths is that described route matrix G filters out described set G ₁Remaining path, back.

Preferably, among the step S1 percent of pass be 1 with the difference of packet loss.

Preferably, among the step S2 described route matrix G screened specifically and comprises:

S21: described route matrix G is divided into groups according to the path node j that sets out, and the row of the path of each grouping in described route matrix G number is recorded in S set _jIn;

S22: extract S set _jIn first data m, the path that obtains the capable correspondence of matrix m is vectorial u, judge described vectorial u whether with G _IndeIn SYSTEM OF LINEAR VECTOR irrelevant, if linear independence, jump procedure S23 then is if linear correlation then deposits the number of going m in set T _j

S23: survey the path percent of pass of described vectorial u correspondence, and judge whether the path percent of pass of described vectorial u correspondence is 1, if 1, then with S set _jIn data m from S set _jMiddle deletion, and upgrade described route matrix G, and then upgrade G _IndeIf be not 1, then the path with described vectorial u correspondence joins G _IndeIn.

Preferably, judge the described vectorial u first time and G _IndeIn vector whether during linear independence, establish G _IndeBe null matrix.

Preferably, the described route matrix G of described renewal specifically comprises: be 1 column vector deletion with the value of the capable correspondence of m among the described route matrix G.

Preferably, the G that finally obtains among the determining step S23 _IndeOrder whether equal to upgrade the way of escape by the order of matrix G, if unequal, then will gather T _jAssignment is to S set _j, jump procedure S22; If equate, then continue step S3.

Preferably, routing algorithm is specially described in the step S4:

S41: get initial path set G _IndeFirst row be made as v, calculate R ₁=orth (G ₁ ^Τ) ^Τv ^ΤAnd R ₂=‖ v ‖ ²-‖ R ₁‖ ², wherein, T represents transposition, || || expression delivery, G ₁Be very big linear independence set of paths;

S42: judge R ₂Whether be 0, if be not 0, then v, continue step S43 if satisfying linear independence; If 0, jump procedure S44 then;

S43: v is joined G ₁In, and judge whether v is G ₁Last column, if, then with G ₁Return, otherwise execution in step S44;

S44: initial path is gathered G _IndeThe next line assignment give v, execution in step S41.

Preferably, G among the step S41 ₁Initial value be null matrix.

Preferably, calculating the residual paths packet loss among the step S5 specifically comprises:

S51: establishing the residual paths matrix is G _Rem, with G _RemWith described very big linear independence set of paths G ₁Recombinate, obtain the matrix of recombinating $G_{\mathrm{final}}=\left[\begin{array}{c}{G}_1\\ G_{\mathrm{rem}}\end{array}\right];$

S52: to reorganization matrix G _FinalCarry out elementary rank transformation, then residual paths percent of pass logarithm is reorganization matrix G _FinalCarry out matrix and described very big linear independence set of paths G that elementary rank transformation obtains ₁In path percent of pass logarithm long-pending.

Three) beneficial effect

The present invention proposes a kind of end-to-end path packet loss detection method of chromatography Network Based, by obtaining network topology path packet loss and link packet drop rate are carried out modeling, and support the network environment of UDP, less to the load that network causes, and the detective path quantity that obtains obviously reduces.

Description of drawings

Fig. 1 is method flow diagram provided by the invention;

Fig. 2 is network topology structure figure in the embodiment of the invention.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the invention, the technical scheme in the embodiment of the invention is clearly and completely described.

Embodiment 1:

The invention provides a kind of end-to-end path packet loss detection method of chromatography Network Based, this method flow diagram as shown in Figure 1, described method comprises:

S1: the relation to path packet loss and link packet drop rate in the network is carried out modeling, obtains Gx=b, and wherein, G is route matrix, and x is the logarithm of every link percent of pass, and b is the logarithm of every paths percent of pass;

By surveying the packet loss of topological and every the detective path that obtains network, set up algebraic equation Gx=b.G is route matrix, a line display one paths among the G, and a link, G are shown in a tabulation _IkBe the element of the capable k row of the i of matrix, value is 0 or 1, works as G _IkRepresented that path i did not comprise link k at=0 o'clock, work as G _Ik=1 expression path i comprises link k.

Among the step S1 percent of pass be 1 with the difference of packet loss.

S2: G screens to the route matrix, obtains initial path set G _Inde, in screening process, detect described initial path set G _IndeThe path packet loss;

Among the step S2 described route matrix G screened specifically and comprises:

S21: described route matrix G is divided into groups according to the path node j that sets out, and the row of the path of each grouping in described route matrix G number is recorded in S set _jIn;

S22: extract S set _jIn data m, the path that obtains the capable correspondence of matrix m is vectorial u, judge described vectorial u whether with G _IndeIn SYSTEM OF LINEAR VECTOR irrelevant, if linear independence, if jump procedure S23 then is linear correlation, then with the capable set T that deposits in of m _j

S23: survey the path percent of pass of described vectorial u correspondence, and judge whether the path percent of pass of described vectorial u correspondence is 1, if 1, then with S set _jIn m capable of S set _jMiddle deletion, and upgrade described route matrix G, and then upgrade G _IndeIf be not 1, then the path with described vectorial u correspondence joins G _IndeIn.

Judge the described vectorial u first time and G _IndeIn vector whether during linear independence, establish G _IndeBe null matrix.Form by a series of 0,1 among the route matrix G, then upgrade described route matrix G and specifically comprise: be 1 column vector deletion with the value of the capable correspondence of m among the described route matrix G.The G that finally obtains among the determining step S23 _IndeOrder whether equal to upgrade the way of escape by the order of matrix G, if unequal, then will gather T _jAssignment is to S set _j, jump procedure S22; If equate, then continue step S3.

S3: judge described initial path set G _IndeWhether be very big linear independence set of paths if not, then continue execution in step S4; If then establish described initial path set G _IndeBe set G ₁, and jump procedure S5;

S4: utilize routing algorithm that initial path is gathered G _IndeProceed screening;

Routing algorithm is specially following steps described in the step S4:

S41: get initial path set G _IndeFirst row be made as v, calculate R ₁=orth (G ₁ ^Τ) ^Τv ^ΤAnd R ₂=‖ v ‖ ²-‖ R ₁‖ ², wherein, T represents transposition, || || expression delivery, G ₁Be very big linear independence set of paths;

S42: judge R ₂Whether be 0, if be not 0, then v, continue step S33 if satisfying linear independence; If 0, jump procedure S44 then;

S43: v is joined G ₁In, and judge whether v is G ₁Last column, if, then with G ₁Return, otherwise execution in step S44;

S44: initial path is gathered G _IndeThe next line assignment give v, execution in step S41.

G among the step S41 ₁Initial value be null matrix.

S5: in step S2, survey the described initial path set G that obtains _IndeThe path packet loss in obtain described set G ₁In the path packet loss, and calculate the residual paths packet loss; Described residual paths is that described route matrix G filters out described set G ₁Remaining path, back.

Calculating the residual paths packet loss among the step S5 specifically comprises:

S51: establishing the residual paths matrix is G _Rem, with G _RemWith described very big linear independence set of paths G ₁Recombinate, obtain the matrix of recombinating $G_{\mathrm{final}}=\left[\begin{array}{c}{G}_1\\ G_{\mathrm{rem}}\end{array}\right];$

S52: to reorganization matrix G _FinalCarry out elementary rank transformation, then the residual paths packet loss is reorganization matrix G _FinalCarry out matrix and described very big linear independence set of paths G that elementary rank transformation obtains ₁In the path packet loss long-pending.

Embodiment 2:

Part topological structure with a practical IP network is that example is showed path of the present invention packet loss detection method below.Exist service end to end in this topological structure between any two main frames.e ₁-e ₆Represent and have a data harvester in the link network in the network, the data message that exists in real-time monitoring and the collection network.This topological structure as shown in Figure 2.

The end-to-end path packet loss detection method implementation step that the present invention proposes a kind of chromatography Network Based is as follows:

101) read network topology and detection information, concrete steps comprise:

Be that end of probe is carried out the traceroute(Internet mutually with 3 main frames in scheming earlier) obtain network topology, it is as follows to obtain route topological:

$G=\left[\begin{array}{cccccc}1& 0& 0& 0& 0& 1\\ 1& 0& 0& 0& 1& 0\\ 0& 0& 1& 1& 0& 0\\ 0& 0& 1& 0& 1& 0\\ 0& 1& 0& 1& 0& 0\\ 0& 1& 0& 0& 0& 1\end{array}\right]$

Each line display one paths is respectively A → B, A → C, B → A, B → C, C → A, C → B; A link is shown in each tabulation, respectively the e in the corresponding diagram ₁To e ₆And their opposite direction (from left to right).

102) route matrix G is handled, survey while screening, obtain initial path set G _Inde:

$G_{\mathrm{inde}}=\left[\begin{array}{ccc}0& 1& 1\\ 0& 1& 0\\ 1& 0& 1\end{array}\right]\stackrel{\‾}{b}=\ln \left[\begin{array}{c}0.9310\\ 0.9800\\ 0.9500\end{array}\right].$

103) judge G _IndeBe very big linear independence set of paths, then G ₁=G _Inde, obtain initial path set G _IndeThe logarithm of path percent of pass be

104) according to the packet loss of the very big linear independence set of paths reasoning residual paths that obtains, according to G ₁Obtain G _FinalAs follows:

$G_{\mathrm{final}}=\left[\begin{array}{ccc}0& 1& 1\\ 0& 1& 0\\ 1& 0& 1\\ 0& 0& 0\\ 0& 0& 0\\ 1& 0& 0\end{array}\right]$

To G _FinalCarry out elementary rank transformation and obtain its row simplest formula G _sFor:

$G_s=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\\ 0& 0& 0\\ 0& 0& 0\\ -1& 1& 1\end{array}\right]$

Then The value that obtains b is:

$b=\ln \left[\begin{array}{c}0.9310\\ 0.9800\\ 0.9500\\ 1.0000\\ 1.0000\\ 1.0000\end{array}\right]$

The value that obtains the path percent of pass after the rearrangement is 1.0000,1.0000,0.9310,0.9800,0.9500,1.0000, and is identical with actual value.

Because the order of route matrix G is 5, all are not when using our algorithm, and detective path bar number is 5, and use after our algorithm herein, by the route matrix deleted that the detective path bar number that obtains is reduced to 3, has saved communication overhead to a great extent.

Above execution mode only is used for explanation 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 make a variety of changes 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 (9)

1. the end-to-end path packet loss detection method of chromatography Network Based is characterized in that described method comprises:

S1: the relation to path packet loss and link packet drop rate in the network is carried out modeling, obtains Gx=b, and wherein, G is route matrix, and x is the logarithm of every link percent of pass, and b is the logarithm of every paths percent of pass;

S2: G screens to the route matrix, obtains initial path set G _Inde, in screening process, detect described initial path set G _IndeThe path packet loss;

S3: judge described initial path set G _IndeWhether be very big linear independence set of paths, if not, execution in step S4 then continued; If then establish described initial path set G _IndeBe set G ₁, and jump procedure S5;

S4: utilize routing algorithm that initial path is gathered G _IndeProceed screening;

S5: in step S2, survey the described initial path set G that obtains _IndeThe path packet loss in obtain described set G ₁In the path packet loss, and calculate the residual paths packet loss; Described residual paths is that described route matrix G filters out described set G ₁Remaining path, back.

2. method according to claim 1 is characterized in that, among the step S1 percent of pass be 1 with the difference of packet loss.

3. method according to claim 1 is characterized in that, among the step S2 described route matrix G is screened specifically to comprise:

S21: described route matrix G is divided into groups according to the path node j that sets out, and the row of the path of each grouping in described route matrix G number is recorded in S set _jIn;

S22: extract S set _jIn first data m, the path that obtains the capable correspondence of matrix m is vectorial u, judge described vectorial u whether with G _IndeIn SYSTEM OF LINEAR VECTOR irrelevant, if linear independence, jump procedure S23 then is if linear correlation then deposits the number of going m in set T _j

S23: survey the path percent of pass of described vectorial u correspondence, and judge whether the path percent of pass of described vectorial u correspondence is 1, if 1, then with S set _jIn data m from S set _jMiddle deletion, and upgrade described route matrix G, and then upgrade G _IndeIf be not 1, then the path with described vectorial u correspondence joins G _IndeIn.

4. method according to claim 3 is characterized in that, judges described vectorial u for the first time and G _IndeIn vector whether during linear independence, establish G _IndeBe null matrix.

5. method according to claim 3 is characterized in that, the described route matrix G of described renewal specifically comprises: be 1 column vector deletion with the value of the capable correspondence of m among the described route matrix G.

6. according to claim 1 or 3 each described methods, it is characterized in that the G that finally obtains among the determining step S23 _IndeOrder whether equal to upgrade the way of escape by the order of matrix G, if unequal, then will gather T _jAssignment is to S set _j, jump procedure S22; If equate, then continue step S3.

7. method according to claim 1 is characterized in that, routing algorithm is specially described in the step S4:

S41: get initial path set G _IndeFirst row be made as v, calculate R ₁=orth (G ₁ ^Τ) ^Τv ^TAnd R ₂=‖ v ‖ ²-‖ R ₁‖ ², wherein, T represents transposition, || || expression delivery, G ₁Be very big linear independence set of paths;

S42: judge R ₂Whether be 0, if be not 0, then v, continue step S43 if satisfying linear independence; If 0, jump procedure S44 then;

S43: v is joined G ₁In, and judge whether v is G ₁Last column, if, then with G ₁Return, otherwise execution in step S44;

S44: initial path is gathered G _IndeThe next line assignment give v, execution in step S41.

8. method according to claim 7 is characterized in that, G among the step S41 ₁Initial value be null matrix.

9. method according to claim 1 is characterized in that, calculates the residual paths packet loss among the step S5 and specifically comprises:

S51: establishing the residual paths matrix is G _Rem, with G _RemWith described very big linear independence set of paths G ₁Recombinate, obtain the matrix of recombinating $G_{\mathrm{final}}=\left[\begin{array}{c}{G}_1\\ G_{\mathrm{rem}}\end{array}\right];$

S52: to reorganization matrix G _FinalCarry out elementary rank transformation, then residual paths percent of pass logarithm is reorganization matrix G _FinalCarry out matrix and described very big linear independence set of paths G that elementary rank transformation obtains ₁In path percent of pass logarithm long-pending.

Images