CN102025628B - Distribution method of network flow - Google Patents

Abstract

The embodiment of the invention relates to a distribution method of network flow. The distribution method comprises the following steps of: establishing a plurality of optional routing paths between a source node and a destination node of a network, and initially distributing the network flow between the source node and the destination node to one or more optional routing paths; according to the network flow initially distributed to the optional routing paths, determining the initial optimal routing path in the optional routing paths by a hopfield neural network algorithm, and distributing the network flow between the source node and the destination node to the initial optimal routing path; and according to the network flow initially distributed to the optional routing paths and the network flow distributed to the initial optimal routing path, adjusting the network flow distributed to the optional routing paths by an FD flow deviation algorithm until the network transmission time delay meets the predetermined requirements. In the invention, the multiple optional routings are established between the source node and the destination node, and the hopfield neural network algorithm and the FD algorithm are combined for adjusting the service load of each link in the network, thereby adjusting the flow distribution and optimizing the network transmission time delay.

Description

The distribution method of network traffics

Technical field

The present invention relates to communication technical field, relate in particular to a kind of distribution method of network traffics.

Background technology

In packet-switched communication networks network or computer network; Route Selection has extremely important influence for the performance of network; Increasing application such as real-time audio frequency, video traffic etc. all need network that strict QoS (Quality of Service can be provided; Service quality) guarantee, as ensureing network delay or cell loss ratio or peak transfer rate etc.

At present; When carrying out Route Selection according to assignment of traffic; Usually only consider the routed path of setting up of network source node and network destination node; And do not have the real-time variation according to network state to carry out the flow adjustment, therefore, the route of selecting for the user can not well satisfy the demand of user to QoS.

Summary of the invention

The embodiment of the invention provides a kind of distribution method of network traffics, optimizes network transfer delay.

The purpose of the embodiment of the invention realizes through following technical scheme:

Between the source node of network and destination node, set up many alternative routed paths, on the said alternative routed path of the network traffics original allocation to one or more between said source node and the said destination node;

Network traffics based on original allocation on the said alternative routed path; Confirm the initial optimum routed path in the said alternative routed path through the hopfield neural network algorithm, and the network traffics between said source node and the said destination node are assigned on the said initial optimum routed path;

According to the network traffics of distributing on the network traffics of original allocation on the said alternative routed path and the said initial optimum routed path, adjust the network traffics of distributing on the said alternative routed path through FD flow deviation algorithm;

When repeatedly adjusting the network traffics of distributing on the said alternative routed path through the FD algorithm; Confirm the optimum routed path in the adjusted said alternative routed path of current these network traffics through the hopfield neural network algorithm; And the network traffics between said source node and the said destination node are assigned on the said optimum routed path; To carry out network traffics adjustment next time, satisfy predetermined requirement until network transfer delay.

Technical scheme by the invention described above embodiment can be found out: through between source node and destination node, setting up many alternative routes; In conjunction with each link traffic load in hopfield neural network algorithm and the FD flow deviation algorithm adjustment network; Carry out the assignment of traffic adjustment, to reach the optimization network transfer delay.

Description of drawings

Fig. 1 provides the schematic flow sheet of the distribution method of network traffics for the embodiment of the invention;

Fig. 2 provides the topological sketch map of network in the distribution method of network traffics for the embodiment of the invention;

Fig. 3 provides many schematic flow sheets that alternative routed path is set up in the distribution method of network traffics for the embodiment of the invention.

Embodiment

As shown in Figure 1, a kind of distribution method of network traffics comprises:

Step 11, between the source node of network and destination node, set up many alternative routed paths, on the alternative routed path of network traffics original allocation to one or more between source node and the destination node.

Step 12, according to the network traffics of original allocation on the alternative routed path; Confirm the initial optimum routed path in the alternative routed path through the hopfield neural network algorithm, and the network traffics between source node and the destination node are assigned on the initial optimum routed path.

Step 13, according to the network traffics of distributing on the network traffics of original allocation on the alternative routed path and the initial optimum routed path, adjust the network traffics of distributing on the alternative routed path through FD (Flow Deviation, flow deviation) algorithm.

Step 14, when repeatedly adjusting the network traffics of distributing on the alternative routed path through the FD algorithm; Confirm the optimum routed path in the adjusted alternative routed path of current these network traffics through the hopfield neural network algorithm; And the network traffics between source node and the destination node are assigned on the optimum routed path; To carry out network traffics adjustment next time, satisfy predetermined requirement until network transfer delay.

Technical scheme by the invention described above embodiment can be found out: between source node and destination node, set up many alternative routes; In conjunction with each link traffic load in hopfield neural network algorithm and the flow deviation algorithm adjustment network; Carry out the assignment of traffic adjustment, to reach the optimization network transfer delay.And, realized the routing optimality under many alternative paths situation.

Optional, in the above-mentioned steps 11, when the original allocation network traffics, network traffics can be assigned on the one or more alternative routed path, and are unrestricted.

Step 12, step 13 are in conjunction with link business load in hopfield neural network algorithm and the FD algorithm adjustment network.

Need to prove,, realize that promptly network transfer delay satisfies predetermined requirement owing to can not guarantee once to adjust the network traffics of distributing on the alternative routed path.After each network traffics adjustment, the network traffics distribution changes, and the propagation delay time of each link (link between the adjacent node) also can change, and then need continue adjustment.Therefore, in the step 14, continue to confirm optimum routed path in the alternative routed path, continue network traffics on the alternative routed path of adjustment through the FD algorithm then, satisfy predetermined requirement up to network transfer delay through the hopfield neural network algorithm.

In the step 14, it can be that network transfer delay satisfies certain threshold values that network transfer delay satisfies predetermined requirement, or the like.

It is thus clear that initial optimum routed path and optimum routed path are represented the preferred routed path of different phase, do not give unnecessary details at this.

Concrete, between the destination node of the source node of network and network, set up many alternative routed paths in the above-mentioned steps 11, can comprise:

Source node sends a plurality of route exploration bags and confirms that with hop-by-hop source node arrives the intermediate node of destination node, carries intermediate node information and link information during route exploration bag process intermediate node.

Destination node is set up many first alternative routed paths according to the nodal information and the link information that carry in a plurality of route exploration bags that receive.

Destination node is sent a plurality of acknowledge messages through many first alternative routed paths, carries the intermediate node information of renewal and the link information of renewal during acknowledge message process intermediate node.

Source node is set up many alternative routed paths according to the intermediate node information of the renewal of carrying in a plurality of acknowledge messages that receive and the link information of renewal.

Optional, between the destination node of the source node of network and network, set up many alternative routed paths, can also comprise:

Destination node is set first timer after receiving first route exploration bag, with a plurality of route exploration bags of confirming that first timer is received before overtime, so that guarantee can be consuming time not long because of waiting for that all route exploration bags cause.

In like manner, source node receives that first acknowledge message bag sets second timer, with a plurality of acknowledge messages of confirming that second timer is received before overtime, so that guarantee can be consuming time not long because of waiting for that all acknowledge messages cause.

Exemplary, the topological sketch map of network as shown in Figure 2, wherein network comprises: source node 21, destination node 23, and intermediate node 221, intermediate node 222, intermediate node 223 only exemplaryly illustrate 3 intermediate nodes.

As shown in Figure 3, specify many alternative routed paths and set up process:

Step 31 when source node has data to send, at first can be sent same route exploration bag to a plurality of alternative intermediate nodes, comprises source node information and destination node information in the route exploration bag.Like Fig. 2, source node 21 respectively sends 1 route exploration bag to 3 intermediate nodes 22.For fear of network complexity, the number of route exploration bag can be in 5, but this quantity is not regarded as the restriction to route detection packet quantity.

Step 32; After intermediate node is received the route exploration bag; According to the network performance parameter that is kept at this node for example link channel capacity, load, time delay etc. calculate next-hop nodes, and be kept at the network informations such as selected link information, nodal information in the route exploration bag.Route exploration bag hop-by-hop is selected node, and collection network information, and when the route exploration bag arrived destination node, the route exploration bag had just intactly been preserved the information of whole piece route.

Step 33; Destination node can be set first timer after receiving first route exploration bag; The route that the route exploration bag of before first timer is overtime, receiving from same source node is comprised is as the first alternative routed path, and the first alternative routed path can be one or more.Exemplary, as shown in Figure 2, the first alternative routed path as, source node 21 is through 3 paths shown in intermediate node 221, intermediate node 222, intermediate node 223 and 23 arrows of destination node.The route exploration bag of receiving after overtime can directly abandon.

Step 34, destination node is sent ACK (Acknowledge) acknowledge message according to the first alternative routed path.The first alternative routing iinformation is included in the corresponding ACK message.

Step 35, ACK message is returned source node along the first alternative route, and through intermediate node the time, upgrades link and nodal information, upgrades link and nodal information and already is included in the corresponding ACK message.

Step 36, source node can be set second timer after receiving first ACK message, and the route that the ACK message of receiving before second timer is overtime is comprised is as alternative routed path, and alternative routed path can be one or more.Directly abandon ACK message after overtime.

So far, source node can be set up many alternative routed paths of source node to destination node according to routing iinformation in the ACK message.

Particularly, above-mentioned steps 12 is confirmed the initial optimum routed path in the alternative routed path according to the network traffics of original allocation on the alternative routed path through the hopfield neural network algorithm, can comprise:

According to the network traffics of original allocation on the alternative routed path, confirm in the network network traffics of link between each adjacent node.

Based on the network traffics of link between each adjacent node, and the channel capacity of link between each adjacent node, confirm link overhead between each adjacent node.

According to link overhead between each adjacent node, confirm the initial optimum routed path in the alternative routed path through the hopfield neural network algorithm.

Particularly, above-mentioned steps 13 is adjusted the network traffics of distributing on the alternative routed path according to initial optimum routed path through the FD algorithm, can comprise:

According to the network traffics of distributing on the network traffics of original allocation on the alternative routed path and the initial optimum routed path, adjust the network traffics of distributing on the alternative routed path through the FD algorithm.

According to the relation of the network traffics of network transfer delay and link assignment,, confirm the network traffics of distributing on the corresponding alternative routed path when network transfer delay satisfies predetermined requirement.

Step 14 is confirmed the optimum routed path in the adjusted alternative routed path of current these network traffics through the hopfield neural network algorithm; And the network traffics between source node and the destination node are assigned on the optimum routed path; To carry out network traffics adjustment next time, can be able to understand with reference to above-mentioned steps 12, step 13.

Technical scheme by the invention described above embodiment can be found out: combine with the FD algorithm through the hopfield neural network algorithm; Network traffics are carried out the iteration adjustment; Drop to minimum up to network transfer delay; Obtain the optimal flow assignment strategy, source node is according to this result distribution service load on each alternative route.

Through adopting the hopfield neural network algorithm; Reduce the complexity of FD algorithm iteration effectively, only needed iteration number of times seldom just can reach convergence, obtained the optimum route of network; And along with the increase of network size, iterations can both remain at very in the reasonable range.

Exemplary, below in conjunction with hopfield neural network algorithm and FD algorithm, be target to be optimized for network transfer delay, the distribution method of embodiment of the invention network traffics is described.

At first, number of nodes is N in the network;

Number of links is L in the network;

The channel capacity of link l is expressed as C _l

The flow load of link l is expressed as f _l

The output packet propagation delay time is expressed as T ^*

According to the network delay model, independently suppose like Kleinrock, obtain the relational expression of network transfer delay and link flow load:

$T^{*}=\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{D}_l\left(f_l\right)=\underset{l=1}{\overset{L}{\mathrm{\Σ}}}\frac{f_l}{C_l-f_l}---\left(1\right)$

Then, the alternative route quantity between source node and the destination node is Q;

Flow load on the alternative route q is expressed as f (q);

Network traffics between source node and the point of destination node are expressed as λ;

All alternative routing traffic load distribution is expressed as

initial traffic load allocation vector is represented as

Therefore, obtain the flow load of certain bar link l in the network $f_l=\underset{q=1}{\overset{Q}{\mathrm{\Σ}}}f\left(q\right)\mathrm{\ζ}(q,l)---\left(2\right)$

Wherein,

Then, there is matrix ρ in connect matrix V and link;

Wherein, x and i representation node, the link between lxi representation node x and the node i.

Set up energy function according to the hopfield neural network algorithm:

$E=\frac{{\mathrm{\μ}}_1}{2}\underset{x=1}{\overset{N}{\mathrm{\Σ}}}\underset{\underset{(x,i)\≠(d,s)}{\underset{i\≠x}{i=1}}}{\overset{N}{\mathrm{\Σ}}}{C}_{\mathrm{xi}}\·V_{\mathrm{xi}}+\frac{{\mathrm{\μ}}_2}{2}\underset{x=1}{\overset{N}{\mathrm{\Σ}}}\underset{\underset{(x,i)\≠(d,s)}{\underset{i\≠x}{i=1}}}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ρ}}_{\mathrm{xi}}\·V_{\mathrm{xi}}$

$+\frac{{\mathrm{\μ}}_3}{2}\underset{x=1}{\overset{N}{\mathrm{\Σ}}}{\{\underset{\underset{i\≠x}{i=1}}{\overset{N}{\mathrm{\Σ}}}{V}_{\mathrm{xi}}-\underset{\underset{i\≠x}{i=1}}{\overset{N}{\mathrm{\Σ}}}{V}_{\mathrm{ix}}\}}^2$

$+\frac{{\mathrm{\μ}}_4}{2}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\underset{\underset{x\≠i}{x=1}}{\overset{N}{\mathrm{\Σ}}}{V}_{\mathrm{xi}}\·(1-V_{\mathrm{xi}})+\frac{{\mathrm{\μ}}_5}{2}(1-V_{\mathrm{ds}})---\left(3\right)$

Wherein, As link overhead; D represents destination node, and s represents source node; μ ₁-μ ₅Can get empirical value.Because the dynamical equation of hopfield neural network algorithm is:

$\frac{d{U}_{\mathrm{xi}}}{\mathrm{dt}}=\underset{y=1}{\overset{N}{\mathrm{\Σ}}}\underset{\underset{y\≠j}{j=1}}{\overset{N}{\mathrm{\Σ}}}{T}_{\mathrm{xi},\mathrm{yi}}\·V_{\mathrm{yj}}-\frac{U_{\mathrm{xi}}}{\mathrm{\τ}}+I_{\mathrm{xi}}---\left(4\right)$

Wherein, y and j representation node.

Can obtain the transfer weight matrix T=[T of hopfield neural network algorithm according to above-mentioned dynamical equation (4) and above-mentioned energy function (3) _{Xi, yj}] and deviant I _Xi

$I_{\mathrm{xi}}=-\frac{{\mathrm{\μ}}_1}{2}{C}_{\mathrm{xi}}(1-{\mathrm{\δ}}_{\mathrm{xd}}{\mathrm{\δ}}_{\mathrm{is}})-\frac{{\mathrm{\μ}}_2}{2}{\mathrm{\ρ}}_{\mathrm{xi}}(1-{\mathrm{\δ}}_{\mathrm{xd}}{\mathrm{\δ}}_{\mathrm{is}})-\frac{{\mathrm{\μ}}_4}{2}+\frac{{\mathrm{\μ}}_5}{2}{\mathrm{\δ}}_{\mathrm{xd}}{\mathrm{\δ}}_{\mathrm{is}};$

T _xi，yj＝μ ₄δ _xyδ _ij-μ ₃δ _ij-μ ₃δ _ij+μ ₃δ _jx+μ ₃δ _iy；

Wherein, δ representes the Kronecker function:

${\mathrm{\δ}}_{\mathrm{ab}}=\left\{\begin{array}{cc}1& \mathrm{ifa}=b\\ 0& \mathrm{otherwise}\end{array}\right.;$

Therefore, through shifting weight matrix T=[T _{Xi, yj}] and deviant I _Xi, can set up the hopfield neural net, obtain initial optimum route, and the network traffics λ between source node and the point of destination node is assigned on the initial optimum route, obtain flow adjustment vector The flow that optimum route is distributed in this vector is the network traffics between source node and the destination node, and all the other alternative routes are dispense flow rate not.

Need to prove that concrete meaning of parameters and effect that above-mentioned hopfield neural network algorithm relates to specifically can be able to understand with reference to hopfield neural network algorithm related content in the prior art, do not give unnecessary details at this.

Afterwards; According to;

and

network traffics through distributing on each alternative route of FD algorithm iteration adjustment obtain flow allocation vector on adjusted each alternative route

If need carry out repeatedly the flow adjustment through the FD algorithm; Then after each FD algorithm adjustment network traffics; Continuation is confirmed optimum routed path in the alternative routed path through the hopfield neural network algorithm; Continue network traffics on the alternative routed path of adjustment through the FD algorithm then, satisfy predetermined requirement up to network transfer delay.

After the k time iteration adjustment; The assignment of traffic vector

of each alternative route obtains optimum route through the hopfield neural network algorithm; And the network traffics λ between source node and the point of destination node is assigned on the optimum route; Obtain

and can calculate

therefore according to

; Can guarantee that up to the assignment of traffic of each alternative route network transfer delay

meets the demands through loop iteration; Obtain the optimal flow assignment strategy, source node is according to the distribution service load on each alternative route of optimal flow assignment strategy.

Concrete, can be according to the network traffics of distributing on each alternative route of formula (5) iteration adjustment

${\stackrel{\→}{f}}^{(k+1)}={\stackrel{\→}{f}}^{\left(k\right)}+{\mathrm{\α}}_k({\stackrel{\→}{\tilde{f}}}^{\left(k\right)}-{\stackrel{\→}{f}}^{\left(k\right)}),{\mathrm{\α}}_k\∈\left[\mathrm{0,1}\right]---\left(5\right)$

α wherein _kBe the adjustment step-length, calculate by formula (6).

${\mathrm{\α}}_k=\min \{1,\frac{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}({\tilde{f}}_l^k-f_l^k)D_l^{\′\′}\left(f_l^k\right)}{\underset{l=1}{\overset{L}{\mathrm{\Σ}}}({\tilde{f}}_l^k-f_l^k)D_l^{\′}\left(f_l^k\right)}\}---\left(6\right)$

In the formula (6), $f_l^k=\underset{q=1}{\overset{Q}{\mathrm{\Σ}}}{f}^k\left(q\right)\mathrm{\ζ}(q,l),$ ${\tilde{f}}_l^k=\underset{q=1}{\overset{Q}{\mathrm{\Σ}}}{\tilde{f}}^k\left(q\right)\mathrm{\ζ}(q,l),$ And $D_l^{\′\′}\left(f_l^k\right)=\frac{2\·C_l}{{(C_l-f_l)}^3}.$

Need to prove; The FD algorithm is used to solve to requiring down with a given topological structure and given external flow; Seek and do not have constraint, non-linear minimizing mathematical tool; The FD algorithm relates to related content, specifically can be able to understand with reference to FD algorithm related content in the prior art, does not give unnecessary details at this.

Technical scheme by the invention described above embodiment can be found out: combine with the FD algorithm through the hopfield neural network algorithm; Network traffics are carried out the iteration adjustment; Drop to minimum up to network transfer delay; Obtain the optimal flow assignment strategy, source node is according to this result distribution service load on each alternative route.

Through adopting the hopfield neural network algorithm; Reduce the complexity of FD algorithm iteration effectively, only needed iteration number of times seldom just can reach convergence, obtained the optimum route of network; And along with the increase of network size, iterations can both remain at very in the reasonable range.

The above; Be merely the preferable embodiment of the present invention, but protection scope of the present invention is not limited thereto, any technical staff who is familiar with the present technique field is in the technical scope that the present invention discloses; The variation that can expect easily or replacement all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection range of claims.

Claims (4)

1. the distribution method of network traffics is characterized in that, comprising:

Between the source node of network and destination node, set up many alternative routed paths, on the said alternative routed path of the network traffics original allocation to one or more between said source node and the said destination node;

Network traffics based on original allocation on the said alternative routed path; Confirm the initial optimum routed path in the said alternative routed path through the hopfield neural network algorithm, and the network traffics between said source node and the said destination node are assigned on the said initial optimum routed path;

According to the network traffics of distributing on the network traffics of original allocation on the said alternative routed path and the said initial optimum routed path, adjust the network traffics of distributing on the said alternative routed path through FD flow deviation algorithm;

When repeatedly adjusting the network traffics of distributing on the said alternative routed path through the FD algorithm; Confirm the optimum routed path in the adjusted said alternative routed path of current these network traffics through the hopfield neural network algorithm; And the network traffics between said source node and the said destination node are assigned on the said optimum routed path; To carry out network traffics adjustment next time, satisfy predetermined requirement until network transfer delay;

Set up many alternative routed paths between said source node and the destination node at network, comprising:

Source node sends a plurality of route exploration bags and confirms that with hop-by-hop said source node arrives the intermediate node of said destination node, and said route exploration bag carries said intermediate node information and link information during through said intermediate node;

Said destination node is set up many first alternative routed paths according to the intermediate node information and the link information that carry in a plurality of said route exploration bag that receives;

Said destination node is sent a plurality of acknowledge messages through many said first alternative routed paths, and said acknowledge message is carried the intermediate node information of renewal and the link information of renewal during through said intermediate node;

Said source node is set up many alternative routed paths according to the intermediate node information of the renewal of carrying in a plurality of said acknowledge message that receives and the intermediate line link information of renewal.

2. method according to claim 1 is characterized in that, sets up many alternative routed paths between said source node and the destination node at network, also comprises:

Said destination node is set first timer after receiving first said route exploration bag, with a plurality of said route exploration bag of confirming that said first timer is received before overtime;

Said source node receives that first said acknowledge message bag sets second timer, with a plurality of said acknowledge message of confirming that said second timer is received before overtime.

3. method according to claim 1 is characterized in that, according to the network traffics of original allocation on the said alternative routed path, confirms the initial optimum routed path in the said alternative routed path through the hopfield neural network algorithm, comprising:

According to the network traffics of original allocation on the said alternative routed path, confirm in the network network traffics of link between each adjacent node;

According to the network traffics of link between said each adjacent node, and the channel capacity of link between said each adjacent node, confirm link overhead between said each adjacent node;

Based on link overhead between said each adjacent node, confirm the initial optimum routed path in the said alternative routed path through the hopfield neural network algorithm.

4. method according to claim 1 is characterized in that, according to said initial optimum routed path, adjusts the network traffics of distributing on the said alternative routed path through the FD algorithm, comprising:

According to the network traffics of distributing on the network traffics of original allocation on the said alternative routed path and the said initial optimum routed path, adjust the network traffics of distributing on the said alternative routed path through the FD algorithm;

According to the relation of the network traffics of network transfer delay and link assignment,, confirm the network traffics of distributing on the corresponding said alternative routed path when network transfer delay satisfies predetermined requirement.

Images