CN103581037B - A kind of level and smooth suppressing method of load based on power communication soft switch gateway - Google Patents

Abstract

A kind of level and smooth suppressing method of load based on power communication soft switch gateway of the present invention, belong to load balancing and the resource optimization technical field of powerline network, by considering fairness and the validity of load suppressing method, with soft switch gateway traffic overload for rejection condition, detect cpu busy percentage and the service spatial cache utilance of soft switch gateway, determine whether soft switch gateway is in traffic overload, and set up Model for Multi-Objective Optimization, utilize the optimization of ant group swarm intelligence to obtain the optimal feasible solution of this model, realize the level and smooth suppression to different business; The optimal feasible solution that the present invention proposes to be obtained by ant colony intelligence optimization multiple target efficiency Optimized model innovatively may carry out the effectively suppression of justice by overloaded traffic to different; The present invention proposes to adopt the level and smooth mode suppressed to prevent soft switch gateway load from occurring catastrophe, thus effectively can improve systematic function.

Description

A kind of level and smooth suppressing method of load based on power communication soft switch gateway

Technical field

The invention belongs to load balancing and the resource optimization technical field of powerline network, be specifically related to a kind of level and smooth suppressing method of load based on power communication soft switch gateway.

Background technology

Along with the development of power system telecommunications technology, powerline network of future generation has developed into business-driven type, opening, distributivity and the many data transmission networks of comprehensive multi-service.Powerline network of future generation is separated with Call-Control1, calls out and be separated with carrying due to business, its business can really independent of bearer network, and powerline network of future generation itself also can provide multi-service most certificate flexibly effectively, thus third party service provider is joined in the middle of the definition of new business, design and operation.As the support platform for business development, operation and management that powerline network of future generation provides, the network that oneself runs by soft switch gateway opens to third party service provider, and new business can develop independent of network-external and run.Soft switch gateway, as the important component part of powerline network architecture of future generation, while the call request of different entities processes in network, also will provide service processing function to the application demand of third party service provider.So a large amount of Business Processing requests is easy to cause soft switch gateway to occur that high capacity even occurs traffic overload, thus reduce the traffic handing capacity of soft switch gateway.Therefore, effective control of soft switch gateway load has become the focus theme of research both at home and abroad.

Existing a large amount of researcher controls to expand research to the load of soft switch gateway now, the people such as Zhang propose a kind of adaptive control algorithm based on soft switch gateway server, service request is processed by utilizing priority treatment principle the earliest, and utilize token bucket algorithm service control arrival rate, judge whether overload by the queue wait time of up-to-date business, thus make this algorithm simple possible.The people such as Liu have studied the fairness that load controls, and propose and effectively can solve the method that soft exchanging network closes premature beats fairness.The people such as Deng, using soft exchanging network Central Shanxi Plain cpu busy percentage as load overload threshold, propose the overload control algorithms based on bank money controlling mechanism, by considering the fairness of premature beats, have studied the effective control method controlling different overloaded traffic.

These methods are effective for solution high capacity or traffic overload to a certain extent above, certain fairness and validity can be met, but it is larger to service impact that it controls result, processed service request is usually caused to fluctuate larger, load control effects is unsatisfactory, and does not consider the practical problem of the high energy consumption of soft switch gateway, poor efficiency and low-energy-efficiency.

Summary of the invention

For the deficiencies in the prior art, the present invention proposes a kind of level and smooth suppressing method of load based on power communication soft switch gateway, to reach the object improving soft switch gateway efficiency in existing powerline network.

Based on the level and smooth suppressing method of load of power communication soft switch gateway, comprise the following steps:

Step 1, according to the arrival rate of number of services, different business in this sense cycle and processing speed, obtain the utilance of CPU, and according to the free space in services cache space, obtain services cache space availability ratio;

Step 2, judge whether soft switch gateway transships, namely judge whether the utilance of CPU is greater than threshold value set according to demand, judge whether services cache space availability ratio is greater than threshold value set according to demand simultaneously, if above-mentioned, both are greater than threshold value simultaneously, then soft switch gateway overload, stop the business in next cycle of reception and perform step 3, otherwise, extend sense cycle and also return execution step 1;

Step 3, according to the total throughout of gateway with process the total energy consumption that whole business consumes, in conjunction with the restrictive condition of different business speed control, adopt ant group algorithm obtain efficiency maximum time different business speed control;

Step 3-1, be set in the qualifications of different business speed control in premature beats process, comprise the lower limit that different business speed control is greater than its speed control, cpu busy percentage is less than or equal to set threshold value;

Step 3-2, according to number of services in Business Processing cycle of this gateway and this cycle, set up the relation of this gateway total throughout and different business speed control;

Step 3-3, according to basal energy expenditure power proportion in number of services, this gateway link capacity and link in Business Processing cycle of this gateway, this cycle, set up the relation of total energy consumption that all business of this gateway processes consume and different business speed control;

The relation of the relation of step 3-4, restrictive condition, gateway total throughout and different business speed control in conjunction with different business speed control, total energy consumption that all business of this gateway processes consume and different business speed control, sets up efficiency Optimized model;

Different business speed control when step 3-5, employing ant group algorithm acquisition efficiency are maximum;

Step 3-6, different business speed control when adopting level and smooth suppressing method maximum to efficiency carry out premature beats, obtain the final speed control of different business;

Step 4, according to the final speed control of different business obtained, counter is set, namely sets the quantity of reception business in next cycle, realize the control that actual gateway is transshipped;

Step 5, judge that in this cycle, whether all business have all processed, and if so, then shorten sense cycle, wait-receiving mode new business, and return and perform step 1; Otherwise continue to process surplus lines.

The arrival rate according to number of services, different business described in step 1 and processing speed, obtain the utilance of CPU;

Cpu busy percentage ρ formula is as follows:

$\mathrm{\ρ}=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}(v_i^A/v_i^P)---\left(1\right)$

Wherein, n represents number of services, represent the arrival rate of business i in this cycle, A represents Arrive, 1≤i≤n, the processing speed of expression business i, P represents process;

The described free space according to services cache space, obtain services cache space availability ratio η, formula is as follows:

η＝(P _o-P _c)/P _o（2）

Wherein, P _ofor services cache queue total length; P _crepresent the free space in services cache space.

Set up efficiency Optimized model described in step 3-4, formula is as follows:

$\left\{\begin{array}{c}{v}_i^C=\arg \max J\left(v_i^C\right)\\ s.t.v_i^C\≥v_{i\_\min }^C\\ \underset{i=1}{\overset{n}{\mathrm{\Σ}}}{v}_i^C/v_i^P\≤{\mathrm{\ρ}}^{*}\\ 1\≤i\≤n\end{array}\right.---\left(3\right)$

Wherein, the speed control of expression business i, C represents Control, the minimum speed control of expression business i; N represents number of services; ρ ^*represent cpu busy percentage threshold value, the processing speed of expression business i, P represents process; $J=P_{\mathrm{total}}//E_{\mathrm{total}}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{v}_i^C/({\mathrm{n\&beta;c}}_e+\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(1-\mathrm{\&beta;})v_i^C),$ P _tatalthe total throughout of gateway in indication cycle, E _totalrepresent the total energy consumption that process this cycle all business consume; c _erepresent link capacity, β represents basal energy expenditure power proportion in link, 0 < β < 1.

Different business speed control when employing ant group algorithm acquisition efficiency described in step 3-5 is maximum, specifically comprises the following steps:

Step 3-5-1, carry out initialization to ant group algorithm, namely arrange the number M that total iterations is A, solution vector, described each solution vector is made up of the processing speed of all business in certain iteration;

Arrange iterations a=1, meeting under the restrictive condition in efficiency Optimized model, draw the initial solution of speed control vector as algorithm of one group of different business at random, solution vector matrix is be expressed as:

$V_{M\&times;n}^{C\_a}=\begin{array}{c}\left[\begin{array}{c}{V}_1^{C\_a}\\ ...\\ V_m^{C\_a}\\ ...\\ V_M^{C\_a}\end{array}\right]=\left[\begin{array}{ccccc}{v}_{1\_1}^{C\_a}& ...& v_{1\_i}^{C\_a}& ...& v_{1\_n}^{C\_a}\\ ...& ...& ...& ...& ...\\ v_{m\_1}^{C\_a}& ...& v_{m\_i}^{C\_a}& ...& v_{m\_n}^{C\_a}\\ ...& ...& ...& ...& ...\\ v_{M\_1}^{C\_a}& ...& v_{M\_i}^{C\_a}& ...& v_{M\_n}^{C\_a}\end{array}\right]---\left(4\right)\end{array}$

Wherein, represent the first to the M group solution vector; represent in a time iteration, the control rate of first group of solution vector, first business is to the speed control of M group solution vector n-th business, and n represents number of services;

The pheromones intensity arranging M solution vector is Q, initial information prime matrix Plist _afor:

Wherein, ρ ₁..., ρ _m.., ρ _mrepresent the pheromones of first solution vector to M solution vector respectively;

The efficiency matrix J of homography _afor:

$J_a=[J\left(V_1^{C\_a}\right),J\left(V_2^{C\_a}\right),...,J\left(V_M^{C\_a}\right)]---\left(6\right)$

Now iterations a is 1;

Speed control in step 3-5-2, selection efficiency matrix corresponding to efficiency minimum value, Gaussian mutation is carried out to remaining M-1 group solution in solution vector matrix, if the solution vector gained energy valid value after variation is greater than the energy valid value of former solution vector, then solution vector matrix is upgraded, the solution vector after variation is replaced former solution vector; If the solution vector gained energy valid value after variation is less than or equal to the energy valid value of original solution vector, then solution vector is upgraded by setting probability according to roulette algorithm, the value that even roulette algorithm draws is less than or equal to setting probability, then carry out renewal to solution vector matrix to replace, if be greater than setting probability, do not carry out renewal and replace, finally obtain the solution vector matrix after upgrading;

Step 3-5-3, according to efficiency matrix J _amiddle efficiency maximum and efficiency minimum value, obtain the Pheromone update coefficient μ in each solution vector _m, the pheromones ρ of M solution vector after renewal _m' be:

ρ _m′=ρ _m(1-Rho)+Qμ _m(7)

Wherein, μ _m=(J (V _m)-J _min)/(J _max-J _min), J (V _m) represent solution vector V _mcorresponding energy valid value, J _maxrepresent efficiency maximum, J _minrepresent efficiency minimum value, Rho represents pheromones volatility coefficient, 0 < Rho < 1; And to Pheromone Matrix Plist _aupgrade, obtain the Pheromone Matrix after upgrading, efficiency matrix is upgraded simultaneously, obtain the efficiency matrix after upgrading;

In step 3-5-4, solution vector matrix in the updated, random selecting x group solution vector makes a variation:

Group number x formula is as follows:

x＝M×(A+a)/2A（8）

Select pheromones in x group solution vector to be worth the speed control corresponding to maximum solution vector according to the Pheromone Matrix after upgrading, according to following formula, the x a chosen solution vector made a variation:

$V_m^{\text{'}C\_a}=(1-w)V_m^{C\_a}+w\&times;V_{\max \_x}^{C\_a}---\left(9\right)$

Wherein, 0 < w < 1; represent the solution vector after variation, represent that in x group solution vector, pheromones is worth the speed control corresponding to maximum solution vector;

If the solution vector gained energy valid value after variation is greater than the energy valid value of former solution vector, then the solution vector matrix after renewal is upgraded again, the solution vector be about to after variation replaces former solution vector, if the solution vector gained energy valid value after variation is less than the energy valid value of original solution vector, then adopt roulette algorithm to upgrade by probability, and the efficiency matrix after renewal and the Pheromone Matrix after upgrading are upgraded again;

Step 3-5-5, judge whether iterations reaches set point, if so, then obtain final efficiency matrix and perform step 2-5-6; Otherwise return and perform step 2-5-2;

Step 3-5-6, in final efficiency matrix, select can valid value maximum time corresponding different business speed control, namely obtain the different business speed control that this gateway efficiency is maximum formula is as follows:

$V_{\max }^C=\arg \left(\max \left(J_a\left(V_m^{C\_a}\right)\right)\right)=[v_1^{C\_\max },...,v_i^{C\_\max },...,v_n^{C\_\max }]---\left(10\right)$

Different business speed control when the level and smooth suppressing method of employing described in step 3-6 is maximum to efficiency carries out premature beats, and formula is as follows:

$v_i^C=v_i^A\&times;(1-e)---\left(11\right)$

Wherein, the speed control of expression business i, C represents Control, represent the arrival rate of business i in this cycle, A represents Arrive, and 1≤i≤n, n represents number of services, and e represents level and smooth rejection coefficient,

Sense cycle described in step 1 is α T ₀, α ∈ [1, T _m/ T ₀], wherein, T ₀for the least measuring time cycle, T _mfor the maximum measuring time cycle.

Prolongation sense cycle described in step 2, namely sense cycle maximum is regulated, be specially: sense cycle maximum is multiplied by adjustment factor, obtain new sense cycle maximum, the span of adjustment factor is 1 ~ 2, if the sense cycle maximum after regulating is greater than sense cycle maximum higher limit, then sense cycle maximum is extended for sense cycle maximum higher limit.

Shortening sense cycle described in step 5, namely regulates sense cycle maximum, is specially: sense cycle maximum is multiplied by adjustment factor, obtains new sense cycle maximum, and the span of adjustment factor is T ₀/ T _m~ 1, wherein, T ₀for the least measuring time cycle, T _mfor the maximum measuring time cycle.

Advantage of the present invention:

A kind of level and smooth suppressing method of load based on power communication soft switch gateway of the present invention, by considering fairness and the validity of load suppressing method, with soft switch gateway traffic overload for rejection condition, detect cpu busy percentage and the service spatial cache utilance of soft switch gateway, determine whether soft switch gateway is in traffic overload, and set up Model for Multi-Objective Optimization, utilize the optimization of ant group swarm intelligence to obtain the optimal feasible solution of this model, realize the level and smooth suppression to different business; The optimal feasible solution that the present invention proposes to be obtained by ant colony intelligence optimization multiple target efficiency Optimized model innovatively may carry out the effectively suppression of justice by overloaded traffic to different; The present invention proposes to adopt the level and smooth mode suppressed to prevent soft switch gateway load from occurring catastrophe, thus effectively can improve systematic function.

Accompanying drawing explanation

Fig. 1 is the level and smooth suppressing method flow chart of the load based on power communication soft switch gateway of an embodiment of the present invention;

Fig. 2 is that the load of an embodiment of the present invention smoothly suppresses system block diagram;

Fig. 3 is that the load of an embodiment of the present invention smoothly suppresses iterations on the impact of efficiency;

Fig. 4 is that the load of an embodiment of the present invention smoothly suppresses cpu busy percentage;

Fig. 5 is that the load of an embodiment of the present invention smoothly suppresses multi-service Energy Efficiency Analysis.

Embodiment

Below in conjunction with accompanying drawing, a kind of embodiment of invention is described further.

Based on the level and smooth suppressing method of load of power communication soft switch gateway, method flow diagram as shown in Figure 1, comprises the following steps:

Step 1, according to the arrival rate of number of services, different business in this sense cycle and processing speed, obtain the utilance of CPU, and according to the free space in services cache space, obtain services cache space availability ratio;

As shown in Figure 2, for soft switch gateway in the embodiment of the present invention, server is processed by the service request of soft switch gateway to different business; Wherein at server end, services cache queue total length is 200; In 160 traffic handling time units T (T=1s), carry out Service control to the different business of soft switch gateway in embodiment, there is overload in the business in cycle 20T ~ 100T that is arranged on, arranges T ₀=100ms, arranges T _mscope be 1 ~ 3s, initial T _m=2s; Three different business S are set ₁, S ₂, S ₃carry out premature beats, be all 5/second in the arrival rate of time period three business of nonoverload, in the time period 20T ~ 100T of overload, the arrival rate of three business is all 100/second, and its processing speed is respectively the arrival rate arranging three kinds of business for the condition that meets fairness and validity when premature beats is all satisfied $v_i^a\&GreaterEqual;\mathrm{10,1}\&le;i\&le;3.$

In the embodiment of the present invention, dynamic sampling is adopted to measure soft switch gateway cpu busy percentage ρ and service spatial cache utilance η; To measure step-length α T ₀dynamic random measures the cpu busy percentage ρ of soft switch gateway, measures the free space P between services cache queue empty simultaneously _c, and draw services cache space availability ratio η;

Cpu busy percentage ρ formula is as follows:

$\mathrm{\&rho;}=\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}(v_i^A/v_i^P)---\left(1\right)$

Obtain services cache space availability ratio η, formula is as follows:

η＝(P _o-P _c)/P _o（2）

Step 2, judge whether soft switch gateway transships, namely judge whether the utilance of CPU is greater than threshold value set according to demand, judge whether services cache space availability ratio is greater than threshold value set according to demand simultaneously, if above-mentioned, both are greater than threshold value simultaneously, then soft switch gateway overload, stop the business in next cycle of reception and perform step 3, otherwise, extend sense cycle and also return execution step 1;

In the embodiment of the present invention, setting corresponds to the utilance ρ of CPU, the overload threshold ρ of services cache space availability ratio η ^*and η ^*, ρ ^*=0.9, η ^*=0.9; Initializing variable l _ρ=0 and l _η=0, perform following full load degree deterministic process: if ρ ^*≤ ρ, soft switch gateway cpu load transships, then l _ρ=1; If η ^*≤ η, the load of services cache space is transshipped, then l _η=1.Finally, following judgement is done: if (l _ρaMP.AMp.Amp l _η)=1, then soft switch gateway overload;

In the embodiment of the present invention, extend sense cycle, namely sense cycle maximum regulated, be specially:

T _m′=βT _m(12)

Wherein, T _m' represent the rear sense cycle maximum of adjustment; β is regulating system, and 1 < β < 2; If T _m' > T _u, wherein, T _ufor maximum measuring time upper cycle limit, make T _m'=T _u, wherein T _u=3;

Step 3, according to the total throughout of gateway with process the total energy consumption that whole business consumes, in conjunction with the restrictive condition of different business speed control, adopt ant group algorithm obtain efficiency maximum time different business speed control;

Step 3-1, be set in the qualifications of different business speed control in premature beats process, comprise different business speed control and be greater than minimum speed control, cpu busy percentage is less than or equal to set threshold value;

The span of different business speed control in load process of inhibition is set according to fair principle:

$v_i^C\&GreaterEqual;10---\left(13\right)$

According to validity principle, cpu busy percentage must be had can not to be greater than its threshold value, so the speed control of different business also demand fulfillment:

$\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}{v}_i^C/v_i^P\&le;0.9---\left(14\right)$

Step 3-2, according to number of services in Business Processing cycle of this gateway and this cycle, set up the relation of this gateway total throughout and different business speed control;

Formula is as follows:

$P_{\mathrm{total}}=T\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}{v}_i^C---\left(15\right)$

Wherein, T is process service period;

Step 3-3, according to basal energy expenditure power proportion in number of services, this gateway link capacity and link in Business Processing cycle of this gateway, this cycle, set up the relation of total energy consumption that all business of this gateway processes consume and different business speed control;

$E_{\mathrm{total}}=T(3{\mathrm{\&beta;c}}_e+\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}(1-\mathrm{\&beta;})v_i^C)---\left(16\right)$

The relation of the relation of step 3-4, restrictive condition, gateway total throughout and different business speed control in conjunction with different business speed control, total energy consumption that all business of this gateway processes consume and different business speed control, sets up efficiency Optimized model;

Efficiency Optimized model formula is as follows:

$\left\{\begin{array}{c}{v}_i^C=\arg \max J\left(v_i^C\right)\\ s.t.v_i^C\&GreaterEqual;10\\ \underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{v}_i^C/v_i^P\&le;0.9\\ 1\&le;i\&le;3\end{array}\right.---\left(3\right)$

Wherein, $J=P_{\mathrm{total}}/E_{\mathrm{total}}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{v}_i^C/({\mathrm{n\&beta;c}}_e+\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(1-\mathrm{\&beta;})v_i^C)$

Different business speed control when step 3-5, employing ant group algorithm acquisition efficiency are maximum, specifically comprises the following steps:

Step 3-5-1, carry out initialization to ant group algorithm, namely arrange the number M that total iterations is A, solution vector, described each solution vector is made up of the processing speed of all business in certain iteration;

In the embodiment of the present invention, arranging total iterations is 25 times, and ant (solution vector) number is 20, so the solution vector of 20 of correspondence ants is 20 × 3 matrixes;

Arrange iterations a=1, meeting under the restrictive condition in efficiency Optimized model, draw the initial solution of speed control vector as algorithm of one group of different business at random, solution vector matrix is be expressed as:

$V_{20\&times;3}^{C\_a}=\left[\begin{array}{ccc}{v}_{1\_1}^{C\_a}& v_{1\_2}^{C\_a}& v_{1\_3}^{C\_a}\\ ...& ...& ...\\ v_{m\_1}^{C\_a}& v_{m\_2}^{C\_a}& v_{m\_3}^{C\_a}\\ ...& ...& ...\\ v_{20\_1}^{C\_a}& v_{20\_2}^{C\_a}& v_{20\_3}^{C\_a}\end{array}\right]---\left(4\right)$

Wherein, represent in a time iteration, the control rate of first group of solution vector, first business is to the speed control of the 20th group of solution vector the 3rd business;

The pheromones intensity arranging M solution vector is Q=1, initial information prime matrix Plist _afor:

The efficiency matrix J of homography _afor:

$J_a=[J\left(V_1^{C\_a}\right),J\left(V_2^{C\_a}\right),...,J\left(V_{20}^{C\_a}\right)]---\left(6\right)$

Now iterations a is 1;

Step 3-5-2, selection efficiency matrix J _aspeed control corresponding to middle efficiency minimum value, to solution vector matrix in remaining M-1 group solution carry out Gaussian mutation, if the solution vector gained energy valid value after variation is greater than the energy valid value of former solution vector, then to solution vector matrix upgrade, the solution vector after variation is replaced former solution vector; If the solution vector gained energy valid value after variation is less than or equal to the energy valid value of original solution vector, then press probability P according to roulette algorithm _m(0 < P _m< 0.1) solution vector is upgraded, P in the present embodiment _m=0.08, the value that even roulette algorithm draws is less than or equal to P _m, then to solution vector matrix carry out renewal to replace, if be greater than P _mthen do not carry out renewal to replace, finally obtain the solution vector matrix after upgrading;

Step 3-5-3, according to efficiency matrix J _amiddle efficiency maximum and efficiency minimum value, obtain the Pheromone update coefficient μ in each solution vector _m, the pheromones ρ of M solution vector after renewal _m' be:

ρ _m′=ρ _m(1-Rho)+Qμ _m(7)

Wherein, pheromones volatility coefficient is Rho=0.5; And to Pheromone Matrix Plist _aupgrade, obtain the Pheromone Matrix after upgrading, efficiency matrix is upgraded simultaneously, obtain the efficiency matrix after upgrading;

In step 3-5-4, solution vector matrix in the updated, random selecting x group solution vector makes a variation:

Group number x formula is as follows:

x＝M×(A+a)/2A（8）

During a=1, namely during first time iteration, from the solution vector matrix after renewal, Stochastic choice goes out x=10 group solution vector, select pheromones in these 10 groups of solution vectors according to the Pheromone Matrix after renewal and be worth the speed control corresponding to maximum solution vector, according to following formula, these 10 solution vectors chosen are made a variation:

$V_m^{\text{'}C\_a}=(1-w)V_m^{C\_a}+w\&times;V_{\max \_x}^{C\_a}---\left(9\right)$

Wherein, w=0.6;

If the solution vector gained energy valid value after variation is greater than the energy valid value of former solution vector, then the solution vector matrix after renewal is upgraded again, the solution vector be about to after variation replaces former solution vector, if the solution vector gained energy valid value after variation is less than the energy valid value of original solution vector, then adopt roulette algorithm to upgrade by probability, and the efficiency matrix after renewal and the Pheromone Matrix after upgrading are upgraded again;

Step 3-5-5, judge whether iterations reaches set point, if so, then obtain final efficiency matrix and perform step 2-5-6; Otherwise return and perform step 2-5-2;

Step 3-5-6, in final efficiency matrix, select can valid value maximum time corresponding different business speed control, namely obtain the different business speed control that this gateway efficiency is maximum formula is as follows:

$V_{\max }^C=\arg \left(\max \left(J_a\left(V_m^{C-a}\right)\right)\right)=\left[v_1^{C\_\max }{v}_2^{C\_\max }{v}_3^{C\_\max }\right]---\left(10\right)$

Step 3-6, different business speed control when adopting level and smooth suppressing method maximum to efficiency carry out premature beats, obtain the final speed control of different business; Formula is as follows:

$v_i^C=v_i^A\&times;(1-e)---\left(11\right)$

Wherein, represent the arrival rate of business i in this cycle, A represents Arrive, and 1≤i≤n, n represents number of services, and e represents level and smooth rejection coefficient, through the iteration in multiple premature beats cycle, level and smooth rejection coefficient e levels off to zero, the value that the arrival rate of business also levels off in formula (10);

Judging whether level and smooth rejection coefficient meets e < 0, as met, then stopping suppressing and performing step 4; Otherwise, return and perform the level and smooth suppression of step 3-6 continuation execution.

Step 4, according to the final speed control of different business obtained, counter is set, namely sets the quantity of reception business in next cycle, realize the control that actual gateway is transshipped;

The final speed control of different business feeds back to the counter in the soft exchanging network Central Shanxi Plain, counter can enter the quantity of the different business of soft switch gateway according to the corresponding often kind of business of speed control value setting that these feed back, thus speed control is carried out to different business: when the value of certain business in corresponding counter is non-vanishing, just allow this service request from soft switch gateway by and processed by server, and often subtract one by value corresponding a traffic meter with regard to corresponding from soft switch gateway, if when the value in the counter of certain business in correspondence is zero, in this treatment cycle, soft switch gateway can not continue to receive this business,

Step 5, judge that in this cycle, whether all business have all processed, and if so, then shorten sense cycle, wait-receiving mode new business, and return and perform step 1; Otherwise continue to process surplus lines.

Shorten sense cycle, namely sense cycle maximum is regulated, be specially: sense cycle maximum is multiplied by adjustment factor T _m=zT _m, wherein z is regulating system, and T ₀/ T _m< z < 1, obtains new sense cycle maximum.

As shown in Figure 3, the level and smooth suppressing method of this load only needs iteration about 25 times, just can draw the optimal solution of energetic efficiency objectives function, has good convergence in this way.

As shown in Figure 4, the level and smooth suppressing method of this load has good validity, can meet the restrictive condition of its nonoverload; And can find out and utilize this level and smooth suppressing method, when premature beats, cpu busy percentage slowly rises until its threshold value, plays a very good protection like this to server.

As shown in Figure 5, the efficiency change in the level and smooth process of inhibition of load of three kinds of business, under the principle meeting fairness, the efficiency of three kinds of business all reaches optimum, is similar to the effect that Fig. 4 smoothly suppresses and is also represented in Figure 5.

Claims (5)

1., based on the level and smooth suppressing method of load of power communication soft switch gateway, it is characterized in that, comprise the following steps:

Step 1, according to the arrival rate of number of services, different business in sense cycle and processing speed, obtain the utilance of CPU, and according to the free space in services cache space, obtain services cache space availability ratio;

Step 2, judge whether soft switch gateway transships, namely judge whether the utilance of CPU is greater than threshold value set according to demand, judge whether services cache space availability ratio is greater than threshold value set according to demand simultaneously, if above-mentioned, both are greater than threshold value simultaneously, then soft switch gateway overload, stop the business in next cycle of reception and perform step 3, otherwise, extend sense cycle and also return execution step 1;

Step 3, according to the total throughout of gateway with process the total energy consumption that whole business consumes, in conjunction with the restrictive condition of different business speed control, adopt ant group algorithm obtain efficiency maximum time different business speed control;

Step 3-1, be set in the qualifications of different business speed control in premature beats process, comprise the lower limit that different business speed control is greater than its speed control, cpu busy percentage is less than or equal to set threshold value;

Step 3-2, according to number of services in Business Processing cycle of this gateway and this cycle, set up the relation of this gateway total throughout and different business speed control;

Step 3-3, according to basal energy expenditure power proportion in number of services, this gateway link capacity and link in Business Processing cycle of this gateway, this cycle, set up the relation of total energy consumption that all business of this gateway processes consume and different business speed control;

The relation of the relation of step 3-4, restrictive condition, gateway total throughout and different business speed control in conjunction with different business speed control, total energy consumption that all business of this gateway processes consume and different business speed control, sets up efficiency Optimized model;

Set up efficiency Optimized model, formula is as follows:

Wherein, the speed control of expression business i, C represents Control, the minimum speed control of expression business i; N represents number of services; ρ ^*represent cpu busy percentage threshold value, the processing speed of expression business i, P represents process; p _tatalthe total throughout of gateway in indication cycle, E _totalrepresent the total energy consumption that process this cycle all business consume; c _erepresent link capacity, β represents basal energy expenditure power proportion in link, 0 < β < 1;

Different business speed control when step 3-5, employing ant group algorithm acquisition efficiency are maximum; Specifically comprise the following steps:

Step 3-5-1, carry out initialization to ant group algorithm, namely arrange the number M that total iterations is A, solution vector, each solution vector is made up of the processing speed of all business in certain iteration;

Arrange iterations a=1, meeting under the restrictive condition in efficiency Optimized model, draw the initial solution of speed control vector as algorithm of one group of different business at random, solution vector matrix is be expressed as:

Wherein, represent the first to the M group solution vector; represent in a time iteration, the control rate of first group of solution vector, first business is to the speed control of M group solution vector n-th business, and n represents number of services;

The pheromones intensity arranging M solution vector is Q, initial information prime matrix Plist _afor:

Wherein, ρ ₁..., ρ _m..., ρ _mrepresent the pheromones of first solution vector to M solution vector respectively;

The efficiency matrix J of homography _afor:

Now iterations a is 1;

Step 3-5-2, selection efficiency matrix J _aspeed control corresponding to middle efficiency minimum value, to solution vector matrix in remaining M-1 group solution carry out Gaussian mutation, if the solution vector gained energy valid value after variation is greater than the energy valid value of former solution vector, then to solution vector matrix upgrade, the solution vector after variation is replaced former solution vector; If the solution vector gained energy valid value after variation is less than or equal to the energy valid value of original solution vector, then upgrade solution vector by probability P m according to roulette algorithm, the value that even roulette algorithm draws is less than or equal to Pm, then to solution vector matrix carry out renewal to replace, if be greater than Pm, do not carry out renewal and replace, finally obtain the solution vector matrix after upgrading;

Step 3-5-3, according to efficiency matrix J _amiddle efficiency maximum and efficiency minimum value, obtain the Pheromone update coefficient μ in each solution vector _m, the pheromones ρ of M solution vector after renewal _m' be:

ρ′ _m＝ρ _m(1-Rho)+Qμ _m(7)

Wherein, μ _m=(J (V _m)-J _min)/(J _max-J _min), J _maxrepresent efficiency maximum, J _minrepresent efficiency minimum value, J (V _m) represent solution vector V _mcorresponding energy valid value, Rho represents pheromones volatility coefficient, 0 < Rho < 1; And to Pheromone Matrix Plist _aupgrade, obtain the Pheromone Matrix after upgrading, efficiency matrix is upgraded simultaneously, obtain the efficiency matrix after upgrading;

In step 3-5-4, solution vector matrix in the updated, random selecting x group solution vector makes a variation:

Group number x formula is as follows:

x＝M×(A+a)/2A(8)

Select pheromones in x group solution vector to be worth the speed control corresponding to maximum solution vector according to the Pheromone Matrix after upgrading, according to following formula, the x a chosen solution vector made a variation:

Wherein, 0 < w < 1; represent the solution vector after variation, represent that in x group solution vector, pheromones is worth the speed control corresponding to maximum solution vector;

If the solution vector gained energy valid value after variation is greater than the energy valid value of former solution vector, then the solution vector matrix after renewal is upgraded again, the solution vector be about to after variation replaces former solution vector, if the solution vector gained energy valid value after variation is less than the energy valid value of original solution vector, then adopt roulette algorithm to upgrade by probability, and the efficiency matrix after renewal and the Pheromone Matrix after upgrading are upgraded again;

Step 3-5-5, judge whether iterations reaches set point, if so, then obtain final efficiency matrix and perform step 3-5-6; Otherwise return and perform step 3-5-2;

Step 3-5-6, in final efficiency matrix, select can valid value maximum time corresponding different business speed control, namely obtain the different business speed control that this gateway efficiency is maximum formula is as follows:

Step 3-6, different business speed control when adopting level and smooth suppressing method maximum to efficiency carry out premature beats, obtain the final speed control of different business; Formula is as follows:

Wherein, the speed control of expression business i, C represents Control, represent the arrival rate of business i in this cycle, A represents Arrive, and 1≤i≤n, n represents number of services, and e represents level and smooth rejection coefficient, λ > 1;

Step 4, according to the final speed control of different business obtained, counter is set, namely sets the quantity of reception business in next cycle, realize the control that actual gateway is transshipped;

Step 5, judge that in this cycle, whether all business have all processed, and if so, then shorten sense cycle, wait-receiving mode new business, and return and perform step 1; Otherwise continue to process surplus lines.

2. the level and smooth suppressing method of load based on power communication soft switch gateway according to claim 1, is characterized in that, the arrival rate according to number of services, different business described in step 1 and processing speed, obtains the utilance of CPU;

Cpu busy percentage ρ formula is as follows:

Wherein, n represents number of services, represent the arrival rate of business i in this cycle, A represents Arrive, 1≤i≤n, the processing speed of expression business i, P represents process;

The described free space according to services cache space, obtain services cache space availability ratio η, formula is as follows:

η＝(P _o-P _c)/P _o(2)

Wherein, P _ofor services cache queue total length; P _crepresent the free space in services cache space.

3. the level and smooth suppressing method of load based on power communication soft switch gateway according to claim 1, it is characterized in that, the sense cycle described in step 1 is α T ₀, α ∈ [1, T _m/ T ₀], wherein, T ₀for the least measuring time cycle, T _mfor the maximum measuring time cycle.

4. the level and smooth suppressing method of load based on power communication soft switch gateway according to claim 1, it is characterized in that, prolongation sense cycle described in step 2, namely sense cycle maximum is regulated, be specially: sense cycle maximum is multiplied by adjustment factor, obtain new sense cycle maximum, the span of adjustment factor is 1 ~ 2, if the sense cycle maximum after regulating is greater than sense cycle maximum higher limit, then sense cycle maximum is extended for sense cycle maximum higher limit.

5. the level and smooth suppressing method of load based on power communication soft switch gateway according to claim 1, it is characterized in that, shortening sense cycle described in step 5, namely sense cycle maximum is regulated, be specially: sense cycle maximum is multiplied by adjustment factor, obtain new sense cycle maximum, the span of adjustment factor is T ₀/ T _m~ 1, wherein, T ₀for the least measuring time cycle, T _mfor the maximum measuring time cycle.