CN100571200C - A kind of fuzzy PD flow control method that is used for the congested control of communication network - Google Patents
A kind of fuzzy PD flow control method that is used for the congested control of communication network Download PDFInfo
- Publication number
- CN100571200C CN100571200C CNB2005101279484A CN200510127948A CN100571200C CN 100571200 C CN100571200 C CN 100571200C CN B2005101279484 A CNB2005101279484 A CN B2005101279484A CN 200510127948 A CN200510127948 A CN 200510127948A CN 100571200 C CN100571200 C CN 100571200C
- Authority
- CN
- China
- Prior art keywords
- fuzzy
- departure
- controller
- rate
- switch
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Abstract
The present invention relates to a kind of fuzzy PD flow control method and equipment that is used for the congested control of communication network, scheme is as follows: every the queue length of a time cycle sampling buffer; Predefined congestion threshold is deducted the controlled error e of described queue length (n); Described departure is deducted value that departure that last time cycle obtains obtains divided by time cycle controlled error change rate Δ e (n); If this departure during greater than certain prior preset threshold, is selected fuzzy controller work; When if this departure is not more than certain prior preset threshold, select the work of PD controller; Adopt method and apparatus of the present invention, realize fairly and reasonably distributing corresponding Internet resources to the terminal use, the data traffic of dynamic adjustments network effectively in real time, Control Network congested guarantees terminal use's service quality.
Description
Technical field
The present invention relates to a kind of communication network congestion control method that is used for, relate in particular to a kind of fuzzy PD flow control method that is used for the congested control of communication network.
Background technology
Growing and mutual fusion along with telecommunication, multimedia and computer industry and technology, data communication network, as computer network, just develop towards high speed, broadband, multimedization direction, this has impelled popularizing of Internet network, makes the data communication network terminal use become volatile growth simultaneously.On the other hand, fast development along with mobile communication technology, mobile communications network, as second generation mobile communications network GSM/GPRS network, CDMA1X network and 3G (Third Generation) Moblie network such as WCDMA network, CDMA2000EV/DO network, TD-SCDMA network also or will obtain in the world popularizing rapidly, this impels mobile communication terminal user that the growth of geometric progression has taken place equally.All of these factors taken together has all caused the carrier as transfer of data--the increase at full speed of data traffic in the data communication network, thereby caused the reduction of network performance and the degradation of service quality, in full reportedly defeated time delay increases, time jitter increases, the data-bag lost rate rises or the like.
For reducing also and then avoiding these phenomenons to take place, relevant international standard has all defined relevant solution framework.Formulate as the Internet international standard and to organize IETF (The Internet Engineering Task Force--the Internet engineering duty group) the data traffic control framework in the Internet that proposed a plurality of RFC normalized definitions, atm forum (ATM Forum) has also defined its flow control framework in its service management standard simultaneously.A common feature of these control frameworks is exactly to have adopted the method for FEEDBACK CONTROL to carry out the flow control of data in each automatic network.
Existing flow control mechanism can be divided into two classes according to the feedback mechanism difference that is adopted: binary system feedback mechanism and explicit rate mechanism.Because explicit rate mechanism is primarily aimed at explicit rate mechanism at present and designs, and propose different control methods with respect to the advantage of binary system feedback mechanism.Yet the major defect of these methods is controlling mechanism or intuitively, do not provide any formal method for designing to guarantee the stability of closed-loop control; Though perhaps provide the assurance of machine-processed stability, owing to the reason of complexity is difficult to realize; Though perhaps satisfy preceding two requirements, the transient response of mechanism exists bigger deficiency, all has deficiency as fairness, robustness, anti-interference.
Summary of the invention
Purpose of the present invention just provides the method for designing of The fuzzy PD in a kind of communication network (Fuzzy PD) flow controller, thereby can be according to the grade of service that the terminal use applied for, fairly and reasonably distribute corresponding Internet resources to the terminal use, the real-time data traffic of dynamic adjustments network effectively, Control Network congested, guarantee terminal use's service quality (QoS), the present invention adopts following scheme to realize purpose of the present invention:
The invention provides a kind of fuzzy PD flow control method that is used for the congested control of communication network, comprise the steps:
1) every the queue length of a time cycle sampling buffer;
2) predefined congestion threshold is deducted the controlled error e of described queue length (n);
3) described departure is deducted departure value of obtaining of last time cycle acquisition divided by time cycle controlled error change rate Δ e (n);
4) judge in step 2) in the departure that obtains, if this departure is during greater than certain prior preset threshold, select fuzzy controller, the rate of change of described departure and described departure is input to fuzzy controller, described fuzzy controller output switch supported demonstration speed r (n+1) is used for Controlling Source end speed; When if this departure is not more than certain prior preset threshold, select the PD controller, the rate of change of described departure and described departure is input to the PD controller, described PD controller is exported supported demonstration speed r (n+1);
Wherein n is an integer.
After described fuzzy controller receives the rate of change of described departure and described departure, can carry out following processing:
1) described departure and described rate of change are carried out Fuzzy processing, obtain corresponding fuzzy language variable;
2) according to rule base described fuzzy language variable is carried out the fuzzy language variable Δ ER that reasoning obtains the increment of feedback velocity parameter ER;
3) the fuzzy language variable Δ ER with the increment of described feedback velocity parameter ER carries out the output controlled quentity controlled variable Δ er (n+1) that ambiguity solution obtains fuzzy controller, obtains the supported demonstration speed of switch r (n+1) according to described output controlled quentity controlled variable.
Described fuzzy controller calculates the supported demonstration speed of switch r (n+1): r (n+1)=Sat according to following formula
β * C{ r (n)+Δ er (n+1) }, wherein C is the bandwidth of link, and β is the bandwidth constraints factor, and r (n) is the supported explicit rate of switch that is calculated in [(n-1) T, nT] time interval, and Sat is a truncation funcation.
Described PD controller is according to formula r (n+1)=max{r (n)+K
PE (n)+K
DΔ e (n), 0} calculate supported demonstration speed r (n+1), and wherein, r (n) is the supported explicit rate of switch that is calculated in [(n-1) T, nT] time interval, K
PAnd K
DBe called proportionality constant and derivative constant.
The present invention also provides a kind of communication system, comprises at least one switch, and described switch comprises a The fuzzy PD flow controller, and described The fuzzy PD flow controller comprises:
Sampling and computing unit, be used for queue length every a time cycle sampling buffer, predefined congestion threshold is deducted the controlled error e of described queue length (n) and described departure is deducted departure value of obtaining that last time cycle obtains divided by time cycle controlled error change rate Δ e (n), described departure and described rate of change are outputed to the mode selector switch;
The mode selector switch is judged described departure, if this departure during greater than certain prior preset threshold, is selected fuzzy controller, the rate of change of described departure and described departure is input to fuzzy controller; When if this departure is not more than certain prior preset threshold, select the PD controller, the rate of change of described departure and described departure is input to the PD controller;
Fuzzy controller is exported switch supported demonstration speed r (n+1) according to the described departure of input and the rate of change of described departure, is used for Controlling Source end speed;
The PD controller is exported switch supported demonstration speed r (n+1) according to the described departure of input and the rate of change of described departure, is used for Controlling Source end speed; Wherein n is an integer.
Described fuzzy controller comprises:
Fuzzier unit is carried out Fuzzy processing with described departure and described rate of change, obtains corresponding fuzzy language variable;
The inference machine unit carries out the fuzzy language variable Δ ER that reasoning obtains the increment of feedback velocity parameter ER according to rule base to described fuzzy language variable;
The ambiguity solution unit carries out the output controlled quentity controlled variable Δ er (n+1) that ambiguity solution obtains fuzzy controller with the fuzzy language variable Δ ER of the increment of described feedback velocity parameter ER.
Described fuzzy controller also comprises demonstration rate calculations device, calculates switch supported demonstration speed r (n+1) according to following formula:
R (n+1)=Sat
β * C{ r (n)+Δ er (n+1) }, wherein C is the bandwidth of link, and β is the bandwidth constraints factor, and r (n) is the supported explicit rate of switch that is calculated in [(n-1) T, nT] time interval, and Sat is a truncation funcation.
The The fuzzy PD flow controller calculates the rate parameter that feeds back to the terminal use by above-mentioned effective control method, the terminal use regulates the terminal use sends speed from data to network according to this rate parameter, thereby control (regulating/avoid) network congestion finally improves network performance and service quality.This method has the advantage of stability, fairness, robustness and anti-interference simultaneously.
By below in conjunction with the accompanying drawing description of the preferred embodiment of the present invention, other characteristics of the present invention, purpose and effect will become clear more and easy to understand.
Description of drawings
Preferred implementation of the present invention is described below with reference to the accompanying drawings, wherein:
Fig. 1 is network architecture figure;
Fig. 2 is a The fuzzy PD flow control mechanism illustraton of model;
Fig. 3 is a The fuzzy PD flow controller structure chart;
Fig. 4 is structure of fuzzy controller figure;
Fig. 5 is language value and the membership function curve chart thereof of variable E;
Fig. 6 is language value and the membership function curve chart thereof of variable EC;
Language value and the membership function curve chart thereof of Fig. 7 variable Δ ER;
Fig. 8 is the network configuration curve chart that contains single bottleneck link;
Fig. 9 is multi-hop (hop) network structure;
Figure 10 a to Figure 10 d is the dynamic characteristic figure of the queue length in the stability experiment of The fuzzy PD flow controller of the present invention;
Figure 11 a to Figure 11 d is the dynamic characteristic figure of the ACR in the stability experiment of The fuzzy PD flow controller of the present invention;
Figure 12 a is the dynamic characteristic figure of the queue length in the anti-interference test of The fuzzy PD flow controller of the present invention;
Figure 12 b is the dynamic characteristic figure of the ACR in the anti-interference test of The fuzzy PD flow controller of the present invention;
Figure 13 is 5 ABR source performance plots during the robust of The fuzzy PD flow controller of the present invention is tested;
Figure 14 a is the dynamic characteristic figure of the queue length during the robust of The fuzzy PD flow controller of the present invention is tested;
Figure 14 b is the dynamic characteristic figure of the ACR during the robust of The fuzzy PD flow controller of the present invention is tested;
Figure 15 a is the dynamic characteristic figure of the queue length during the fairness of The fuzzy PD flow controller of the present invention is tested;
Figure 15 b is the dynamic characteristic figure of the ACR during the fairness of The fuzzy PD flow controller of the present invention is tested;
In all above-mentioned accompanying drawings, identical label represents to have identical, similar or corresponding feature or function.
Embodiment
The present invention is described further below in conjunction with accompanying drawing.
The present invention has utilized fuzzy controller and two kinds of devices of PD controller, at first introduces the control principle of two kinds of controllers below:
The principle of PD (proportion differential) controller
PD controling appliance volume description is as follows: establishing r is the professional explicit rate of its supported whole ABR (Available Bit Rate) that switch is calculated at its a certain delivery outlet (corresponding with an output link), it also is the controlled quentity controlled variable of controlled object as the output of PD controller; Q represents this delivery outlet corresponding buffer region queue length, as controlled variable.Q
TBe this buffering area queue length desired value or congestion threshold, as desired value.T represents the computing cycle of r and feedback velocity parameter ER, the i.e. sampling time of controller.Therefore, the expression formula of the demonstration speed of PD controller output is:
R (n+1)=max{r (n)+K
PE (n)+K
D[e (n)-e (n-1)]/T is in the 0} formula: e (n)=Q
T-q (n) is called departure, and r (n+1) is the supported explicit rate of switch that is calculated in [nT, (n+1) T] time interval; R (n) is the supported explicit rate of switch that is calculated in [(n-1) T, nT] time interval; K
PAnd K
DBe called proportionality constant and derivative constant.[e (n)-e (n-1)]/T represents the rate of change of e (n), with Δ e (n) expression.Here, r (n+1) just can be used as feedback velocity parameter ER, and for carrying out Discrete Time Analysis, the every T of PD controller that runs on the network intermediate node calculates a new feedback velocity parameter ER value second.
The principle of fuzzy (Fuzzy) controller
Fuzzy controller does not need to set up the mathematical models of controlled process.For fuzzy controller, input and output all are accurate numerical value, the principle of fuzzy system then is to adopt people's thinking, be that language rule carries out reasoning, therefore the input data must be become the language value, promptly carry out " obfuscation ",, promptly carry out " reverse gelatinization " through the accurate amount that the result of fuzzy reasoning gained must become a reality.The basic structure of fuzzy controller as shown in Figure 4, this be one typical two the input one export structure.Accurate input variable e (n) is at n sampling instant switch output buffer queue length threshold Q among Fig. 4
T(set-point) is poor with instantaneous queue length q's (n) (measured value), i.e. e (n)=Q
T-q (n); Error rate Δ e (n) is [e (n)-e (n-1)]/T, and T is the sampling period, and e (n-1) is at n-1 sampling instant switch output buffer queue length threshold Q
T(set-point) is poor with instantaneous queue length q's (n-1) (measured value); Δ er (n+1) is the output controlled quentity controlled variable of fuzzy controller, and it is the increment of the available feedback velocity parameter of switch ER in the next sampling period; E, EC and Δ ER are above-mentioned three respectively and accurately measure (e (n), Δ e (n) and Δ er (n+1)) pairing fuzzy language variable; K
E, K
DBe respectively the quantizing factor of input variable, K
UIt is scale factor.Database among the figure is used for ambiguity in definition controller language control law and fuzzy data operation in addition.
Design an effective fuzzy controller and generally include several steps: definition can reflect the input variable of network congestion state and be used to control the method for the output variable of ABR service traffics, the domain of determining these variablees and their membership function curve, design control law storehouse, design fuzzy reasoning structure, the judgement of selection ambiguity solution.
1) definition of input and output variable
As shown in Figure 4, fuzzy controller has adopted the processing structure of typical two inputs, one output, be that fuzzy controller has used two input variables to calculate the increment of feedback velocity parameter ER, here, we select the switch output buffer as controlling object, and its instantaneous queue length is as controlled parameter.Therefore, we adopt input variable e (n) to be used for the current state of trace buffer formation; Its rate of change Δ e (n) is used to provide the prediction of following quene state.By the ER increment that provides with switch in next sampling period as the output Control Parameter, can be so that ABR (Available Bit Rate) service source end be easier to the unexpected variation of response to network load or buffering area queue length.
2) domain and membership function determines
By the actual conditions of controlled process (switch output buffer) as can be known, the excursion of queue length q (n) is [0, B], and wherein B is the size of this buffering area.Therefore, the measuring range of two input variables is respectively: e (n) ∈ [Q
T-B, Q
T], Δ e (n) ∈ [B/T ,+B/T]; According to the bandwidth C of link, the sphere of action that can determine to export control variables is Δ ER ∈ [C, C].By top, we can see, no matter are input variable or output variable, and they all are accurate amounts.Adopt fuzzy control technology just must at first convert them to the membership function of fuzzy set.Each input value can corresponding fuzzy set, and a scope continually varying value just can have unlimited a plurality of fuzzy set, and this is skimble-skamble when practical application.For the ease of realizing, usually the input variable scope is defined as discrete some levels artificially.Here, for the amount of calculation that reduces fuzzy controller and keep higher control resolution, three linguistic variables are all fuzzy be divided into five fuzzy subsets as NB, NS, ZE, PS, PB}, we select to quantize factor K
E=K
D=1.The membership function that defines top three amounts can adopt and hang bell, trapezoidal and triangle, hang in theory bell ideal, but calculation of complex; Facts have proved not have fairly obvious difference with triangle and its performance of trapezoidal function; Therefore in order to simplify calculating, we adopt the most frequently used triangle and the trapezoidal function that combines.Fig. 5--Fig. 7 has shown the pairing membership function curve of each linguistic variable fuzzy subset.Q among the figure
E=M
E* Q
T, Q
EC=M
EC* Q
T/ T, Q
Δ=M
Δ* C; Here select M
E=M
EC=0.08, M
Δ=0.5, Q
T=150cells, B=5000cells, T=1ms.
3) design of traffic control rule and fuzzy logic inference machine
According to the number of input and output variable, just can obtain the maximum number of desire rule:
N＝n_out×(n_level)
n_m
N_in is the number of input variable in the formula, is 2 herein, i.e. e (n) and Δ e (n); N_out is the number of output variable, is 1 herein, i.e. Δ er (n+1); N_level is the number that grade classification is blured in input and output; Therefore, the maximum number of this paper rule is 25.
Fuzzy reasoning is the core of whole fuzzy controller, and this reasoning process is based in the fuzzy logic that the relation of implication and inference rule carries out.According to fuzzy set and fuzzy relation theory, can be with different fuzzy reasoning methods for dissimilar fuzzy rules.In order to guarantee the stability of controller, reduce the overshoot and the oscillatory occurences of controlled parameter, we adopt the fuzzy rule of following form.As:
(1) " if queue length is little and queue length reduces soon, the supported speed of switch increases fast so.”
(2) " if queue length is moderate and queue length reduces slowly, the supported speed of switch is constant so.”
(3) " if queue length is very big and queue length does not change, the supported speed of switch reduces slow so.”
For formalization, can be expressed as:
(1)IF?E?is?PS?and?EC?is?PB?THEN?ΔER?is?PB
(2)IF?E?is?ZE?and?EC?is?PS?THEN?ΔER?is?ZE
(3)IF?E?is?NB?and?EC?is?ZE?THEN?ΔER?is?NS
Wherein " the x is a and y is b " of IF part is called former piece portion, and " the z is c " of THEN part is called consequent portion.According to buffering area characteristic and experimental result, fuzzy traffic control rule is designed to shown in the table 1.
Fuzzy controller adopts minimum-maximum rationalistic method (min-max) to carry out fuzzy reasoning.This inference method is to choose earlier the adaptive degree of the value (i.e. the degree of membership of " least adaptive ") of degree of membership minimum in each condition as this rule in the reasoning former piece, and to obtain the conclusion of this rule, this is called gets little min operation; Again each regular conclusion is comprehensively chosen the part of maximum adaptation degree, promptly got big max operation.
Table 1 fuzzy Control rule list
The selection of ambiguity solution decision method
Scale factor K is multiply by in the fuzzy output of inference machine
UAfter send into the ambiguity solution module and handle, the present invention selects K
U=0.08.Here, the control output that obtains through fuzzy reasoning is a fuzzy membership functions or fuzzy subset, and it has reflected the ambiguity of control language, and this is a kind of combination of different values.Yet to control a physical object in actual applications, can only when some, be carved with a definite controlled quentity controlled variable, this just must find out one from fuzzy output membership function to represent this fuzzy set be the accurate amount that fuzzy control effect possibility distributes, ambiguity solution judgement that Here it is.From mathematics, this is a mapping from the defined fuzzy control action space of output domain to accurate control action space.The most frequently used method is maximal criterion method, the maximum membership degree method of average and gravity model appoach at present.Wherein, gravity model appoach is a kind of widely used method.Find out in this way the membership function curve that cuts and abscissa surround the center of gravity of area, its essence is and find out the center of gravity that the control action possibility distributes.In output is that the output controlled quentity controlled variable of fuzzy controller can be tried to achieve with following formula under the situation of centrifugal pump set:
Δer(n+1)＝[∑μ(z
1)×z
1]/∑μ(z
1)
μ (z wherein
i) be each regular conclusion degree of membership, z
1Be the representative point of each rule conclusion, i.e. center of gravity abscissa.
The supported demonstration speed of switch is: r (n+1)=Sat
β * CBandwidth constraints factor-beta=1.1 in { r (n)+Δ er (n+1) } formula, it is used to limit because the long controlled parameter overshoot that causes of link range is excessive, thereby can reduce the cell loss concealment of controller; Sat is a truncation funcation.Calculating the process that shows speed r (n+1) can finish in the The fuzzy PD flow controller, also can finish in the outside of The fuzzy PD flow controller.
The PD controller above the following basis and the theory of fuzzy controller describe the concrete control method of The fuzzy PD flow controller of the present invention in detail:
Described a kind of network system figure among Fig. 1, comprised source end subscriber, destination user and n switch in this network, the The fuzzy PD flow controller among the present invention is exactly a part of forming switch, and the concrete running of The fuzzy PD flow controller is as follows:
The queue length (as q) of the buffering area (as input block or output buffer) of the The fuzzy PD flow controller in the switch (see figure 2) in a time cycle (as T second) once sampling switch is made as q (n) in the queue length of n sampling period sampling.Set a buffering area queue length desired value or congestion threshold, as desired value (Q
T).N time cycle, q (n) and Q
TIt is the input variable of The fuzzy PD flow controller.Switch computes goes out Q
T-q (n) is called departure e (n), i.e. e (n)=Q
T-q (n), and preserve e (n).This switch computes [e (n)-e (n-1)]/T then, the rate of change that is called e (n) is with Δ e (n) expression, i.e. Δ e (n)=[e (n)-e (n-1)]/T, e (n-1) is the amount of calculating in the time cycle and preserving at n-1, is called the departure of n-1 in the time cycle.Sampling buffer queue length and the calculating of e (n) and Δ e (n) carried out in the sampling of The fuzzy PD flow controller and computing unit wherein, as shown in Figure 3, the mode selector switch that the described e (n) that calculates and Δ e (n) are input to the The fuzzy PD flow controller, the mode selector switch is carried out selection operation, its size according to e (n) selects to adopt PD controller or fuzzy controller, as e (n) during greater than certain prior preset threshold (can be provided with as the case may be), select fuzzy controller, e (n), Δ e (n) are input to fuzzy controller; Otherwise, select the PD controller, e (n), Δ e (n) are input to the PD controller.
If selection fuzzy controller, the fuzzier unit of fuzzy controller is carried out Fuzzy processing respectively with e (n) and Δ e (n), draw two input fuzzy quantity E and EC, carry out fuzzy reasoning according to the rule base of formulating then, draw an output fuzzy quantity Δ ER, then, draw an accurate output valve Δ er (n+1), show that at last the rate calculations device is according to formula: r (n+1)=Sat by the ambiguity solution module
β * C{ r (n)+Δ er (n+1) } draws the output controlled quentity controlled variable r (n+1) that a n+1 will use in the time cycle, and this output controlled quentity controlled variable is used as feedback velocity parameter ER and is sent to the source end subscriber by switch.
If select the PD controller, the PD controller is according to formula:
R (n+1)=max{r (n)+K
PE (n)+K
D[e (n)-e (n-1)]/T, 0} calculates r (n+1).In having a plurality of switches and switch, have in the system of The fuzzy PD flow controller (with reference to figure 1), a minimum value computing module in the switch compares this r (n+1) and the r (n+1) that The fuzzy PD flow controller in other switch of receiving produces and sends, as the minimum computing Min that asks among Fig. 2, this computing can be to carry out in the computing module in switch, the minimum value that computing is obtained feeds back to source end subscriber (as the source end subscriber I among Fig. 2) as feedback velocity parameter ER, each source end subscriber is regulated separately data (as cell Cell according to the minimum r (n+1) of this feedback, grouping) emission rate, for the data transmission rate of how to regulate separately, all are ripe prior aries, just are not described in detail here.
Enumerate several examples of application and respective performances below, Fig. 8 is the network configuration that contains single bottleneck link, and wherein 5 ABR (Available Bit Rate) users and a synthetic VBR (variable bit-rate) user share a bottleneck link.
The link range between active end/destination system and switch sw1 or the sw2 be 10km, link rate is 149.79Mbits/s, trunk bottleneck link parameter as shown in Figure 8.S1 to S5 represents the source end system of ABR service among the figure, and D1 to D5 is its corresponding destination system.
For the ABR business, we use the source with greedy character, can make each user can make full use of its available bandwidth like this.Table 2 has been listed the used major parameter value of ABR end system among Fig. 8;
Table 2ABR service end system major parameter
The represented VBR service source of Svbr is an ON-OFF process that belongs to the Poisson random process among the figure.In the ON state time, it produces cell with the speed of 20Mbits/s; Its speed that produces cell is 0 under the OFF state.The duration of each ON or OFF state is 2.5ms, so the Mean Speed of VBR source generation cell is 10Mbits/s; The actual ON and the interval obeys index distribution of OFF state.Institute is active to begin to send cell in the moment 0.
In this system, we mainly investigate the transient state and the steady-state characteristic of network controlled parameter.We are with maximum instantaneous queue length and response time two main mapping indexs as the reflection network performance; With steady-state error (error between the stationary value of controlled parameter and set-point-switch output buffer queue length threshold) as main steady-state behaviour index.
We suppose VBR VC, and (be Svbr, Dvbr) inoperative, so available link bandwidth and active vc s number are all fixed in the network, used network model as shown in Figure 8.Figure 10 and Figure 11 have provided the dynamic characteristic of bottleneck switch (sw1) output buffer queue length and an ABR (Available Bit Rate) source end system (as S1) ACR (decay crosstalk ratio) respectively.We know that the stability of mechanism and the selection of mode switching threshold have nothing to do from Figure 10, and promptly controlled parameter (queue length q) always converges to its desired value Q
TFigure 11 shows that the same selection with the mode switching threshold of convergence of the Control Parameter ACR of mechanism has nothing to do, and always to converge to 1/5 of bottleneck link total capacity 50Mbits/s be 10Mbits/s; This ACR sum that shows whole ABR source end systems equals the bottleneck link bandwidth capacity, so this mechanism has very high bandwidth availability ratio.From Figure 10 and 11, we can see that also when stable state, all there be not " burr " in the controlled parameter and the Control Parameter of The fuzzy PD flow controller.By last, we may safely draw the conclusion: Fuzzy PD Control mechanism is stable, and its steady-state error is 0 (making full use of buffer resource), can make full use of link bandwidth; In addition, the selection of mode switching threshold is to the not influence of steady-state behaviour of controller.Therefore, the The fuzzy PD flow controller has better steady-state characteristic.
From Figure 10 and 11 as can be known: the selection of mode switching threshold is very big to the influence of controller transient response.Table 3 has been listed the influence of the selection of mode switching threshold to the relevant performance parameter of controller transient response (controlled parameter).
Table 3 mode switching threshold is to the influence of controller transient response
From table as can be known: along with the increase of mode switching threshold, will change the PD controller gradually into from fuzzy controller to the influence of controller transient response, the analysis when this designs the The fuzzy PD flow controller with us is consistent.When mode switching threshold during less than 45cells (30%), fuzzy control plays a major role; When the mode switching threshold greater than 75 (50%) time, PD control plays a major role.The mode switching threshold is when 45cells is between 75cells, and mechanism combines the speciality of two kinds of control actions, and has obtained transient response preferably: fast response time, low overshoot.
Hence one can see that, and when selecting suitable mode switching threshold, the The fuzzy PD flow controller can obtain low overshoot, thereby can improve the performance of network; Compare with the PD controller, the The fuzzy PD flow controller has the response time faster.
Describing below when the available link capacity changes, also is to have when disturbing the dynamic characteristic of bottleneck switch queue length and ABR source end ACR value in the network.Network model such as Fig. 8.Figure 12 shows that the ACR of an ABR source end system such as S1 and the dynamic characteristic of switch sw1 queue length.
By Figure 12 (b) as can be known the ACR value be near the vibration steady-state value of 8Mbps at (50-10)/5Mbps, so link bandwidth is fully utilized.Similar to ACR, the queue length shown in Figure 12 (a) is at desired value Q
TNear vibration; Because queue length is vibrated near being controlled at desired value, so the switch buffers district also is fully utilized.By on can reach a conclusion, the The fuzzy PD flow controller also has interference free performance, it can be applied to equally ABR and VBR the coexistence network environment in and can utilize Internet resources fully.
Fig. 9 is multi-hop (hop) network simulation model, and all link rates are 150Mbits/s among the figure.Used parameter value of ABR end system and table 2 parameter value are basic identical, and unique difference is its PCR=200Mbits/s.SA1, SA2, SB1, SB2, SC1 to SC6 represent the source end system of ABR service among the figure, and DA1, DA2, DB1, DB2, DC1 to DC6 are its corresponding destination systems.Here robustness is meant that it is stable that controller remains when movable ABR VCs number changes.In order to investigate robustness, we adopt ABR source characteristic shown in Figure 13 in this experiment, and purpose is that movable ABR VCs number is changed.Figure 14 is a simulation result, and as we know from the figure, the The fuzzy PD flow controller is the variation of response activity ABRVCs number rapidly.From Figure 14 (a) as can be known: queue length is that segmentation is stable and always converge to object queue length Q in finite time
TFrom Figure 14 (b) as can be known: ACR increases and reduces along with movable ABR VCs number, raises along with the minimizing of movable ABR VCs number.By on can reach a conclusion: the The fuzzy PD flow controller is a robust.
We utilize the dynamic characteristic of ACR to investigate fairness, still use network model as shown in Figure 9.Can see that from Figure 15 (a) the tight control to switch queue length makes the ACR of each source end system converge to the expectation fair share value of VC group separately, as shown in Figure 15 (b).Fairness between different VC in same group (as the VC of SA1 and SA2 correspondence) can reach equally, here because layout space is limit the ACR curve of the VC in every group of only having drawn.But as can be seen, the The fuzzy PD flow controller is fair equally.
In addition, we also can know from Figure 15, and on the same group link not, their distances are separately also made a world of difference, and the queue length of bottleneck switch all restrains.Can reach a conclusion simultaneously thus, the The fuzzy PD flow controller also is a robust for the influence of distance (or round-trip transmission time RTT).
By the foregoing description as can be known, The fuzzy PD flow controller of the present invention has better steady-state characteristic, has interference free performance, and the The fuzzy PD flow controller is fair and robust equally.Method of the present invention can be applied to the data flow control method in the computer network, also can be applied to core net, Access Network partial data flow control methods in the mobile communication system.
The above only is a preferred implementation of the present invention; should be pointed out that for those skilled in the art, under the prerequisite that does not break away from the principle of the invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.
Claims (8)
1. a fuzzy proportion differential flow control methods that is used for the congested control of communication network comprises the steps:
1) every the queue length of a time cycle sampling buffer;
2) predefined congestion threshold is deducted the controlled error e of described queue length (n);
3) described departure is deducted value that departure that last time cycle obtains obtains divided by time cycle controlled error change rate Δ e (n);
4) judge in step 2) in the departure that obtains, if this departure is during greater than certain prior preset threshold, select fuzzy controller, the rate of change of described departure and described departure is input to fuzzy controller, described fuzzy controller output switch supported demonstration speed r (n+1) is used for Controlling Source end speed; When if this departure is not more than certain prior preset threshold, the selection percentage derivative controller, the rate of change of described departure and described departure is input to proportional plus derivative controller, and described proportional plus derivative controller is exported supported demonstration speed r (n+1); Wherein n is an integer.
2. a kind of fuzzy proportion differential flow control methods that is used for the congested control of communication network according to claim 1 is characterized in that, after described fuzzy controller receives the rate of change of described departure and described departure, carries out following processing:
1) rate of change with described departure and described departure carries out Fuzzy processing, obtains corresponding fuzzy language variable;
2) according to rule base described fuzzy language variable is carried out the fuzzy language variable Δ ER that reasoning obtains the increment of feedback velocity parameter ER;
3) the fuzzy language variable Δ ER with the increment of described feedback velocity parameter ER carries out the output controlled quentity controlled variable Δ er (n+1) that ambiguity solution obtains fuzzy controller, obtains the supported demonstration speed of switch r (n+1) according to described output controlled quentity controlled variable.
3. a kind of fuzzy proportion differential flow control methods that is used for the congested control of communication network according to claim 2 is characterized in that described fuzzy controller calculates the supported demonstration speed of switch r (n+1): r (n+1)=Sat according to following formula
β * C{ r (n)+Δ er (n+1) }, wherein C is the bandwidth of link, and β is the bandwidth constraints factor, and r (n) is the supported explicit rate of switch that is calculated in [(n-1) T, nT] time interval, and Sat is a truncation funcation.
4. according to each described a kind of fuzzy proportion differential flow control methods that is used for the congested control of communication network in the claim 1 to 3, it is characterized in that, described proportional plus derivative controller calculates supported demonstration speed r (n+1) according to following formula, r (n+1)=max{r (n)+K
PE (n)+K
DΔ e (n), 0}, wherein, r (n) is the supported explicit rate of switch that is calculated in [(n-1) T, nT] time interval, K
PAnd K
DBe called proportionality constant and derivative constant.
5. a communication system comprises at least one switch, it is characterized in that, described switch comprises a fuzzy proportion differential flow controller, and described fuzzy proportion differential flow controller comprises:
Sampling and computing unit, be used for queue length every a time cycle sampling buffer, predefined congestion threshold is deducted the controlled error e of described queue length (n) and described departure is deducted value that departure that last time cycle obtains obtains divided by time cycle controlled error change rate Δ e (n), described departure and described rate of change are outputed to the mode selector switch;
The mode selector switch is judged described departure, if this departure during greater than certain prior preset threshold, is selected fuzzy controller, the rate of change of described departure and described departure is input to fuzzy controller; When if this departure is not more than certain prior preset threshold, the selection percentage derivative controller is input to proportional plus derivative controller with the rate of change of described departure and described departure;
Fuzzy controller according to the described departure of input and the output controlled quentity controlled variable Δ er (n+1) or the supported demonstration speed of the switch r (n+1) of the rate of change of described departure output fuzzy controller, is used for Controlling Source end speed;
Proportional plus derivative controller is exported switch supported demonstration speed r (n+1) according to the described departure of input and the rate of change of described departure, is used for Controlling Source end speed; Wherein n is an integer.
6. a kind of communication system according to claim 5 is characterized in that, described fuzzy controller comprises:
Fuzzier unit is carried out Fuzzy processing with the rate of change of described departure and described departure, obtains corresponding fuzzy language variable;
The inference machine unit carries out the fuzzy language variable Δ ER that reasoning obtains the increment of feedback velocity parameter ER according to rule base to described fuzzy language variable;
The ambiguity solution unit carries out the output controlled quentity controlled variable Δ er (n+1) that ambiguity solution obtains fuzzy controller with the fuzzy language variable Δ ER of the increment of described feedback velocity parameter ER.
7. a kind of communication system according to claim 6 is characterized in that, described fuzzy controller also comprises demonstration rate calculations device, calculates switch supported demonstration speed r (n+1): r (n+1)=Sat according to following formula
β * C{ r (n)+Δ er (n+1) }, wherein, r (n) is the supported explicit rate of switch that is calculated in [(n-1) T, nT] time interval, and C is the bandwidth of link, and β is the bandwidth constraints factor, and Sat is a truncation funcation.
8. according to each described a kind of communication system in the claim 5 to 7, it is characterized in that described proportional plus derivative controller calculates supported demonstration speed r (n+1) according to following formula, r (n+1)=max{r (n)+K
PE (n)+K
DΔ e (n), 0}, wherein, r (n) is the supported explicit rate of switch that is calculated in [(n-1) T, nT] time interval, K
PAnd K
DBe called proportionality constant and derivative constant.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005101279484A CN100571200C (en) | 2005-12-07 | 2005-12-07 | A kind of fuzzy PD flow control method that is used for the congested control of communication network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2005101279484A CN100571200C (en) | 2005-12-07 | 2005-12-07 | A kind of fuzzy PD flow control method that is used for the congested control of communication network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1980188A CN1980188A (en) | 2007-06-13 |
| CN100571200C true CN100571200C (en) | 2009-12-16 |
Family
ID=38131201
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2005101279484A Expired - Fee Related CN100571200C (en) | 2005-12-07 | 2005-12-07 | A kind of fuzzy PD flow control method that is used for the congested control of communication network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100571200C (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101534519B (en) * | 2008-03-14 | 2014-03-12 | 摩托罗拉移动公司 | Method for displaying package switching congestion state of wireless communication network |
| CN101635674B (en) * | 2009-08-20 | 2013-01-16 | 上海交通大学 | Adaptive congestion control method for communication network |
| CN104730927B (en) * | 2015-03-27 | 2018-02-16 | 西南石油大学 | The fuzzy PD variable structure control method of Intelligent artificial leg |
-
2005
- 2005-12-07 CN CNB2005101279484A patent/CN100571200C/en not_active Expired - Fee Related
Non-Patent Citations (8)
| Title |
|---|
| A Fuzzy Flow Control Approach for ABR Servicein ATM Networks. Zhang Xiaolin,Zhang Sabing,Wu Jieyi.Journal of Southeast University(English Edition),Vol.18 No.1. 2002 |
| A Fuzzy Flow Control Approach for ABR Servicein ATM Networks. Zhang Xiaolin,Zhang Sabing,Wu Jieyi.Journal of Southeast University(English Edition),Vol.18 No.1. 2002 * |
| ATM网络中基于模糊PD的显示速率流量控制机制的研究. 刘鑫，张少博，吴介一，张飒兵.计算机应用研究，第3期. 2004 |
| ATM网络中基于模糊PD的显示速率流量控制机制的研究. 刘鑫，张少博，吴介一，张飒兵.计算机应用研究，第3期. 2004 * |
| 一种基于PD与模糊复合控制的船舶航向变结构控制器. 姚刚，汤天浩.上海海运学院学报，第24卷第3期. 2003 |
| 一种基于PD与模糊复合控制的船舶航向变结构控制器. 姚刚，汤天浩.上海海运学院学报，第24卷第3期. 2003 * |
| 高速计算机网络中拥塞控制系统设计方法综述. 张孝林，吴介一，王宏宇，费翔.信息与控制，第30卷第1期. 2001 |
| 高速计算机网络中拥塞控制系统设计方法综述. 张孝林，吴介一，王宏宇，费翔.信息与控制，第30卷第1期. 2001 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1980188A (en) | 2007-06-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Adas et al. | On resource management and QoS guarantees for long range dependent traffic | |
| Kelly et al. | Distributed admission control | |
| Kalyanaraman et al. | The ERICA switch algorithm for ABR traffic management in ATM networks | |
| Alpcan et al. | A globally stable adaptive congestion control scheme for internet-style networks with delay | |
| CN100539546C (en) | Current control in the network equipment | |
| Di Fatta et al. | A genetic algorithm for the design of a fuzzy controller for active queue management | |
| Marbach | Pricing differentiated services networks: Bursty traffic | |
| CN106572019A (en) | Network energy-saving flow scheduling method based on mixing of time delay guaranteeing and SDN | |
| CN105515880A (en) | Token bucket traffic shaping method suitable for fusion network | |
| CN1964326A (en) | A method to monitor flow and flow monitoring equipment | |
| Wu et al. | Burst-level congestion control using hindsight optimization | |
| CN100571200C (en) | A kind of fuzzy PD flow control method that is used for the congested control of communication network | |
| Kesidis et al. | Extremal shape-controlled traffic patterns in high-speed networks | |
| Bhatnagar et al. | Optimal structured feedback policies for ABR flow control using two-timescale SPSA | |
| Qiu et al. | A predictive flow control scheme for efficient network utilization and QoS | |
| US6768744B1 (en) | Methods and apparatus for managing communication networks supporting multiple quality of service classes utilizing generalized processor sharing | |
| Kulkarni et al. | Performance analysis of a rate-based feedback control scheme | |
| Ren et al. | ABR traffic control over ATM network using fuzzy immune-PID controller | |
| Kulkarni et al. | Leaky buckets: Sizing and admission control | |
| Guan et al. | Adaptive fuzzy sliding mode active queue management algorithms | |
| Rao et al. | Stochastic control and analysis of two-node tandem communication network model with DBA and binomial bulk arrivals with phase type transmission | |
| Liu et al. | A fuzzy-logic control algorithm for active queue management in IP networks | |
| Chalfoun et al. | Mist: Microscopy image stitching tool | |
| Li et al. | A new bandwidth allocation strategy considering the network applications | |
| Sekercioglu et al. | Intelligent control techniques for efficient regulation of ABR queue length in ATM switches |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20091216 Termination date: 20201207 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |