CN100389580C - Network packet scheduling method applicable to wireless high-speed adaptive channels - Google Patents

Abstract

The invention relates to a network packet scheduling method suitable for wireless high-speed adaptive channels, which belongs to the technical field of networks. The steps are: (1) Initialize; (2) Determine the current queue to be served; (3) Assign the current round of sending quota to the queue, the advanced queue compensates the lagging or synchronous queue, and adjusts the leading/lagging amount of the corresponding queue; (4) ) to adjust the value of the balance counter; (5) get the head grouping of the queue, and calculate the length of the group; (6) judge the size of the packet length and the difference counter value; (7) the scheduler sends the grouping; (8) judge the current Whether the queue is empty; (9) delete the index of the current queue from the polling linked list, make the difference counter value zero, and adjust the value of the lead/lag amount of each queue, then go to step (2); (10) put the queue The index of is added to the end of the polling linked list, go to step (2). The present invention is applicable to variable packet length and low complexity.

Description

A network packet scheduling method suitable for wireless high-speed adaptive channels

technical field

The present invention relates to a method in the field of network technology, more specifically, to a network packet scheduling method suitable for wireless high-speed adaptive channels.

Background technique

Wireless network packet scheduling is derived from wired network packet scheduling. At present, there are two main types of wired network packet scheduling methods: GPS-based (general processor sharing) and RR (round-robin based) scheduling methods. The packet scheduling of the wireless network is also correspondingly developed based on the above two scheduling methods combined with the characteristics of the physical layer of the wireless network. Typical GPS-based methods include Fair Scheduling in Wireless Packet published by S.Lu et al. in "IEEE/ACM Transactions on Networking" (IEEE/ACM Network Journal) (1999, Vol. IWFQ proposed in Networks (Fair Scheduling of Wireless Packet Networks); T.S.E.Ng et al. in "INFOCOM'98, Proceedings.IEEE" (Proceedings of IEEE INFOCOM'98 Conference) (Volume 3, pp. 1103-1111, March 1998) CIF-Q proposed in the published Packet fair queuing algorithms for wireless networks with location-dependent errors (wireless network packet fair queuing algorithm based on location errors); S.Lu et al. in "Wireless Networks" (wireless network) (No. WFS mentioned in the Design and Analysis of an Algorithm for Fair Service in Error-prone Wireless Channels (Design and Analysis of a Fair Service Algorithm under Wireless Error Channels) published on Volume 4, Issue 323-343). In order to approach the characteristics of GPS scheduling, the error-free models of the above methods all adopt the method of virtual time. The disadvantage of this method is that the time and space complexity is too high, and it is not suitable for the environment of high-speed network.

Through the retrieval of prior art documents, it is found that RR-based network packet scheduling methods include WRR (Weighted Round Robin); M.Shreedhar et al. ) (Volume 4, Issue 3, Page 375-385, 1996) published Efficient Fair Queuing Using Deficit Round-Robin (Efficient Fair Queuing Using Difference Polling), the DRR method proposed in this article. The WRR and DRR methods are packet scheduling methods based on wired networks and cannot be directly applied to high-speed wireless adaptive channels. It is characterized by low complexity, but the QoS guarantee it provides is worse than the GPS-based method. Because the high-speed network requires the scheduler to use as little time as possible when processing the packet, the method based on polling has been well applied in the high-speed network.

Adaptive modulation technology makes the link rate dynamically variable so that the link rate is no longer a constant, but a random variable that changes randomly with the channel quality. It makes the wireless channel a rate adaptive channel. The channel model of the current (including the above) wireless network packet scheduling method is mainly based on the GE (Gillbert Elliott) model of the two-state Markov chain, which divides the channel into two states: channel good (G) and channel bad (B), but the channel state of the adaptive channel is multi-state, and there is still an intermediate state between the best channel and the worst channel, so the GE model can no longer be applied to the current wireless channel characteristics.

Contents of the invention

Aiming at the deficiencies of the prior art and the characteristics of the adaptive channel, the present invention proposes a network packet scheduling method suitable for the wireless high-speed adaptive channel. It discards the GE model that does not conform to the current adaptive channel characteristics, adopts a new multi-state wireless network channel model with more practical significance, each state represents a channel rate, and proposes a multi-state model that can be applied to Indefinite packet length, low complexity, long-term and short-term fairness guaranteed, good degradation performance, good isolation performance, polling-based wireless network packet scheduling method that can provide differentiated services for different mobile stations.

The present invention is achieved through the following technical solutions, comprising the following steps:

(1) Initialization, complete the initialization and assignment of all variables.

(2) Take the index of a queue from the head of the polling list to determine the current queue to be served.

(3) According to the channel state corresponding to the current queue and the lead/lag status of the queue, the queue is allocated the sending quota of the current round, the lead queue compensates the lagging or synchronous queue, and adjusts the lead/lag amount of the corresponding queue.

(4) Adjust the value of the difference counter DC _i . Add the newly allocated sending quota of this round to the difference counter, and the difference counter saves the sending quota of the queue in this round.

(5) Take the head (HOL, Head of Line) grouping of the queue, and calculate the length of this grouping.

(6) Judge the size of the grouping length and the difference counter value, if the length of the grouping is less than or equal to the value of the difference counter then turn to step (7), otherwise turn to step (10).

(7) The scheduler sends the packet.

(8) Determine whether the current queue is empty, if empty then go to step (9), if not empty then go to step (5).

(9) Delete the index of the current queue from the polling linked list, make the difference counter value zero, and adjust the value of the lead/lag amount of each queue, and then go to step (2).

(10) Add the index of the queue to the end of the polling linked list, and turn to step (2). Repeat the above steps until the polling list in the system is empty.

Among the above steps, step (3) is the key point of the present invention, and the distribution of queue sending quota and the amount of penalty compensation are all determined by this step. Step (3) will be described in detail below.

The lead/lag of the queue in step (3) is determined according to the variable lag _i . (The value of lag _i is the amount of service that queue i should receive minus the actual amount of service received). In the present invention, the leading queue is defined as: at a certain moment, when a certain queue i has received more services than its reserved quota in the optimal channel state, lag _i <0 at this time. A synchronous queue is defined as: at a certain moment, when the service obtained by queue i is equal to the service of its reserved quota under the optimal channel state, lag _i =0 at this time. The lagging queue is defined as: at a certain moment, when the service received by the queue i is less than the service reserved in the optimal channel state, lag _i >0 at this time.

In step (3), the allocation of the queue sending quota is performed dynamically. Penalize or compensate the sending quota of the queue according to the channel status of the current queue and the advanced/lag status of the queue. The principles of punishment or compensation are as follows:

① Punish the queue in the non-optimal channel to reduce the amount of data it sends in a round of service, thereby reducing the time it occupies the channel, giving the quota to the stream transmission with the optimal channel, and maximizing the system throughput.

② Determine the size of the penalty amount according to the channel state of the queue, that is, the worse the channel state, the more the penalty amount.

③ In order to improve the throughput of the system, only the queues in the optimal channel are compensated, because only when the channel is optimal, the system throughput is the largest and the channel rate is the fastest, so that the compensation can be completed in the shortest time. When no queue in the system is on the optimal channel, the compensation is abandoned.

④ Compensation shall be performed in the order of prioritizing lagging queue, synchronous queue, and leading queue.

⑤When compensating for lagging queues or leading queues, it is carried out according to the weighting method of lag _i descending order of the queues, that is, the queue with the largest lagging amount is compensated first, so that it can catch up as soon as possible: if there is no lagging queue in the system, then compensate the synchronous queue; if the above If neither queue exists, the queue with the smallest lead is compensated preferentially.

⑥In order to make the advanced queue have good degradation performance, the present invention defines compensation coefficient μ, μ∈(0, 1), so that the advanced queue does not completely give up its sending quota in the current round but only partially releases it when compensating the lagging queue, so that The queue can get the service volume of μQ _i in the leading situation. In order to prevent the advanced queue from releasing too much advanced amount and becoming a lagging queue during compensation, it is stipulated that the released advanced amount is at most min[(1-μ)Q _i , |lag _i |].

In step (9), in order to ensure the fairness of the system, the entire system guarantees ${\mathrm{\&Sigma;}}_{i=1}^N{\mathrm{lag}}_i=0.$ In order to ensure the long-term fairness of the system, the following two situations need to be considered:

①If the queue i becomes empty after receiving some services in this round, but it still belongs to the lagging condition (lag _i > 0), its lagging amount should be allocated to other lagging backlog queues in proportion according to the weight, ${\mathrm{lag}}_j={\mathrm{lag}}_j+{\mathrm{lag}}_i\frac{Q_j}{{\mathrm{\&Sigma;}}_{i\&Element;B}{Q}_i},$ B is the set of all lagging backlog queues in the system.

② If the queue i becomes empty after receiving some services, but it is still leading at this time (lag _i <0), then the leading amount should be allocated to other leading queues, ${\mathrm{lag}}_j={\mathrm{lag}}_j+{\mathrm{lag}}_i\frac{Q_j}{{\mathrm{\&Sigma;}}_{i\&Element;C}{Q}_i},$ C is the collection of all advanced backlog queues in the system.

The packet scheduling method suitable for wireless adaptive channel network proposed by the present invention is based on polling, overcomes the shortcomings of high time complexity and space complexity of the GPS-based method, and has the advantages of low complexity and easy implementation . The biggest innovation of the present invention is that it is based on the wireless self-adaptive multi-rate channel, which is more in line with the characteristics of the physical layer of the current wireless channel. Due to the adoption of the penalty/compensation mechanism in this scheduling method, the stations in the non-optimal channel are punished, the amount of data sent is reduced, and the bandwidth is given to the station with the optimal channel, thereby greatly improving the throughput of the system. The adoption of the compensation mechanism can satisfy the long-term fairness and short-term fairness of the system. A partial compensation mechanism is used in the compensation to ensure that the advanced queue has good degradation performance. When a station accesses the system, it can determine the scheduled sending quota according to its own requirements for service quality, so that the present invention can guarantee different services for different stations, and meet the different requirements for bandwidth in the current multi-service environment.

Description of drawings

FIG. 1 is a schematic structural diagram of the present invention applicable to downlink packet data scheduling in a wireless communication system.

Fig. 2 is a flow chart when the present invention is implemented.

Fig. 3 is a flow chart of allocating sending quotas for queues when the present invention is implemented.

Detailed ways

The applicable network model of the grouping scheduling method proposed by the present invention is as follows: assuming that there are N mobile terminals accessing the system through APs or base stations, the scheduling method proposed by the present invention operates in APs (access points) or base stations, The terminal creates a queue (a total of N queues), and all data sent to a specific terminal is put into the corresponding queue in a FIFO (first-in, first-out) manner. The wireless channel has at least two channel states, each channel state corresponds to a different channel rate, and the channel state is a time-varying random variable. The present invention serves each backlog (that is, there are packets waiting to be transmitted in the queue) queues in a polling manner. Before starting the service, the scheduler can accurately estimate the channel corresponding to the current backlog queue through the channel state detector/estimator state and serve the queue at the rate corresponding to the current channel state, and the service rate remains unchanged in this service.

In order to describe the content of the present invention more specifically, the symbols used in the text are shown in Table 1.

Table 1

The channel model adopted in the present invention is a multi-state adaptive channel, and the rate of the adaptive channel is a time-varying random variable. The applicable channel state transition of the present invention is as follows: the channel state corresponding to the queue during the service queue of the scheduler does not change (that is, a certain channel rate is kept constant), so as to ensure that the service rate of the scheduler is constant during the service of the queue , channel state transitions only occur at other times. There are M (1~M) channel states, and the corresponding channel transmission rates from large to small are (V ₁ , V ₂ ...V _m-1 , V _m ) where (V ₁ >V ₂ >v ₃ ...V _{m -2} >V _m-1 >V _m ). V ₁ corresponds to the optimal channel. At this time, the transmission rate of the channel is the fastest. It is the optimal situation for the system to transmit at this rate, and the maximum throughput can be obtained. V _m corresponds to the worst channel state, when the channel is the "worst" channel (V _m =0), the system cannot transmit data. (V ₂ . . . V _m-1 ) represents the channel rate of the intermediate channel state.

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 1 shows a specific embodiment of a wireless communication system to which the packet scheduling method proposed by the present invention can be applied. It shows the present invention and the relationship between various parts of the wireless communication system in the form of a simplified block diagram.

As shown in Figure 1, the scheduling method shown in the present invention is located in the access point (AP) or base station of the wireless communication system, and is used to schedule the packets transmitted from the AP or base station to the mobile terminal, so as to meet the fairness of the site, time Delay and other QoS indicators. Assuming that there are N mobile stations accessing the system, corresponding N queues are maintained for the N stations in the scheduler. The data source transmitted to the terminal first passes through the classifier. The function of the classifier is to classify the data transmitted to different mobile stations, and then put the data into corresponding queues in FIFO mode. The round-robin scheduler services non-empty backlog queues one by one in a round-robin fashion. The channel state detector/estimator uses the current and historical data sent by the wireless transmitter to judge the state of the current channel, and the channel state information is fed back to the polling scheduler to provide the scheduler with necessary channel information. When the scheduler serves the queue, it first uses the information of the channel where the current queue is located provided by the channel state detector/estimator to determine the size of the current queue data sending quota, and then takes out the packet from the head of the queue, and if the packet meets the scheduling condition, the packet will be is sent to the wireless transmitter to complete the transmission of the packet. After the scheduler finishes servicing one queue, it moves on to serving the next queue.

FIG. 2 shows a flow chart of the specific implementation of the packet scheduling method proposed by the present invention. It can be seen from the flow chart that the present invention can complete group scheduling through 10 steps.

step 1:

As shown in Figure 2, the initialization of the system is completed in step 1. Allocate corresponding queues for each mobile station accessing the system; determine the variables corresponding to each queue according to the QoS requirements of each station, specifically assigning sending quotas Q _i , Q _i ^k , and setting DC _i = 0, lag _i = 0; determine the compensation coefficient μ and the channel state function f(k) according to the system environment (here determine the function $f\left(k\right)=\frac{{\mathrm{<figure-callout\; id="1"\; label="V\; k\; V"\; filenames="C20061002407900161.png"\; state="\{\{state\}\}">V</figure-callout>}}_k}{V_{}},$ k=1, 2, ..., M, note: this is only the function defined in the present embodiment, and the definition of function f (k) can have multiple forms under different system environments); The index is put into polling list A.

Step 2:

In step 2, the scheduler starts to work, and takes out the index of a certain queue i from the head of the polling list A to determine the current queue to be served.

Step 3:

In step 3, assign sending quotas to the queues determined in step 2 and adjust the lead/lag amounts of the corresponding queues. Quota distribution is a relatively complicated part of the present invention. In order to illustrate the implementation of this step more clearly, the flow chart of quota distribution is shown in FIG. 3 , which continues from step 2 in FIG. 2 . In Fig. 3, three boxes are used to represent the quota allocation situation when the queue i is in three different channel states, that is, in the optimal channel (with the highest channel rate, steps 301 to 305), the worst channel state ( Bad channel cannot send data, steps 306 to 311) and intermediate channel status (channel rate between maximum and minimum, steps 312 to 318).

Step 301:

Judging whether the channel state corresponding to the current queue i is optimal, if yes, go to step 302, otherwise go to step 306.

Step 302:

Judging whether the current queue is in a lagging or synchronous state, that is, determining the queue state according to the value of lag _i , if lag _i > 0, it means the queue is lagging, and if lag _i = 0, it means the queue is synchronous. If i is lagging or synchronous, go to step 304, otherwise go to step 303.

Step 303:

Judging whether there is a lagging queue j in the optimal channel in the current backlog queue, if yes, go to step 305, otherwise go to step 304. Entering this step indicates that the queue i is in the leading state. If there is a lagging flow in the optimal channel in the system, it needs to be compensated, and the queue i needs to release part of the leading queue to compensate the lagging queue.

Step 304:

Allocate the sending quota Q _i for the current queue i, and then go to step 4 in Figure 2.

Step 305:

Since the queue i is in the leading state and there is a lagging queue j in the optimal channel in the system, the queue i releases the leading amount of min[(1-μ)Q _i , |lag _i |], and the values of lag _i and lag _j are given by the formula 1 Adjustment, the queue i gets the sending quota of μQ _i , which ensures the good degradation performance of i, and then go to step 4 in Figure 2.

$\left\{\begin{array}{c}{\mathrm{lag}}_i={\mathrm{lag}}_i+\min [(1-\mathrm{\&mu;})Q_i,|{\mathrm{lag}}_i\left|\right]\\ {\mathrm{lag}}_j={\mathrm{lag}}_j-\min [(1-\mathrm{\&mu;})Q_i,|{\mathrm{lag}}_i\left|\right]\end{array}\right.$ Formula 1

Step 306:

Judging whether the channel where the current queue i is located is the worst channel, if so, go to 307, otherwise go to 312. When in the worst channel, the data in this queue cannot be served, and the quota can only be allocated to other queues. Compensation is performed in the priority order of lagging, synchronizing, and leading.

Step 307:

Judging whether there is an optimal channel and delayed queue j in the system, if so, go to 310 , otherwise go to 308 . If there are multiple lagging queues, the queue j with the largest lagging amount is preferentially selected according to the size of the leading/lagging amount lag _j .

Step 308:

Determine whether there is an optimal channel and synchronized queue j in the system, if yes, go to 310, otherwise go to 309.

Step 309:

Judging whether there is an advanced queue j in the optimal channel in the system, if yes, go to 310, otherwise go to 311. If there are multiple leading queues, the queue j with a small leading amount is preferentially selected according to the leading/lagging amount lag _j (note: when the queue is leading, lag _j <0).

Step 310:

Since queue i is in the worst channel and cannot transmit data in this round, allocate the sending quota of this round to queue j, adjust the variable according to formula 2, and then go to step 4 in Figure 2.

$\left\{\begin{array}{c}{\mathrm{lag}}_i={\mathrm{lag}}_i+Q_i\\ {\mathrm{lag}}_j={\mathrm{lag}}_j-Q_i\end{array}\right.$ Formula 2

Step 311:

Entering this step indicates that there is no queue in the optimal channel in the system, so the sending quota is wasted, and no queue gets sending quota, and then go to step 4 in Figure 2.

Step 312:

Determine whether the queue status of i is lagging or synchronous, if yes go to step 314, otherwise go to step 313. Entering this step indicates that the queue i is in the middle channel state, and the channel rate has not reached the maximum value, and the queue should be punished.

Step 313:

Entering this step indicates that queue i is an advanced queue, and part of the advanced amount needs to be released. Judging whether there is a delay queue j in the optimal channel in the system, if there is, go to 305 , otherwise go to step 315 .

Step 314:

Allocate quota Q _i ^k for queue i, where $Q_i^k=Q_i\&CenterDot;f\left(k\right),$ $f\left(k\right)=\frac{{\mathrm{<figure-callout\; id="1"\; label="V\; k\; V"\; filenames="C20061002407900161.png"\; state="\{\{state\}\}">V</figure-callout>}}_k}{V_{}}.$ Then turn to step 316 .

Step 315:

Judging whether there is a synchronous or advanced queue j in the optimal channel in the system, if yes, go to step 317, otherwise go to step 311. When there is a queue in the optimal channel, the priority order of compensation is synchronization first and then advance.

Step 316:

Judging whether there are lagging, synchronous, and leading streams in the optimal channel in the system, if there is, go to step 318; otherwise, go to step 311. In this step, the selection order of the queue j is that the queue with a large lag is selected first, then the synchronized queue, and then the queue with the smallest lead.

Step 317:

Allocate quota Q _i ^k for queue i, where $Q_i^k=Q_i\&Center\; Dot;f\left(k\right),$ $f\left(k\right)=\frac{{\mathrm{<figure-callout\; id="1"\; label="V\; k\; V"\; filenames="C20061002407900161.png"\; state="\{\{state\}\}">V</figure-callout>}}_k}{V_{}}.$ Then turn to step 318 .

Step 318:

Entering this step indicates that there is a queue i in the system that has released part of the sending quota and that there is a queue j that can be compensated. Allocate (Q _i -Q ⁱ _k ) quota for queue j, and adjust the lead/lag amount according to formula 3. Then turn to step 4 in Figure 2.

$\left\{\begin{array}{c}{\mathrm{lag}}_i={\mathrm{lag}}_i+(Q_i-Q_i^k)\\ {\mathrm{lag}}_j={\mathrm{lag}}_j-(Q_i-Q_i^k)\end{array}\right.$ Formula 3

Step 4:

This step completes the adjustment of the difference counter, and the amount of data that can be sent in one service to queue i depends on the value of the difference counter DC _i . Here, not only the adjustment of the balance counter corresponding to the queue i is completed, but also the adjustment of the balance counter of the compensation queue j determined in FIG. 3 is included. The adjustment principle is to add the sending quota allocated for the queue to the balance counter, that is, DC _i =DC _i +Q' _i , DC _j =DC _j +Q' _j . There are three sources of Q′ _i and Q′ _j here, see Figure 3 for details. These three sources are: (1) Q _i or Q _i ^k allocated for the queue; (2) sending quota (Q _i -Q _i ^k ) released by other queues; (3) quota released by this queue lag _i (this Queue advances, releasing μQ _i ). Go to step 5 after this step is completed.

Step 5:

Take the header group of queue i, and calculate the length L _i ^P of the group, and then go to step 6. The calculation of the length of the packet is to compare with the balance counter of this queue to determine whether to send the packet.

Step 6:

Compare L _i ^P with the size of the difference counter DC _i , if $L_i^P\&le;{\mathrm{DC}}_i$ Then go to step 7, otherwise go to 10.

Step 7:

Send packets. Here it means that the scheduler serves the packet and hands the packet to the wireless transmitter to send on the channel.

Step 8:

Determine whether the current queue is empty, if it is empty, go to step 9, otherwise go to step 5, indicating that the scheduler can try to serve the next group in this queue.

Step 9:

Entering this step indicates that the current queue i becomes an empty queue after getting the service, and there is no longer a backlog. The next round of polling the queue is no longer necessary, so the index corresponding to the queue i is deleted from the polling list A. When the lag i of queue _i ≠ 0, in order to ensure that ${\mathrm{\&Sigma;}}_{i=1}^N{\mathrm{lag}}_i=0,$ To adjust the lead/lag value for the entire system. If lag _i > 0, adjust according to the formula, ${\mathrm{lag}}_j={\mathrm{lag}}_i+{\mathrm{lag}}_i\frac{Q_j}{{\mathrm{\&Sigma;}}_{i\&Element;B}{Q}_i},$ B is the set of all backlogged lagging queues in the system; if lag _i <0, adjust according to the following formula, ${\mathrm{lag}}_j={\mathrm{lag}}_j+{\mathrm{lag}}_i\frac{Q_j}{{\mathrm{\&Sigma;}}_{i\&Element;C}{Q}_i},$ C is the set of all backlogged advanced queues in the system, and then go to step 2.

Step 10:

Add the index corresponding to queue i to the end of the polling list. Entering this step indicates that queue i still has a backlog after exhausting its sending quota for this round, and needs to continue polling in the next round.

The embodiment of the present invention repeats the work according to the steps shown in the flowchart of FIG. 2 until the polling linked list is empty, that is, there is no data to be sent in the system.

Claims (5)

1. A network packet scheduling method applicable to wireless high-speed adaptive channel, characterized in that, comprising the following steps: (1) Initialization, completing the initialization and assignment of all variables; (2) Get the index of a queue from the head of the polling linked list to determine the current queue to be served; (3) According to the channel state corresponding to the current queue and the lead/lag status of the queue, the queue is allocated the sending quota of the current round, the lead queue compensates the lagging or synchronous queue, and adjusts the lead/lag amount of the corresponding queue; (4) Adjust the value of the balance counter, add the newly allocated sending amount of this round to the difference counter, and the difference counter saves the sending amount of the queue in this round; (5) Get the head grouping of queue, and calculate the length of this grouping; (6) judge the size of grouping length and difference counter value, if the length of grouping is less than or equal to the value of difference counter then go to step (7), otherwise go to step (10); (7) The scheduler sends the packet; (8) judge whether the current queue is empty, if empty then go to step (9), if not empty then go to step (5); (9) delete the index of current queue from the polling linked list, make the difference counter value zero, and adjust the value of each queue's lead/lag amount, then turn to step (2); (10) Add the index of the queue to the end of the polling linked list, turn to step (2), and repeat the above steps until the polling linked list in the system is empty. 2. the network grouping scheduling method that is applicable to wireless high-speed adaptive channel as claimed in claim 1, it is characterized in that, in described step (3), queue ahead represents that queue has obtained service, queue synchronization exceeding its reserved quota Indicates that the queue is served equal to its reserved quota, and queue lag indicates that the queue is served by less than its reserved quota. 3. the network grouping scheduling method that is applicable to wireless high-speed adaptive channel as claimed in claim 1, is characterized in that, in described step (3), preferentially assigns or compensates to the queue with maximum lagging amount; If there is no lagging The queue is redistributed or compensated to the synchronous queue; if the above two queues do not exist, it is allocated or compensated to the queue with the smallest lead. 4. the network packet scheduling method that is applicable to wireless high-speed adaptive channel as claimed in claim 1, it is characterized in that, in described step (3), when advanced queue compensation lags behind or synchronous queue, in order to guarantee good degradation Performance, when the advanced queue releases the quota, only part of the advanced amount is released to ensure that the advanced queue receives partial services; the advanced queue cannot release more than its advanced amount, so as not to become a lagging queue. 5. the network packet scheduling method that is applicable to wireless high-speed adaptive channel as claimed in claim 1, is characterized in that, in described step (3) and step (9), in order to guarantee the fairness of whole system, in adjustment queue Let the sum of the lead/lag amounts of all queues in the system be zero.

Images