CN106101211B - A Carrier Migration Method Based on Memory Page Rewrite Probability Prediction - Google Patents

Abstract

The invention discloses a carrier migration method based on memory page rewriting probability prediction. The method includes: receiving a carrier migration signal, entering a pre-migration stage, and collecting the latest N memory state changes of all memory pages in a virtual base station corresponding to the carrier information; enter the iterative migration stage; predict the memory page rewrite probability of all memory pages to be transferred; put the memory pages to be transferred with memory page rewrite probability exceeding the rewrite probability threshold value into the shutdown migration stage for transmission, and the memory page rewrite probability does not exceed The memory page to be transmitted with the rewriting probability threshold value is placed in the current round of iterations for transmission; it is judged whether the downtime migration condition is met. The present invention effectively reduces redundant iterative copies of virtual base station memory pages in the carrier migration process, thereby reducing the total migration time, migration downtime and transmission data volume of carrier migration, reducing the possibility of carrier migration non-convergence, and improving carrier migration migration performance.

Description

A Carrier Migration Method Based on Memory Page Rewrite Probability Prediction

technical field

The invention belongs to the technical field of mobile communication, and in particular relates to a carrier migration method based on memory page rewriting probability prediction.

Background technique

With the development of wireless access networks for mobile communications, wireless access equipment is undergoing an evolution process from traditional integrated base stations to distributed base stations and then to base station resource pools. The distributed base station separates the radio frequency unit from the base station, and puts the distributed base station and the remote antenna together to form a remote radio unit (Remote Radio Unit, RRU), while the original base station cabinet only leaves the baseband unit (Baseband Unit). , BBU). On the one hand, putting the RRU and the antenna together reduces the attenuation of the antenna feeder and can reduce the transmission power of the base station; The condition is constant temperature, no need for special air-conditioning equipment, further reducing energy consumption. The concept of base station resource pool is proposed on the basis of distributed base stations. By interconnecting BBUs within a certain range, the baseband processing capabilities of each BBU are shared to form a baseband resource pool that is allocated on demand and uniformly scheduled. Through reasonable planning, the base stations in the base station resource pool can not be in the maximum traffic state at the same time, and the carrier processing resources of the baseband resource pool can not be allocated according to the sum of all the maximum requirements, thereby reducing the operator's investment cost and network The overall energy consumption improves the overall utilization of carrier processing resources. Due to the rapid development of cloud computing technology, virtualization technology and virtual machine migration technology are gradually introduced into the base station resource pool. Combined with virtualization technology, the carrier processing resources in the base station resource pool can be abstracted into the form of virtual machines, and baseband pool processing resources are extracted as needed to form corresponding virtual base stations to process carrier baseband signals, which improves resource utilization. In addition, flexible allocation and unified scheduling of different carriers can be performed more conveniently with finer granularity. Combined with virtual machine migration technology, the virtual base station that processes carrier baseband signals can be migrated from one physical server to another physical server, thereby realizing carrier migration; in the case of insufficient carrier processing resources, carrier migration can be used to alleviate carrier processing resources In the case of insufficiency, improve the quality of communication services; through carrier migration, all carriers on a physical server are migrated to other physical servers, and maintenance operations such as maintenance or upgrade can be performed on the physical server, or it can be powered off to achieve energy saving purpose of emission reduction.

At present, most research on virtual machine migration is oriented to traditional Internet services. Telecom services are quite different from traditional Internet services, especially voice-related telecommunication services, which have high requirements on reliability and real-time performance, and are services with high QoS guarantee. At the same time, the carrier baseband signal has the characteristics of high interface rate and large data bandwidth, which causes the data inside the virtual base station to change very quickly, so the memory read and write of the virtual base station is very fast. Therefore, when the current virtual machine migration technology is applied to carrier migration, it cannot provide good migration performance. Convergence fails.

Contents of the invention

The purpose of the present invention is to provide a carrier migration method based on memory page rewriting probability prediction, aiming to solve the problems of too long shutdown migration time, too long total migration time and the total amount of transmitted data when the current virtual machine migration technology is applied to carrier migration Too big a question.

The present invention is achieved in this way, a carrier migration method based on memory page rewriting probability prediction, the carrier migration method based on memory page rewriting probability prediction includes the following steps:

Step 1, receiving the carrier migration signal, entering the pre-migration stage, and collecting the latest N memory state change information of all memory pages in the virtual base station corresponding to the carrier;

Step 2: Enter the iterative migration stage. In the first round of iteration, the latest N memory state change information of all memory pages is updated, and then all memory pages of the virtual base station are sent to the destination, and then go to step 5. If it is not the first round of iteration, then Update the latest N memory state change information of all memory pages, and the memory page information to be transmitted, and turn to step 3;

Step 3, predicting the memory page rewriting probability of all memory pages to be transmitted;

Step 4: Put the memory pages to be transmitted with the memory page rewrite probability exceeding the rewrite probability threshold value into the shutdown migration stage for transmission, and put the memory pages to be transmitted with the memory page rewrite probability not exceeding the rewrite probability threshold value into the current round of iterations transmission;

Step 5, judging whether the downtime relocation condition is met, if not, go to step 2, if satisfied, enter the downtime relocation stage, and complete the carrier relocation.

Further, the collection of the latest N memory page state change information of all memory pages in the virtual base station of the carrier specifically includes:

In the first step, a carrier migration signal is received, and a memory page status information table state_table with the number of rows N and the number of columns of the virtual base station memory pages M is generated;

The second step is to use the memory page state collection cycle T _c to count the rewritten memory page information. Starting from the first row of state_table, according to the statistical rewritten memory page information, the memory page h is rewritten, then the first row in the state_table , the number in column h is set to 1, and the memory page h has not been rewritten, then the number in the first row and column h in the state_table is set to 0; then according to the information of the rewritten memory page in the next statistics, all memory pages The change state information of the page is stored in the second row of state_table, and it is carried out sequentially until the Nth row is stored in the change state information of all memory pages in the virtual base station, and the change state information of the memory page is collected.

Further, the update of the latest N memory state change information of all memory pages specifically includes:

The first step is to update and replace the data in row 1 with the data in row 2 of state_table, update and replace the data in row 2 with the data in row 3 of state_table, and so on until the data in row N is updated and replaced Row N-1 data;

The second step is the first round of iteration, update all the numbers in the Nth row of state_table to 1; if it is not the first round of iteration, count the information of the rewritten memory page during the previous round of iterative migration, and then update the first row of state_table N lines, the memory page h is rewritten, then the number of the Nth row and the hth column in the state_table is set to 1, and the memory page h is not rewritten, then the number of the Nth row and the hth column of the state_table is set to 0.

Further, the update of the information of the memory page to be transmitted, the memory page to be transmitted includes:

(1) Memory pages rewritten during the previous round of iterative migration;

(2) Put the memory page for transmission in the downtime migration stage.

Further, the memory page rewriting probability prediction for all memory pages to be transmitted specifically includes:

(1) Calculate the time memory page rewriting probability P _it of the memory page i to be transmitted according to the latest N memory state change information of the memory page i to be transmitted;

(2) According to the latest memory state change information of the adjacent memory page of the memory page i to be transmitted, calculate the space memory page rewriting probability P _is of the memory page i to be transmitted;

(3) According to P _it and P _is , calculate the memory page rewriting probability P _i of the memory page i to be transmitted:

P _i = ω _t P _it + ω _s P _is ;

Wherein ω _t is the time memory page rewrite probability weight value, ω _s is the space memory page rewrite probability weight value, ω _s +ω _t =1;

(4) Traverse all memory pages to be transmitted, and calculate memory page rewriting probabilities of all memory pages to be transmitted.

Further, the calculation of the time memory page rewriting probability P _it of the memory page i to be transmitted according to the latest N memory state change information of the memory page i to be transmitted specifically includes:

(1) Obtain the latest N memory state change information of the memory page i to be transferred from the state_table, and form the memory state change information vector A _i =(a _1i , a _2i ,...,a _Ni ) of the memory page i to be transferred ^T , A _i is equal to the i-th column of state_table;

(2) A _i ≠(1,1,...,1) ^T _1×N and A _i ≠(0,0,...,0) ^T _{1×N ,} calculate the Weight value, the weight value of the prediction step size k is:

in is the average value of the memory state change information vector of the memory page i to be transferred, j is the maximum prediction step size;

(3) Normalize the weight values of different prediction steps to obtain the weight coefficients of different prediction steps. The weight coefficient of the prediction step k is:

(4) A _i ＝(1,1,...,1) _1×N or A _i ＝(0,0,...,0) _1×N , then directly calculate the weight coefficients of different prediction steps , the weight coefficient of the prediction step size k is:

(5) Calculate the transition probability matrix of different prediction steps through A _i , and the transition probability matrix of prediction step k is:

in m=0 or 1, n=0 or 1,

(6) Calculate the temporal memory page state prediction vector P _itv through the weight coefficients of different prediction steps, the transition probability matrix of different prediction steps and the latest j memory state change information of the memory page i to be transmitted:

(7) P _it =P _itv (2), P _itv (2) represents the second element of P _itv .

Further, the calculation of the space memory page rewriting probability _Pis of the memory page i to be transmitted according to the latest memory state change information of the adjacent memory page of the memory page i to be transmitted specifically includes:

(1) Obtain the latest memory state change of the adjacent memory page i-2, adjacent memory page i-1, adjacent memory page i+1, and adjacent memory page i+2 of the memory page i to be transferred from the state_table Information a _N(i-2) , a _N(i-1) , a _N(i+1) , a _N(i+2) , a _N(i-2) is equal to row N of state_table, row i- 2 columns, a _N(i-1) is equal to row N and column i-1 of state_table, a _N(i+1) is equal to row N and column i+1 of state_table, a _N(i+2) Equal to row N and column i+2 of state_table;

(2) According to a _N(i-2) , a _N(i-1) , a _N(i+1) , a _N(i+2) , calculate P _is :

P _is ＝α _i-2 a _N(i-2) +α _i-1 a _N(i-1) +α _i+1 a _N(i+1) +α _i+2 a _N(i+2) ;

Among them, α _i-2 is the weight coefficient of memory page i-2, α _i-1 is the weight coefficient of memory page i-1, α _i+1 is the weight coefficient of memory page i+1, and α _i+2 is the weight coefficient of memory page The weight coefficient of i+2, α _i-1 ＝α _i+1 , α _i-2 ＝α _i+2 , α _i-1 ＝2α _i-2 , α _i-2 +α _i-1 +α _{i+ 1} +α _i+2 = 1;

(3) For the situation that the memory page i to be transmitted is an edge memory page, the calculation parameters of P _is are processed as follows:

When i=1, a _N(i-2) =0, a _N(i-1) =0, α _i-2 =0, α _i-1 =0, α _i+1 =2α _i+2 , α _i+1 +α _i+2 =1;

When i=2, a _N(i-2) ＝0, α _i-2 ＝0, α _i-1 ＝α _i+1 , α _i+1 ＝2α _i+2 , α _i-1 +α _i+1 +α _i+2 = 1;

When i=M-1, a _N(i+2) ＝0, α _i+2 ＝0, α _i-1 ＝α _i+1 , α _i-1 ＝2α _i-2 , α _i-1 +α _i+1 +α _i+2 = 1;

When i=M, a _N(i+2) =0, a _N(i+1) =0, α _i+2 =0, α _i+1 =0, α _i-1 =2α _i-2 , α _i-1 +α _i-2 =1.

Further, the shutdown migration conditions include:

(1) The number of iterative migrations exceeds 30;

(2) The sum of the number of memory pages transferred during the shutdown migration phase and the number of memory pages rewritten during the current round of iterative migration does not exceed 50 pages.

Further, the step of entering the downtime relocation stage and completing the carrier relocation specifically includes:

(1) close the source virtual base station;

(2) Transfer the memory pages rewritten in the last round of iterative migration process and the memory pages transferred in the downtime phase to the destination virtual base station, and transfer the CPU status to the destination virtual base station at the same time;

(3) The destination virtual base station is started, and the carrier migration is completed.

The carrier migration method based on the memory page rewriting probability prediction provided by the present invention optimizes the redundant iterative copy of the memory page of the virtual base station during the carrier migration process, and can reduce the cost by at most about 90% compared with the traditional pre-copy iterative carrier migration method. The total migration time and the amount of transmitted data can reduce the downtime migration time by about 80% at most, and improve the migration performance of carrier migration.

Description of drawings

FIG. 1 is a flowchart of a carrier migration method based on memory page rewrite probability prediction provided by an embodiment of the present invention.

FIG. 2 is a flow chart of calculating the rewriting probability of a temporal memory page provided by an embodiment of the present invention.

Fig. 3 is a flow chart of calculating the rewriting probability of a space memory page provided by an embodiment of the present invention.

Fig. 4 is a schematic diagram of the total migration time of the carrier migration method based on memory page rewriting probability prediction provided by the embodiment of the present invention and the traditional pre-copy iterative carrier migration method under different user service numbers.

Fig. 5 is a schematic diagram of the downtime migration time of the carrier migration method based on memory page rewriting probability prediction provided by the embodiment of the present invention and the traditional pre-copy iterative carrier migration method under different user service numbers.

Detailed ways

In order to make the object, technical solution and advantages of the present invention more clear, the present invention will be further described in detail below in conjunction with the examples. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

The application principle of the present invention will be described in detail below in conjunction with the accompanying drawings.

As shown in Figure 1, the carrier migration method based on memory page rewriting probability prediction according to the embodiment of the present invention includes the following steps:

S101: Receive the carrier migration signal, enter the pre-migration stage, and collect the latest N memory state change information of all memory pages in the virtual base station corresponding to the carrier;

S102: Enter the iterative migration stage. If it is the first round of iteration, update the latest N memory state change information of all memory pages, and then send all the memory pages of the virtual base station to the destination, and go to step S105. If it is not the first round of iteration , update the latest N memory state change information of all memory pages, and the memory page information to be transmitted, and turn to step S103;

S103: Predict memory page rewriting probabilities of all memory pages to be transmitted;

S104: Put the memory pages to be transmitted with the memory page rewrite probability exceeding the rewrite probability threshold value in the shutdown migration stage for transmission, and put the memory pages to be transmitted with the memory page rewrite probability not exceeding the rewrite probability threshold value into the current round of iteration for transmission ;

S105: Judging whether the shutdown migration condition is met, if not, proceed to step S102, if yes, enter the shutdown migration stage, and complete the carrier migration.

In step S101, the latest N memory page state change information of all memory pages in the virtual base station of the carrier is collected, and the specific steps are as follows:

(1) Receive the carrier migration signal, generate a memory page state information table state_table with the number of rows as N and the number of columns as the number of memory pages of the virtual base station memory of the carrier;

(2) Use the memory page state collection cycle T _c to count the rewritten memory page information, starting from the first line of state_table, according to the statistics of the rewritten memory page information, if the memory page h is rewritten, the first row in the state_table The number of row and column h is set to 1. If the memory page h has not been rewritten, the number of the first row and column h in the state_table is set to 0, and then according to the information of the rewritten memory page in the next statistics, Store the change state information of all memory pages into the second row of state_table, and proceed sequentially until the Nth row is stored in the change state information of all memory pages in the virtual base station, and the memory page change state information is collected;

In step S102, the latest N memory state change information of all memory pages is updated, specifically performed according to the following steps:

(1) Use the data in row 2 of state_table to update and replace the data in row 1, use the data in row 3 of state_table to update and replace the data in row 2, and so on until the data in row N is updated and replaced N-1 rows of data;

(2) If it is the first round of iteration, update all the numbers in row N of state_table to 1. If it is not the first round of iteration, count the information of the rewritten memory page during the previous round of iterative migration, and then update accordingly In line N of state_table, if the memory page h is rewritten, set the number in line N and column h in state_table to 1; if the memory page h has not been rewritten, set the number in line N and column h in state_table The number is set to 0.

In step S102, the information of the memory page to be transmitted is updated, and the memory page to be transmitted includes:

(1) Memory pages rewritten during the previous round of iterative migration;

(2) Put the memory page for transmission in the downtime migration stage.

In step S103, memory page rewriting probability prediction is performed on all memory pages to be transmitted, specifically according to the following steps:

(1) Calculate the time memory page rewriting probability P _it of the memory page i to be transmitted according to the latest N memory state change information of the memory page i to be transmitted;

(2) According to the latest memory state change information of the adjacent memory page of the memory page i to be transmitted, calculate the space memory page rewriting probability P _is of the memory page i to be transmitted;

(3) According to P _it and P _is , calculate the memory page rewriting probability P _i of the memory page i to be transmitted:

P _i = ω _t P _it + ω _s P _is ;

Wherein ω _t is the time memory page rewrite probability weight value, ω _s is the space memory page rewrite probability weight value, ω _s +ω _t =1;

(4) Traverse all memory pages to be transmitted, and calculate memory page rewriting probabilities of all memory pages to be transmitted.

As shown in Figure 2, according to the latest N memory state change information of the memory page i to be transmitted, the time memory page rewriting probability P _it of the memory page i to be transmitted is calculated, and the specific steps are as follows:

(1) Obtain the latest N memory state change information of the memory page i to be transferred from the state_table, and form the memory state change information vector A _i =(a _1i , a _2i ,...,a _Ni ) of the memory page i to be transferred ^T , A _i is equal to the i-th column of state_table;

(2) If A _i ≠(1,1,...,1) ^T _1×N and A _i ≠(0,0,...,0) ^T _1×N , calculate different prediction steps by A _i The weight value of , the weight value of the prediction step size k is:

in is the average value of the memory state change information vector of the memory page i to be transferred, j is the maximum prediction step size;

(3) Normalize the weight values of different prediction steps to obtain the weight coefficients of different prediction steps. The weight coefficient of the prediction step k is:

(4) If A _i ＝(1,1,...,1) _1×N or A _i ＝(0,0,...,0) _1×N , then directly calculate the weights of different prediction steps Coefficient, the weight coefficient of the prediction step size k is:

(5) Calculate the transition probability matrix of different prediction steps through A _i , and the transition probability matrix of prediction step k is:

in m=0 or 1, n=0 or 1,

(6) Calculate the temporal memory page state prediction vector P _itv through the weight coefficients of different prediction steps, the transition probability matrix of different prediction steps and the latest j memory state change information of the memory page i to be transmitted:

(7) P _it =P _itv (2), P _itv (2) represents the second element of P _itv .

As shown in Figure 3, according to the latest memory state change information of the adjacent memory pages of the memory page i to be transmitted, the space memory page rewriting probability P _is of the memory page i to be transmitted is calculated, and the specific steps are as follows:

(1) Obtain the latest memory state change of the adjacent memory page i-2, adjacent memory page i-1, adjacent memory page i+1, and adjacent memory page i+2 of the memory page i to be transferred from the state_table Information a _N(i-2) , a _N(i-1) , a _N(i+1) , a _N(i+2) , a _N(i-2) is equal to row N of state_table, row i- 2 columns, a _N(i-1) is equal to row N and column i-1 of state_table, a _N(i+1) is equal to row N and column i+1 of state_table, a _N(i+2) Equal to row N and column i+2 of state_table;

(2) According to a _N(i-2) , a _N(i-1) , a _N(i+1) , a _N(i+2) , calculate P _is :

P _is ＝α _i-2 a _N(i-2) +α _i-1 a _N(i-1) +α _i+1 a _N(i+1) +α _i+2 a _N(i+2) ;

Among them, α _i-2 is the weight coefficient of memory page i-2, α _i-1 is the weight coefficient of memory page i-1, α _i+1 is the weight coefficient of memory page i+1, and α _i+2 is the weight coefficient of memory page The weight coefficient of i+2, α _i-1 ＝α _i+1 , α _i-2 ＝α _i+2 , α _i-1 ＝2α _i-2 , α _i-2 +α _i-1 +α _{i+ 1} +α _i+2 = 1;

(3) For the situation that the memory page i to be transmitted is an edge memory page, the calculation parameters of P _is are processed as follows:

When i=1, a _N(i-2) =0, a _N(i-1) =0, α _i-2 =0, α _i-1 =0, α _i+1 =2α _i+2 , α _i+1 +α _i+2 =1;

When i=2, a _N(i-2) ＝0, α _i-2 ＝0, α _i-1 ＝α _i+1 , α _i+1 ＝2α _i+2 , α _i-1 +α _i+1 +α _i+2 = 1;

When i=M-1, a _N(i+2) ＝0, α _i+2 ＝0, α _i-1 ＝α _i+1 , α _i-1 ＝2α _i-2 , α _i-1 +α _i+1 +α _i+2 = 1;

When i=M, a _N(i+2) =0, a _N(i+1) =0, α _i+2 =0, α _i+1 =0, α _i-1 =2α _i-2 , α _i-1 +α _i-2 =1.

The shutdown migration condition in step S105 is specifically: one of the following is met:

(1) The number of iterative migrations exceeds 30 times.

(2) The sum of the number of memory pages transferred during the shutdown migration phase and the number of memory pages rewritten during the current round of iterative migration does not exceed 50 pages.

Enter the downtime migration stage and complete the carrier migration. Specifically, follow the steps below:

(1) Close the source virtual base station.

(2) Transfer the rewritten memory pages in the last round of iterative migration process and the memory pages transferred in the downtime phase to the destination virtual base station, and transfer the CPU status to the destination virtual base station at the same time.

(3) The destination virtual base station is started, and the carrier migration is completed.

The application effect of the present invention will be described in detail below in conjunction with simulation.

In order to test the migration performance of the present invention, the parameters are set as follows: the virtual base station memory size of the carrier is 2048MB; the migration bandwidth is 1000Mb/s; the acquisition cycle is 2s; the rewriting probability threshold is 0.7; N=20; the maximum prediction step size is 4. The time memory page rewrite probability weight value is 0.5; the space memory page rewrite probability weight value is 0.5. The traditional pre-copy and iterative carrier migration method was selected as a comparison method, and 30 Monte Carlo experiment simulations were performed to obtain the total migration time of carrier migration as shown in Figure 4 and the downtime migration time of carrier migration as shown in Figure 5.

It can be seen from FIG. 4 and FIG. 5 that the present invention can reduce the total migration time and transmission data volume by about 90% at most, reduce the downtime migration time by about 80% at most, and improve the migration performance of carrier migration.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. within range.

Claims (1)

1. A carrier migration method based on memory page rewriting probability prediction, characterized in that, the carrier migration method based on memory page rewriting probability prediction comprises the following steps: Step 1, receiving the carrier migration signal, entering the pre-migration stage, and collecting the latest N memory state change information of all memory pages in the virtual base station corresponding to the carrier; Step 2: Enter the iterative migration stage. If it is the first round of iteration, update the latest N memory state change information of all memory pages, and then send all the memory pages of the virtual base station to the destination, and go to step 5. If it is not the first round Iterate, update the latest N memory state change information of all memory pages, and the memory page information to be transmitted, and turn to step 3; Step 3, predicting the memory page rewriting probability of all memory pages to be transmitted; Step 4: Put the memory pages to be transmitted with the memory page rewrite probability exceeding the rewrite probability threshold value into the shutdown migration stage for transmission, and put the memory pages to be transmitted with the memory page rewrite probability not exceeding the rewrite probability threshold value into the current round of iterations transmission; Step 5, judge whether the downtime migration condition is met, if not, go to step 2, if satisfied, enter the downtime migration stage, and complete the carrier migration; Predicting the memory page rewriting probability for all memory pages to be transferred specifically includes the following steps: The first step is to calculate the time memory page rewriting probability P _it of the memory page i to be transferred according to the latest N memory state change information of the memory page i to be transferred; The second step is to calculate the space memory page rewriting probability P _is of the memory page i to be transmitted according to the latest memory state change information of the adjacent memory page of the memory page i to be transmitted; The third step is to calculate the memory page rewriting probability P _i of the memory page i to be transferred according to P _it and P _is : P _i = ω _t P _it + ω _s P _is ; Wherein ω _t is the time memory page rewrite probability weight value, ω _s is the space memory page rewrite probability weight value, ω _s +ω _t =1; The fourth step is to traverse all memory pages to be transmitted, and calculate the memory page rewriting probability of all memory pages to be transmitted; The first step is specifically carried out in the following steps: (1) Obtain the latest N memory state change information of the memory page i to be transmitted from the memory page state information table state_table, and form the memory state change information vector A _i of the memory page i to be transmitted = (a _1i , a _2i , .. ., a _Ni ) ^T , A _i is equal to the ith column of state_table; (2) If A _i ≠(1, 1,...,1) ^T _1×N and A _i ≠(0, 0,...,0) ^T _1×N , calculate different prediction steps by A _i The weight value of , the weight value of the prediction step size k is: in is the average value of the memory state change information vector of the memory page i to be transferred, j is the maximum prediction step size; (3) Normalize the weight values of different prediction steps to obtain the weight coefficients of different prediction steps. The weight coefficient of the prediction step k is: (4) If A _i = (1, 1, ..., 1) _1×N or A _i = (0, 0, ..., 0) _1×N , then directly calculate the weights of different prediction steps Coefficient, the weight coefficient of the prediction step size k is: (5) Calculate the transition probability matrix of different prediction steps through A _i , and the transition probability matrix of prediction step k is: in m=0 or 1, n=0 or 1, Equal to the number of times a _ji is m and a _(j+k)i is n in A _i , 1≤j≤Nk; (6) Calculate the temporal memory page state prediction vector P _itv through the weight coefficients of different prediction steps, the transition probability matrix of different prediction steps and the latest j memory state change information of the memory page i to be transmitted: (7) P _it =P _itv (2), P _itv (2) represents the second element of P _itv ; According to the latest memory state change information of the adjacent memory page of the memory page i to be transmitted, the space memory page rewriting probability Pis of the memory page i to be transmitted _is calculated, specifically according to the following steps: (1) Obtain the latest memory state change of the adjacent memory page i-2, adjacent memory page i-1, adjacent memory page i+1, and adjacent memory page i+2 of the memory page i to be transferred from the state_table Information a _N(i-2) , a _N(i-1) , a _N(i+1) , a _N(i+2) , a _N(i-2) is equal to row N of state_table, row i- 2 columns, a _N(i-1) is equal to row N and column i-1 of state_table, a _N(i+1) is equal to row N and column i+1 of state_table, a _N(i+2) Equal to row N and column i+2 of state_table; (2) According to a _N(i-2) , a _N(i-1) , a _N(i+1) , a _N(i+2) , calculate P _is : P _is ＝α _i-2 a _N(i-2) +α _i-1 a _N(i-1) +α _i+1 a _N(i+1) +α _i+2 a _N(i+2) ; Among them, α _i-2 is the weight coefficient of memory page i-2, α _i-1 is the weight coefficient of memory page i-1, α _i+1 is the weight coefficient of memory page i+1, and α _i+2 is the weight coefficient of memory page The weight coefficient of i+2, α _i-1 ＝α _i+1 , α _i-2 ＝α _i+2 , α _i-1 ＝2α _i-2 , α _i-2 +α _i-1 +α _{i+ 1} +α _i+2 = 1; (3) For the situation that the memory page i to be transmitted _is an edge memory page, the calculation parameters of Pis are processed as follows: M is the number of virtual base station memory pages of the carrier; When i=1, a _N(i-2) =0, a _N(i-1) =0, α _i-2 =0, α _i-1 =0, α _i+1 =2α _i+2 , α _i+1 +α _i+2 = 1; When i=2, a _N(i-2) ＝0, α _i-2 ＝0, α _i-1 ＝α _i+1 , α _i+1 ＝2α _i+2 , α _i-1 +α _{i +1} +α _i+2 = 1; When i=M-1, a _N(i+2) ＝0, α _i+2 ＝0, α _i-1 ＝α _i+1 , α _i-1 ＝2α _i-2 , α _i-1 + α _i+1 +α _i+2 = 1; When i=M, a _N(i+2) =0, a _N(i+1) =0, α _i+2 =0, α _i+1 =0, α _i-1 =2α _i2 , α _{i -1} +α _i-2 =1.