CN101388833A - Network controlling method based on adaptive threshold mechanism - Google Patents

Abstract

The invention relates to a network control method based on a self-adaptive threshold value mechanism (Hereinafter referred to ATRED), which is an active queue management algorithm which is implemented on a router. In an ATRED, the maximum threshold value and the minimum threshold value can be adjusted on line according to the current congestion state, thereby the adaptability and the robustness of the algorithm can be increased, compared with a traditional tail drop and the early stage random detecting algorithm RED, the ATRED algorithm can carry out the strong and effective control to the queue, can enable the queue to have small vibration and be more stable, thereby the performances of the system can be improved, and more stable network service quality can be provided.

Description

Network control method based on adaptive threshold mechanism

Technical field

The present invention relates to the congested control method of a kind of computer network, particularly relate to a kind of router queue management and congestion control method.

Background technology

Congested control has become the indispensable guarantee of network service quality.The congested control of TCP is end to end adjusted the transmission rate of source end according to feedback information.But only rely on control end to end to be difficult to provide the favorable service quality assurance, must introduce congested control based on intermediate node.Congested control can detect congested generation in advance, and it is congested to take certain measure to extenuate, and the generation congestion feedback information makes the source end take measures to avoid congested deterioration.Adopt active queue management mechanism AQM as congested control device on router, wherein, earlier detection RED algorithm is a kind of AQM algorithm of recommendation at random.The average queue length of RED supervision packet on router, find congested approaching after, initiatively mark or packet discard inform that the source end reduces transmission rate randomly.RED abandons grouping in advance with probability mechanism before full queue, thereby has solved the full queue problem; And use average queue length rather than instant team leader to adjust drop probability, can absorb of short duration burst flow as much as possible.

The validity of RED has been passed through the checking of a series of practices, but because the parameter of RED lacks adaptivity, has brought many adverse effects to network performance:

1.RED performance sensitive in the setting of parameter, make to be difficult to find the parameter that adapts to dynamic environment, and the minor variations of parameter can be brought very big influence to network performance;

2.RED performance sensitive in the number of competition source/stream;

3.RED performance sensitive in the bag size;

4. when network environments such as load were suddenlyd change, the parameter that adapts to may no longer shake down originally, and bigger vibration appears in buffer queue, brings adverse effect to system.

Sensitiveness for fear of above-mentioned RED parameter is brought problem, a lot of non-RED strategies have been produced: simultaneously, also the kernel based on RED improves, these non-RED strategies have been alleviated the sensitiveness that parameter is provided with, strengthened the robustness of system, but the hysteresis that the self adaptation of its parameter is adjusted is bigger, and the control dynamics of formation is not strengthened.

Therefore, in the control of network congestion, realize that the online self adaptation adjustment of parameter threshold is that problem to be solved is arranged in the prior art.

Summary of the invention

In order to strengthen the control to formation under network environment complicated and changeable, effectively the crowded situation of Control Network the present invention is based on the RED kernel and has proposed a kind of network control method based on adaptive threshold mechanism (being called for short ATRED).

According to the above-mentioned problem that need to solve designed a kind of can online adjustment minimum threshold min _ThWith max-thresholds max _ThThe adaptive Active Queue Management Algorithm ATRED based on RED.This invention contains successively and has the following steps:

(1) parameter-definition and initialization: w _qUsed weights when calculating average queue length; p _MaxBe maximum drop probability; The count representative is the packet count of success transmission continuously, and initial value is-1; q _AvgRepresent average queue length, initial value is 0; Min _ThThe expression minimum threshold; Max _ThThe expression max-thresholds;

(2) set up the controller of an ATRED method at router;

(3) when new grouping arrives, if formation this moment is sky, then execution in step (4); Otherwise execution in step (5);

( ₄) calculating average queue length: q _Avg=(1-w _q) ^m* q _AvgAnd will be made as the zero-time q_time=time of formation the current time for sky; Wherein, the time that on behalf of empty queue, m continue, m=f (time-q_time);

(5) calculate average queue length: q _Avg=(1-w _q) q _Avg+ w _q* q;

(6) upgrade threshold value:

${\min }_{\mathrm{th}}\left(t\right)={\min }_{\mathrm{th}}(t-T)+\mathrm{\α}(q_{\max }-\frac{1}{\mathrm{\β}}{q}_{\mathrm{avg}}(t-T)),$

${\max }_{\mathrm{th}}\left(t\right){=\max }_{\mathrm{th}}(t-T)+\mathrm{\α}(q_{\max }-\frac{1}{\mathrm{\β}}{q}_{\mathrm{avg}}(t-T));$

Wherein, the smoothing factor of α; β is the buffer memory occupancy of expectation; T is the current time; T is the sampling time;

(7) by q _AvgCalculate packet loss p _aIf q _Avg≤ min _Th, change and go step (8); If q _Avg〉=max _Th, change and go step (9); Go step (10) otherwise change;

(8) packet is fallen in lines, p _a=0, count=0 forwards step (11) to;

(9) packet discard, p _a=1, count=-1 forwards step (11) to;

(10) calculate interim loss ratio: $p_b=\frac{p_{\max }(q_{\mathrm{avg}}-{\max }_{\mathrm{th}})}{{\max }_{\mathrm{th}}-{\min }_{\mathrm{th}}},$ Then $p_a=\frac{p_b}{1-\mathrm{count}\×p_b};$

(11) go to step (3), repeated execution of steps (3) is to (10), until end.

Based on the network control method of adaptive threshold mechanism and traditional tail drop with in early days at random detection algorithm RED compare adaptivity and robustness significantly strengthens, this method can be implemented powerful effectively control to formation, make formation vibration reduce, more steady, thereby improve systematic function, stable network service quality more is provided.

Description of drawings

Fig. 1 is an AQM control block diagram of the present invention;

Fig. 2 is an ATRED algorithm flow chart of the present invention;

Fig. 3 for average queue length of the present invention less than expectation during the team leader, threshold value becomes big schematic diagram (at this moment, the loss ratio curve has become dotted line to right translation by original solid line);

Fig. 4 for average queue length of the present invention greater than expectation during the team leader, the threshold value schematic diagram (at this moment, the loss ratio curve has become dotted line to left by original solid line) that diminishes;

Embodiment

The present invention proposes and a kind ofly can online adaptive adjust threshold value RED strategy based on the RED algorithm.ATRED has stronger adaptivity and robustness, can obtain littler instantaneous team leader's variance, makes instantaneous team leader more steady.Be elaborated with reference to accompanying drawing below in conjunction with embodiment, so that the technical characterictic and the advantage of the inventive method are carried out more deep annotation.

The concrete implementation step that the present invention is based on the network control method of adaptive threshold mechanism is:

(1) parameter-definition and initialization: w _qUsed weights when calculating average queue length; p _MaxBe maximum drop probability; The count representative is the packet count of success transmission continuously, and initial value is-1; q _AvgRepresent average queue length, initial value is 0; Min _ThThe expression minimum threshold; Max _ThThe expression max-thresholds;

(2) set up the controller of an ATRED method at router;

(3) when new grouping arrives, if formation this moment is sky, then execution in step (4); Otherwise execution in step (5);

(4) calculate average queue length: q _Avg=(1-w _q) ^m* q _AvgAnd will be made as the zero-time q_time=time of formation the current time for sky; Wherein, the time that on behalf of empty queue, m continue, m=f (time-q_time);

(5) calculate average queue length: q _Avg=(1-w _q) q _Avg+ w _q* q;

(6) upgrade threshold value:

${\min }_{\mathrm{th}}\left(t\right)={\min }_{\mathrm{th}}(t-T)+\mathrm{\α}(q_{\max }-\frac{1}{\mathrm{\β}}{q}_{\mathrm{avg}}(t-T)),$

${\max }_{\mathrm{th}}\left(t\right){=\max }_{\mathrm{th}}(t-T)+\mathrm{\α}(q_{\max }-\frac{1}{\mathrm{\β}}{q}_{\mathrm{avg}}(t-T));$

Wherein, the smoothing factor of α; β is the buffer memory occupancy of expectation; T is the current time; T is the sampling time;

(7) by q _AvgCalculate packet loss p _aIf q _Avg≤ min _Th, change and go step (8); If q _Avg〉=max _Th, change and go step (9); Go step (10) otherwise change;

(8) packet is fallen in lines, p _a=0, count=0 forwards step (11) to;

(9) packet discard, p _a=1, count=-1 forwards step (11) to;

(10) calculate interim loss ratio: $p_b=\frac{p_{\max }(q_{\mathrm{avg}}-{\max }_{\mathrm{th}})}{{\max }_{\mathrm{th}}-{\min }_{\mathrm{th}}},$ Then $p_a=\frac{p_b}{1-\mathrm{count}\×p_b};$

(11) go to step (3), repeated execution of steps (3) arrives (10) until end.

Nonlinear model with reference to the TCP flow control:

$\left\{\begin{array}{c}\stackrel{\·}{W\left(t\right)}=\frac{1}{R\left(t\right)}-\frac{W\left(t\right)W(t-R\left(t\right))p(t-R\left(t\right))}{2R(t-R\left(t\right))}\\ \stackrel{\·}{q\left(t\right)}=\frac{W\left(t\right)N\left(t\right)}{R\left(t\right)}-C\end{array}\right.$

Wherein, the physical significance of each parameter is: W is a TCP expectation window (unit: bag); Q is expectation team leader's (bag); R is two-way time (second); C is link capacity (bag/second); N is a load factor, i.e. the TCP linking number; P is a loss ratio.In steady operation this nonlinear model linearisation of naming a person for a particular job, and,, TCP/AQM is described as feedback control system (block diagram as shown in Figure 1) with small-signal theory in conjunction with classical control theory in conjunction with the AQM method on the router.The open-loop transfer function of system is as (1) formula, and how described the AQM controller influences q (t) by changing p (t).

Wherein,

For robustness and the adaptivity that strengthens the RED algorithm, threshold value fixing originally among the RED is reset, allow it carry out the online adjustment of self adaptation according to current average queue length, the implementation step that has obtained ATRED algorithm .ATRED controller is as follows:

Step 1): under the framework of RED algorithm, introduce expectation average queue length q _{Avg_e}, come evaluation algorithm control dynamics suitably whether, to the minimum threshold min of RED _ThWith max-thresholds max _ThCarry out online adjustment.If:

1) q _Avg=q _{Avg_e}, then the control of system is suitable, and threshold value remains unchanged;

2) q _Avg＜q _{Avg_e}, then the control of system is strong excessively, increases threshold value;

3) q _AvgQ _{Avg_e}, then a little less than the control of system, reduce threshold value.

Step 2): the average occupancy of establishing the expectation of formation buffer memory is β, and $\mathrm{\β}=\frac{q_{\mathrm{avg}\_e}}{q_{\max }};$ Correspondingly, at moment t, the average occupancy of instantaneous formation is β _a, and ${\mathrm{\β}}_a=\frac{q_{\mathrm{avg}}\left(t\right)}{q_{\max },},$ Wherein, q _MaxRepresent buffer memory capacity.Average occupancy β with formation reality _aWhether reach desired value β as foundation, the online adjustment conceptual design of threshold value is as follows:

$\stackrel{\·}{{\min }_{\mathrm{th}}\left(t\right)}=\mathrm{\α}(q_{\max }-\frac{1}{\mathrm{\β}}{q}_{\mathrm{avg}}\left(t\right))---\left(2\right)$

$\stackrel{\·}{{\max }_{\mathrm{th}}\left(t\right)}=\mathrm{\α}(q_{\max }-\frac{1}{\mathrm{\β}}{q}_{\mathrm{avg}}\left(t\right))---\left(3\right)$

Being adjusted into of threshold value:

1) q _Avg=q _{Avg_e}The time, β=β _a, promptly the occupancy of formation buffer memory is a desired value, threshold value remains unchanged: $\stackrel{\·}{{\min }_{\mathrm{th}}\left(t\right)}=0,$ $\stackrel{\·}{{\max }_{\mathrm{th}}\left(t\right)}=0;$

2) q _Avg＜q _{Avg_e}The time, β _a＜β, the occupancy of formation buffer memory is less than desired value, $\stackrel{\·}{{\min }_{\mathrm{th}}\left(t\right)}>0,$ $\stackrel{\·}{{\max }_{\mathrm{th}}\left(t\right)}>0,$ Threshold value increases.As shown in Figure 3, min _ThAnd max _ThConstant amplitude increases, the loss ratio curve by original solid black lines to right translation, as shown in phantom in FIG.;

3) q _AvgQ _{Avg_e}The time, β _aβ, the occupancy of formation buffer memory is greater than desired value, $\stackrel{\&CenterDot;}{{\min }_{\mathrm{th}}\left(t\right)<}0,$ $\stackrel{\&CenterDot;}{{\max }_{\mathrm{th}}\left(t\right)}<0,$ Threshold value reduces.As shown in Figure 4, min _ThAnd max _ThConstant amplitude reduces, the loss ratio curve by original solid black lines to left, as shown in phantom in FIG.;

Obviously, 0＜β＜1, the recommendation of β is 0.5.The β value is too small, and then control is excessive, the stability of influence control; The β value is excessive, then control too a little less than, sacrificed time delay and but can't improve throughput.

α is a smoothing factor, and the excursion that is used for controlling threshold value makes team leader's variation more steady in a less zone, and 0＜α＜0.1 is generally arranged.The recommendation of α is 0.005.The α value is too small, does not then reach the purpose that self adaptation is adjusted; The α value is excessive, and then changes of threshold is excessive, can strengthen the formation vibration on the contrary, reduces the stability of system.

Step 3): after the improvement, the ATRED controller can be described as the differential equation (4):

$\left\{\begin{array}{c}p\left(t\right)=\frac{p_{\max }(q_{\mathrm{avg}}\left(t\right)-{\min }_{\mathrm{th}}\left(t\right))}{{\max }_{\mathrm{th}}\left(t\right)-{\min }_{\mathrm{th}}\left(t\right)}\\ \stackrel{\&CenterDot;}{q_{\mathrm{avg}}\left(t\right)}=k(q_{\mathrm{nnp}}\left(t\right)-q_{\mathrm{avg}}\left(t\right))\\ \stackrel{\&CenterDot;}{{\min }_{\mathrm{th}}\left(t\right)}=\mathrm{\&alpha;}(q_{\max }-\frac{1}{\mathrm{\&beta;}}{q}_{\mathrm{avg}}\left(t\right))\\ \stackrel{\&CenterDot;}{{\max }_{\mathrm{th}}\left(t\right)}=\mathrm{\&alpha;}(q_{\max }-\frac{1}{\mathrm{\&beta;}}{q}_{\mathrm{avg}}\left(t\right))\end{array}\right.---\left(4\right)$

Wherein, q _TmpRepresent instantaneous team leader, $k=\frac{{\log }_e(1-w_q)}{\mathrm{\&delta;}}>0,$ δ is a sample frequency.

Step 4): at the balance point place system is carried out linearisation, obtain the transfer function of ATRED controller: $G_{\mathrm{qred}}\left(s\right)=\frac{{\mathrm{kp}}_{\max }(s+\frac{\mathrm{\&alpha;}}{\mathrm{\&beta;}})}{({\max }_{\mathrm{th}0}-{\min }_{\mathrm{th}0})s(s+k)}---\left(5\right)$

Wherein, max _Th0And min _Th0Represent the max-thresholds and the minimum threshold at balance point place respectively.

According to stability of a system theorem, use the ATRED controller, and satisfy R ₀＜R ⁺, N N ^-System's (as shown in Figure 2), as β (k+ η ')〉when α sets up, closed-loop stabilization.Wherein, R _oIt is the two-way time of equilibrium state;

α＝0.005；

β＝0.5；

$\mathrm{\&eta;}\&prime;=\frac{{2N}^{-}}{{\left(R^{+}\right)}^{2}C}---\left(6\right)$

$e^{-s{\mathrm{<figure-callout\; id="0"\; label="R"\; filenames="A200810029581E00202.png"\; state="\{\{state\}\}">R</figure-callout>}}_0}\stackrel{\&CenterDot;}{=}\frac{1}{1+R_{0}s}$ After substitution (6) formula, system's open-loop transfer function of (1) can be approximately:

Know by (7) formula again, when $\&ForAll;\mathrm{\&omega;}\&Element;[0,{\mathrm{\&omega;}}_g]$ Shi Keyou:

Wherein,

${\mathrm{\&omega;}}_g=0.1\frac{1}{R^{+}},$ $k=\frac{{\log }_e(1-w_q)}{\mathrm{\&delta;}}>0.$

By (8) Shi Kede:

$G_{\mathrm{atred}}\left(s\right)=\frac{\mathrm{\&mu;}(\frac{\mathrm{\&alpha;}}{\mathrm{\&beta;}}+s)}{s(k+s)(\mathrm{\&eta;}+s)}---\left(9\right)$

Wherein,

In sum, can get the system features equation is:

$s^3+(k+\mathrm{\&eta;})s^2+(\mathrm{\&mu;}+\mathrm{k\&eta;})s+\frac{\mathrm{\&mu;\&alpha;}}{\mathrm{\&beta;}}=0---\left(10\right)$

The routh table of this system is as follows:

s ³ 1 μ+kη 0

s ² k+η

s ¹

s ⁰

When system satisfies R ₀＜R ^÷The time, get by (6) formula:

0<η′<η (11)

Routh table first row are analyzed, by k 0, η〉0, μ〉0, α〉0, β〉0 know: $k+\mathrm{\&eta;}>0,\frac{\mathrm{\&mu;\&alpha;}}{\mathrm{\&beta;}}>0,$ And routh table first row are except the third line

Outward all obviously greater than zero.

If β (k+ η ')〉α, then have: $1-\frac{\mathrm{\&alpha;}}{\mathrm{\&beta;}(k+\mathrm{\&eta;}\&prime;)}>0---\left(12\right)$

Can get by following formula:

$1-\frac{\mathrm{\&alpha;}}{\mathrm{\&beta;}(k+\mathrm{\&eta;}\&prime;)}>0\&DoubleRightArrow;1-\frac{\mathrm{\&alpha;}}{\mathrm{\&beta;}(k+\mathrm{\&eta;})}>0\&DoubleRightArrow;\mathrm{\&mu;}[1-\frac{\mathrm{\&alpha;}}{\mathrm{\&beta;}(k+\mathrm{\&eta;})}]+\mathrm{k\&eta;}>0\&DoubleRightArrow;\mathrm{\&mu;}+\mathrm{k\&eta;}-\frac{\mathrm{\&mu;\&alpha;}}{\mathrm{\&beta;}(k+\mathrm{\&eta;})}>0$

As seen the third line of routh table first row is also greater than 0, and promptly first of the routh table row are all greater than zero, by Routh Criterion as can be known: system stability.

Suppose that network environment is C=4000 ^Packet/ sec ⁵, N ^-=80, R ⁺=0.25sec.The ATRED controller there is k=0.005, α=0.005, β=0.5, p _Max=0.2, q _Max=800.

Known parameters substitution (6) formula is got: η '=0.64, can release:

$\mathrm{\&beta;}(k+\mathrm{\&eta;}\&prime;)=0.3225>0.005\&DoubleRightArrow;\mathrm{\&beta;}(k+\mathrm{\&eta;}\&prime;)>\mathrm{\&alpha;}$

By the system stability theorem as seen, closed-loop system is stable.

On the NS-2 emulation platform, realized the ATRED algorithm, and it has been carried out performance test.2.31 editions NS-2 platform is used in emulation, and operating system is UBUNTU7.04.Adopt n ftp business source and m HTTP service source to simulate network load condition in the practical application in the experiment, the link capacity between all service sources and the router r1 is 1Mbps, delay time into 160ms to the random value between the 240ms; Bottleneck link is between router r1 and r2, and link capacity is 15Mbps, time-delay 40ms; Except that the linking between router r1 and the r2, the DropTail algorithm is all used in all the other links; Buffer memory on the router r1 is 800 bags (the grouping default size is 500bytes), and the grouping of each service source is set to 500bytes.

Use following a series of experiment that the control of the enhancing that ATRED has instantaneous team leader is described.The prerequisite of experiment is: the parameter of default 3 kinds of algorithms, only change load capacity, and observe 3 kinds of algorithms susceptibility to the different loads amount under the situation that preset parameter is provided with.

● the parameter setting of ATRED: α=0.005, β=0.5, p _Max=0.1, min _Th=150, max _Th=700, w _q=0.00000133;

● the parameter setting of RED: min _Th=150, max _Th=700, p _Max=0.1, w _q=0.00000133;

● the parameter setting of GENTLE_RED: min _Th=150, max _Th=700, p _Max=0.1, w _q=0.00000133.

Embodiment one

20 ftp business sources and 60 HTTP service sources are set, and all service sources are started working from 0s, and experimental period is 200s.Link between router r1 and the r2 adopts ATRED, RED and 3 kinds of methods of GENTLE_RED, carries out 3 experiments respectively.And 10 ftp business sources of every increase and 30 HTTP service sources just carry out the comparative experiments of 3 kinds of methods, until load capacity is increased to 180 ftp business sources and 540 HTTP service sources, obtain as following table 1 experimental data:

The performance index of ATRED, RED and GENTLE_RED under the table 1 different loads situation

As can be seen from the above table, in most loading condition, instantaneous formation mean value and the RED of ATRED are suitable with GENTLE_RED, and variance little than these two kinds of algorithms, perhaps the ratio of its variance increase is little with respect to the ratio that formation mean value increases, and visible ATRED is better than RED and GENTLE_RED generally to the control ability of instantaneous queue length.From the result of mean value, verified that further ATRED can adapt to most loading condition, realize instantaneous queue length is controlled preferably, obtain more stable queue length.Under most load environment, ATRED can obtain more stable service.

According to above data as can be seen, the advantage that the ATRED algorithm is represented is when load is lighter, and is comparatively obvious under the just slight congested situation.In order further to check this point, in underloaded situation, prolong with top topological model, revise parameter, can draw the experimental data of following table 2:

The performance index of ATRED, RED and GENTLE_RED under the table 2 underload loading condition

ATRED is under the less light load condition of number of nodes as can be seen from the above table, and instantaneous formation variance is starkly lower than RED and two kinds of methods of GENTLE_RED, and instantaneous queue length mean value is less, and is perhaps approaching with RED and GENTLE_RED.As seen ATRED obviously is better than RED and GENTLE_RED to the ability of instantaneous formation control.

Embodiment two

Topological structure still adopts the structure of embodiment one, 60 ftp business streams and 180 HTTP Business Streams are set, and the bottleneck bandwidth between router r1 and the router r2 between 30Mbps, is done one group experiment every 2.5Mbps at 5Mbps, totally 11 groups, experimental period is 300 seconds.Draw the mean value of 11 groups of experimental datas at last, as following table 3:

ATRED, RED under the different bottleneck bandwidths of table 3 and the performance index of GENTLE_RED

As seen, in most bandwidth situation, RED and GENTLE_RED method to instantaneous team leader's control ability all a little less than, ATRED then can both realize instantaneous team leader is better controlled in most of situation.Mean value by experimental data can verify further that ATRED goes with GENTLE_RED to instantaneous team leader's control effect than RED under most of bandwidth situation.

Embodiment three

Topological structure still adopts the structure of embodiment one, for relatively ATRED, RED and GENTLE_RED method when load capacity is undergone mutation to the control ability of instantaneous formation, available following a series of experimental verification.Topological structure still uses the structure of embodiment one, bottleneck bandwidth is set at 300 seconds for the 15Mbps experimental period, suppose in 120 seconds, the load capacity in ftp business source changes suddenly, and the load capacity of HTTP service source is constant, chooses 4 kinds of sudden change situations and experimentizes, and draws following 4 groups of experimental datas, and calculate the mean value of 4 groups of data, detailed data such as following table 4:

The performance index of ATRED, RED and GENTLE_RED under the situation of table 4 load capacity sudden change

As seen, in the load changing situation, ATRED has apparent in view advantage than RED and GENTLE_RED, and ATRED obviously is dominant to the control ability of instantaneous queue length.According to the mean value of 4 groups of mutating experiment data, can further prove: in the situation of most load capacity sudden change, ATRED is stronger to the control ability of instantaneous formation than RED and GENTLE_RED method.

The control performance of ATRED and RED and GENTLE RED compares under different network environments, the result shows: the ATRED algorithm that adaptivity strengthens is more effective to the control of formation, the oscillating phase of formation can provide more reliable and stable service quality for the user to less.

Claims (6)

1, a kind of network control method based on adaptive threshold mechanism is characterized in that comprising following implementation step:

(1) parameter-definition and initialization: w _qUsed weights when calculating average queue length; p _MaxBe maximum drop probability; The count representative is the packet count of success transmission continuously, and initial value is-1; q _AvgRepresent average queue length, initial value is 0; Min _ThThe expression minimum threshold; Max _ThThe expression max-thresholds;

(2) set up a controller at router based on the network control method of adaptive threshold mechanism;

(3) when new grouping arrives, if formation this moment is sky, then execution in step (4); Otherwise execution in step (5);

(4) calculate average queue length: q _Avg=(1-w _q) ^m* q _AvgAnd will be made as the zero-time q_time=time of formation the current time for sky; Wherein, the time that on behalf of empty queue, m continue, m=f (time-q_time);

(5) calculate average queue length: q _Avg=(1-w _q) q _Avg+ w _q* q;

(6) upgrade threshold value:

${\min }_{\mathrm{th}}\left(t\right)={\min }_{\mathrm{th}}(t-T)+\mathrm{\&alpha;}(q_{\max }-\frac{1}{\mathrm{\&beta;}}{q}_{\mathrm{avg}}(t-T)),$

${\max }_{\mathrm{th}}\left(t\right)={\max }_{\mathrm{th}}(t-T)+\mathrm{\&alpha;}(q_{\max }-\frac{1}{\mathrm{\&beta;}}{q}_{\mathrm{avg}}(t-T));$

Wherein, the smoothing factor of α; β is the buffer memory occupancy of expectation; T is the current time; T is the sampling time;

(7) by q _AvgCalculate packet loss p _aIf q _Avg≤ min _Th, change and go step (8); If q _Avg〉=max _Th, change and go step (9); Go step (10) otherwise change;

(8) packet is fallen in lines, p _a=0, count=0 forwards step (11) to;

(9) packet discard, p _a=1, count=-1 forwards step (11) to;

(10) calculate interim loss ratio: $p_b=\frac{p_{\max }(q_{\mathrm{avg}}-{\max }_{\mathrm{th}})}{{\max }_{\mathrm{th}}-{\min }_{\mathrm{th}}},$ Then $p_a=\frac{p_b}{1-\mathrm{count}\&times;p_b};$

(11) go to step (3), repeated execution of steps (3) is to (10), until end.

2, the network control method based on adaptive threshold mechanism according to claim 1 is characterized in that: the controller of described network control method based on adaptive threshold mechanism, and implementation step is:

(1) establishing the expectation average queue length is q _{Avg_e}If: and q _Avg=q _{Avg_e}, then threshold value remains unchanged; q _Avg＜q _{Avg_e}, then increase threshold value; q _AvgQ _{Avg_e}, then reduce threshold value;

(2) the average occupancy of establishing the expectation of formation buffer memory is β, and $\mathrm{\&beta;}=\frac{q_{\mathrm{avg}\_e}}{q_{\max }};$ Correspondingly, at moment t, the average occupancy of instantaneous formation is β _a, and ${\mathrm{\&beta;}}_a=\frac{q_{\mathrm{avg}}\left(t\right)}{q_{\max }},$ Wherein, q _MaxRepresent buffer memory capacity; According to $\stackrel{\&CenterDot;}{{\min }_{\mathrm{th}}\left(t\right)}=\mathrm{\&alpha;}(q_{\max }-\frac{1}{\mathrm{\&beta;}}{q}_{\mathrm{avg}}\left(t\right))$ With $\stackrel{\&CenterDot;}{{\max }_{\mathrm{th}}\left(t\right)}=\mathrm{\&alpha;}(q_{\max }-\frac{1}{\mathrm{\&beta;}}{q}_{\mathrm{avg}}\left(t\right)),$ Being adjusted into of threshold value:

Work as q _Avg=q _{Avg_e}The time, β=β _a, promptly the occupancy of formation buffer memory is a desired value, value remains unchanged: $\stackrel{\&CenterDot;}{{\min }_{\mathrm{th}}\left(t\right)}=0,$ $\stackrel{\&CenterDot;}{{\max }_{\mathrm{th}}\left(t\right)}=0;$

Work as q _Avg＜q _{Avg_e}The time, β _a＜β, the occupancy of formation buffer memory is less than desired value, $\stackrel{\&CenterDot;}{{\min }_{\mathrm{th}}\left(t\right)}>0,$ $\stackrel{\&CenterDot;}{{\max }_{\mathrm{th}}\left(t\right)}>0,$ Threshold value increases;

Work as q _AvgQ _{Avg_e}The time, β _aβ, the occupancy of formation buffer memory is greater than desired value, $\stackrel{\&CenterDot;}{{\min }_{\mathrm{th}}\left(t\right)}<0,$ $\stackrel{\&CenterDot;}{{\max }_{\mathrm{th}}\left(t\right)}<0,$ Threshold value reduces;

(3) calculate relevant parameter by the following differential equation;

$\left\{\begin{array}{c}p\left(t\right)=\frac{p_{\max }(q_{\mathrm{avg}}\left(t\right)-{\min }_{\mathrm{th}}\left(t\right))}{{\max }_{\mathrm{th}}\left(t\right)-{\min }_{\mathrm{th}}\left(t\right)}\\ \stackrel{\&CenterDot;}{q_{\mathrm{avg}}\left(t\right)}=k(q_{\mathrm{tmp}}\left(t\right)-q_{\mathrm{avg}}\left(t\right))\\ \stackrel{\&CenterDot;}{{\min }_{\mathrm{th}}\left(t\right)}=\mathrm{\&alpha;}(q_{\max }-\frac{1}{\mathrm{\&beta;}}{q}_{\mathrm{avg}}\left(t\right))\\ \stackrel{\&CenterDot;}{{\max }_{\mathrm{th}}\left(t\right)}=\mathrm{\&alpha;}(q_{\max }-\frac{1}{\mathrm{\&beta;}}{q}_{\mathrm{avg}}\left(t\right))\end{array}\right.$

Wherein, q _TmpRepresent instantaneous team leader, $k=\frac{{\log }_e(1-w_q)}{\mathrm{\&delta;}}>0,$ δ is a sample frequency;

(4) at the balance point place system is carried out linearisation, must transfer function be:

$G_{\mathrm{qred}}\left(s\right)=\frac{{\mathrm{kp}}_{\max }(s+\frac{\mathrm{\&alpha;}}{\mathrm{\&beta;}})}{({\max }_{\mathrm{th}0}-\mathrm{mi}{n}_{\mathrm{th}})s(s+k)}$

Wherein, max _Th0And min _Th0Represent the max-thresholds and the minimum threshold at balance point place respectively.

3, the network control method based on adaptive threshold mechanism according to claim 2 is characterized in that: the scope of described α is 0 to 0.1.

4, net according to claim 3 is characterized in that based on the network control method of adaptive threshold mechanism: the value of described α is 0.005.

5, according to claim 2 or 3 or 4 described network control methods based on adaptive threshold mechanism, it is characterized in that: described β chooses corresponding value according to the virtual condition of network in interval 0.2＜β＜0.7.

6, the network control method based on adaptive threshold mechanism according to claim 5, it is characterized in that: the value of described β is 0.5.

Images