CN106101211A - A kind of carrier wave emigration method rewriting probabilistic forecasting based on page - Google Patents

Abstract

The invention discloses a kind of carrier wave emigration method rewriting probabilistic forecasting based on page, described method includes: receive carrier wave emigration signal, enters the premigrates stage, collects the nearest n times internal storage state change information of all pages in the virtual base station that carrier wave is corresponding；Enter the iterative migration stage；The page of prediction institute available for transmission page rewrites probability；Page is rewritten probability exceed and rewrite the page to be transmitted of probability threshold value and be put into shutdown migration phase and be transmitted, page is rewritten probability and is put into epicycle iteration is transmitted less than the page to be transmitted rewriting probability threshold value；Judge whether to meet and shut down transition condition.The present invention effectively reduces the redundancy iteration copy of virtual base station page during carrier wave emigration; thus decrease the gross migration time of carrier wave emigration, migrate downtime and transmitted data amount; reduce the probability that carrier wave emigration is not restrained, improve the migration performance of carrier wave emigration.

Description

A kind of carrier wave emigration method rewriting probabilistic forecasting based on page

Technical field

The invention belongs to mobile communication technology field, particularly relate to a kind of carrier wave based on page rewriting probabilistic forecasting and move Shifting method.

Background technology

Along with the development of mobile communication wireless access network, radio reception device just experiencing from traditional integrated base station to Distributed base station arrives the evolution process in base station resource pond again.Distributed base station, by being separated from base station by radio frequency unit, is distributed The antenna of formula base station and far-end is put together, becomes remote radio unit (RRU) (Remote Radio Unit, RRU), and original base Rack of standing leaves behind Base Band Unit (Baseband Unit, BBU).On the one hand, RRU puts together with antenna, reduces antenna feeder Decay, the transmitting power of base station can be reduced；On the other hand, peeling off the BBU rack volume after RRU can significantly reduce, And the RRU being placed in terrace relies on natural conditions constant temperature, it is no longer necessary to special air-conditioning equipment, reduce further energy consumption.Base The concept of resource pool of standing proposes, by being interconnected by a range of BBU, by each BBU on the basis of distributed base station Base band processing capacity share, formed distribution according to need, the base band resource pool of United Dispatching.By reasonably planning, so that Base station in base station resource pond is not in traffic peak state in the same time, and the carrier processing resource of base band resource pool just may be used Not to be equipped with by all of greatest requirements summation, thus reduce the cost of investment of operator and the energy consumption that network is overall, carry The high overall utilization rate of carrier processing resource.Owing to the high speed development of cloud computing technology, Intel Virtualization Technology and virtual machine move Shifting technology is progressively introduced in base station resource pond.Combined with virtual technology, the carrier processing resource in base station resource pond is permissible It is abstracted into the form of virtual machine, and on-demand extraction baseband pool processes resource, constitute corresponding virtual base station to process carrier wave Baseband signal, improves the utilization rate of resource, and the elasticity that can more easily carry out different carrier with finer and smoother granularity is divided Join and United Dispatching.Combined with virtual machine migrating technology, can take the virtual base station processing baseband signals from a physics Business device moves to another physical server, thus realizes carrier wave emigration；In the case of carrier processing inadequate resource, by carrying Ripple migrates, and alleviates the situation of carrier processing inadequate resource, improves communication service quality；By carrier wave emigration, a certain physics is taken On business device, all carrier wave emigrations are on other physical servers, and just can overhaul this physical server or upgrading etc. is safeguarded and grasped Make, or it is powered-down to reach the purpose of energy-saving and emission-reduction.

At present, for the research of virtual machine (vm) migration mostly towards conventional internet business.Telecommunication service is mutual with tradition Networking service has the telecommunication service that bigger diversity, especially voice are relevant, and the requirement to reliability and real-time is the highest, is A kind of business of high QoS guarantee.Meanwhile, baseband signals has the characteristic that interface rate is high, data band is roomy, causes virtual The data variation speed of inside of base station is quite fast, and therefore the memory read-write of virtual base station is the fastest.So current virtual machine moves When shifting technology is applied to carrier wave emigration, good migration performance can not be provided, mainly show as shutting down migration time oversize, total Oversize and the total transmitted data amount of migration time is too big, even there will be carrier wave emigration and does not restrains and the situation of failure.

Summary of the invention

It is an object of the invention to provide a kind of carrier wave emigration method rewriting probabilistic forecasting based on page, it is intended to solve Migration time is oversize, the gross migration time is oversize and always passes to there is shutdown when current virtual machine migration technology is applied to carrier wave emigration The problem that transmission of data amount is the biggest.

The present invention be achieved in that a kind of based on page rewrite probabilistic forecasting carrier wave emigration method, described based on Page is rewritten the carrier wave emigration method of probabilistic forecasting and is comprised the following steps:

Step one, receives carrier wave emigration signal, enters the premigrates stage, collects in the virtual base station that carrier wave is corresponding all interior Deposit the nearest n times internal storage state change information of page；

Step 2, enters iterative migration stage, first run iteration, then the nearest n times internal storage state updating all pages becomes Change information, is then sent to destination by the full memory page of virtual base station, proceeds to step 5, be not first run iteration, then update The nearest n times internal storage state change information of all pages, and page information to be transmitted, proceed to step 3；

Step 3, it was predicted that the page of institute's available for transmission page rewrites probability；

Step 4, page is rewritten probability exceed rewrite probability threshold value page to be transmitted be put into shutdown migrate rank Section is transmitted, and page is rewritten probability and is put into epicycle iteration carries out less than the page to be transmitted rewriting probability threshold value Transmission；

Step 5, it may be judged whether meet and shut down transition condition, be unsatisfactory for, proceed to step 2, meets, and enters shutdown and migrates rank Section, completes carrier wave emigration.

Further, the nearest n times page state change information of all pages in the virtual base station of described collection carrier wave Specifically include:

The first step, receives carrier wave emigration signal, generate a line number be N, columns be virtual base station internal memory number of pages M of carrier wave Page status information table state_table；

Second step, with internal memory page status collection period T_cThe page information that is written over of statistics, from the of state_table A line starts, and according to the page information that is written over of statistics, page h is written over, then by the first row, h in state_table The number of row is set to 1, and page h is not written over, then the first row, the number of h row in state_table are set to 0；Further according to next The information being written over page of secondary statistics, is stored in second row of state_table by the variable condition information of all pages, Carrying out successively, until its Nth row is stored into the variable condition information of all pages in virtual base station, page variable condition is believed Breath is collected complete.

Further, the nearest n times internal storage state change information of all pages of described renewal specifically includes:

The first step, goes to update the data of replacement the 1st row, uses state_table by the data of the 2nd row of state_table The data of the 3rd row go to update the data replacing the 2nd row, the like, replace N-1 row until its Nth row data being updated Data；

Second step, is first run iteration, then all numbers of Nth row in state_table are updated to 1；It not first run iteration, Then add up the information being written over page in last round of iterative transition process, then update the Nth row of state_table, Page h is written over, then Nth row, the number of h row in state_table are set to 1, and page h is not written over, then will In state_table, Nth row, the number of h row are set to 0.

Further, described renewal page to be transmitted information, page to be transmitted includes:

(1) page being written over during last round of iterative migration；

(2) page that shutdown migration phase is transmitted it is put into.

Further, the described page rewriting probabilistic forecasting that carries out institute's available for transmission page specifically includes:

(1) according to the nearest n times internal storage state change information of page i to be transmitted, the time of page i to be transmitted is calculated Page rewrites probability P_it；

(2) according to the last internal storage state change information of the adjacent memory page of page i to be transmitted, calculate to be transmitted The space page of page i rewrites probability P_is；

(3) according to P_itAnd P_is, the page calculating page i to be transmitted rewrites probability P_i:

P_i=ω_tP_it+ω_sP_is；

Wherein ω_tIt is that time memory page rewrites probability right value, ω_sIt is that space page rewrites probability right value, ω_s+ ω_t=1；

(4) traversal institute available for transmission page, the page calculating institute's available for transmission page rewrites probability.

Further, the described nearest n times internal storage state change information according to page i to be transmitted, calculate internal memory to be transmitted The time memory page of page i rewrites probability P_itSpecifically include:

(1) from state_table, obtain the nearest n times internal storage state change information of page i to be transmitted, constitute to be passed The internal storage state change information vector A of defeated page i_i=(a_1i, a_2i..., a_Ni)^T, A_iThe i-th row equal to state_table；

(2)A_i≠(1,1,...,1)^T _1×NAnd A_i≠(0,0,...,0)^T _1×N, pass through A_iCalculate the weight of different prediction step Value, it was predicted that the weighted value of step-length k is:

$r_{ik}=\frac{\underset{l=1}{\overset{N-k}{\&Sigma;}}(a_{li}-{\stackrel{\&OverBar;}{A}}_i)\&CenterDot;(a_{(l+k)i}-{\stackrel{\&OverBar;}{A}}_i)}{\underset{l=1}{\overset{N}{\&Sigma;}}{(a_{li}-{\stackrel{\&OverBar;}{A}}_i)}^2},(k=1,2,\mathrm{......},j;j<N);$

WhereinIt is the meansigma methods of the internal storage state change information vector of page i to be transmitted,J is maximum predicted step-length；

(3) weighted value of different prediction step is standardized, obtain the weight coefficient of different prediction step, it was predicted that step The weight coefficient of long k is:

$W_{ik}=\frac{\left|r_{ik}\right|}{\underset{l=1}{\overset{j}{\&Sigma;}}\left|r_{i1}\right|};$

(4)A_i=(1,1 ..., 1)_1×NOr A_i=(0,0 ..., 0)_1×N, the most directly calculate the power of different prediction step Weight coefficient, it was predicted that the weight coefficient of step-length k is:

$W_{ik}=\frac{1}{j};$

(5) A is passed through_iCalculate the transition probability matrix of different prediction step, it was predicted that the transition probability matrix of step-length k is:

$M_i^{\left(k\right)}=\left[\begin{array}{cc}{p}_{00}^{\left(k\right)}& p_{01}^{\left(k\right)}\\ p_{10}^{\left(k\right)}& p_{11}^{\left(k\right)}\end{array}\right];$

WhereinM=0 or 1, n=0 or 1,

(6) by the weight coefficient of different prediction step, the transition probability matrix of different prediction step and internal memory to be transmitted Nearest j the internal storage state change information of page i calculates the page status predication vector P of time_itv:

$P_{itv}=W_{i1}\&CenterDot;(1-a_{Ni},a_{Ni})\&CenterDot;M_i^{\left(1\right)}+W_{i2}\&CenterDot;(1-a_{(N-1)i},a_{(N-1)i})\&CenterDot;M_i^{\left(2\right)}+\mathrm{...}+W_{ij}\&CenterDot;(1-a_{(N-j+1)i},a_{(N-j+1)i})\&CenterDot;M_i^{\left(j\right)};$

(7)P_it=P_itv(2), P_itv(2) P is represented_itvSecond element.

Further, the last internal storage state change information of the described adjacent memory page according to page i to be transmitted, meter The space page calculating page i to be transmitted rewrites probability P_isSpecifically include:

(1) from state_table, obtain the adjacent memory page i-2 of page i to be transmitted, adjacent memory page i-1, adjacent Page i+1, the last internal storage state change information a of adjacent memory page i+2_N(i-2)、a_N(i-1)、a_N(i+1)、a_N(i+2), a_N(i-2)Nth row equal to state_table, the i-th-2 row, a_N(i-1)Nth row equal to state_table, the i-th-1 row, a_N(i+1) Nth row equal to state_table, i+1 row, a_N(i+2)Nth row equal to state_table, the i-th+2 row；

(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-2a_N(i-2)+α_i-1a_N(i-1)+α_i+1a_N(i+1)+α_i+2a_N(i+2)；

Wherein α_i-2For the weight coefficient of page i-2, α_i-1Weight coefficient, α for page i-1_i+1For page i+1 Weight coefficient, α_i+2For the weight coefficient of page i+2, α_i-1=α_i+1, α_i-2=α_i+2, α_i-1=2 α_i-2, α_i-2+α_i-1+α_i+1+ α_i+2=1；

(3) when being edge page for page i to be transmitted, P_isCalculating parameter be handled as follows:

When i=1 when, a_N(i-2)=0, a_N(i-1)=0, α_i-2=0, α_i-1=0, α_i+1=2 α_i+2, α_i+1+α_i+2=1；

When i=2 when, a_N(i-2)=0, α_i-2=0, α_i-1=α_i+1, α_i+1=2 α_i+2, α_i-1+α_i+1+α_i+2=1；

When i=M-1 when, a_N(i+2)=0, α_i+2=0, α_i-1=α_i+1, α_i-1=2 α_i-2, α_i-1+α_i+1+α_i+2=1；

When i=M when, a_N(i+2)=0, a_N(i+1)=0, α_i+2=0, α_i+1=0, α_i-1=2 α_i-2, α_i-1+α_i-2=1.

Further, described shutdown transition condition includes:

(1) iterative migration number of times is more than 30 times；

(2) it is put into and shuts down the internal memory number of pages that is written over during the internal memory number of pages of migration phase transmission and epicycle iterative migration Sum is less than page 50.

Further, described entrance shuts down migration phase, completes carrier wave emigration and specifically includes:

(1) source virtual base station is closed；

(2) last is taken turns the page being written over during iterative migration and is put into the page biography of shutdown stage transmission Deliver to purpose virtual base station, CPU state is sent to purpose virtual base station simultaneously；

(3) starting purpose virtual base station, carrier wave emigration completes.

The carrier wave emigration method rewriting probabilistic forecasting based on page that the present invention provides, during optimizing carrier wave emigration The redundancy iteration copy of virtual base station page, compares the carrier wave emigration method of tradition pre-copy iteration, and its multipotency reduces about The gross migration time of 90% and transmitted data amount, multipotency reduce by the shutdown migration time of about 80%, improve carrier wave emigration Migration performance.

Accompanying drawing explanation

Fig. 1 is the carrier wave emigration method flow diagram rewriting probabilistic forecasting based on page that the embodiment of the present invention provides.

Fig. 2 is the calculation flow chart that the time memory page that the embodiment of the present invention provides rewrites probability.

Fig. 3 is the calculation flow chart that the space page that the embodiment of the present invention provides rewrites probability.

Fig. 4 is that the carrier wave emigration method based on page rewriting probabilistic forecasting that the embodiment of the present invention provides is copied in advance with tradition Shellfish iteration carrier wave emigration method gross migration time diagram of carrier wave emigration in the case of different user business number.

Fig. 5 is that the carrier wave emigration method based on page rewriting probabilistic forecasting that the embodiment of the present invention provides is copied in advance with tradition Shellfish iteration carrier wave emigration method is the shutdown migration time schematic diagram of carrier wave emigration in the case of different user business number.

Detailed description of the invention

In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with embodiment, to the present invention It is further elaborated.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not used to Limit the present invention.

Below in conjunction with the accompanying drawings the application principle of the present invention is explained in detail.

As it is shown in figure 1, the carrier wave emigration method based on page rewriting probabilistic forecasting of the embodiment of the present invention includes following Step:

S101: receive carrier wave emigration signal, enters the premigrates stage, collects all internal memories in the virtual base station that carrier wave is corresponding The nearest n times internal storage state change information of page；

S102: enter the iterative migration stage, if first run iteration, then update the nearest n times internal memory shape of all pages State change information, is then sent to destination by the full memory page of virtual base station, proceeds to step S105, change if not the first run In generation, update the nearest n times internal storage state change information of all pages, and page information to be transmitted, proceed to step S103；

S103: the page of prediction institute available for transmission page rewrites probability；

S104: page is rewritten probability exceed rewrite probability threshold value page to be transmitted be put into shutdown migration phase It is transmitted, page is rewritten probability and is put into epicycle iteration passes less than the page to be transmitted rewriting probability threshold value Defeated；

S105: judge whether to meet and shut down transition condition, if be unsatisfactory for, proceeding to step S102, stopping if it is satisfied, enter Machine migration phase, completes carrier wave emigration.

In step S101, collect the nearest n times page state change information of all pages in the virtual base station of carrier wave, Concrete execution according to the following steps:

(1) receive carrier wave emigration signal, generate a line number be N, columns be virtual base station internal memory number of pages M interior of carrier wave Deposit page status information table state_table；

(2) with internal memory page status collection period T_cThe page information that statistics is written over, from the first row of state_table Start, according to the information being written over page of statistics, if page h is written over, then by the first row, h in state_table The number of row is set to 1, if page h is not written over, then the first row, the number of h row in state_table is set to 0, root the most again According to the information being written over page added up next time, the variable condition information of all pages is stored in state_table's Second row, is carried out successively, until its Nth row is stored into the variable condition information of all pages in virtual base station, page becomes Change state information collection complete；

In step S102, update the nearest n times internal storage state change information of all pages, hold the most according to the following steps OK:

(1) go to update the data of replacement the 1st row, with the 3rd of state_table the by the data of the 2nd row of state_table The data of row go to update the data replacing the 2nd row, the like, replace N-1 row data until its Nth row data being updated；

(2) if first run iteration, then all numbers of Nth row in state_table are updated to 1, if not the first run Iteration, then add up the information being written over page in last round of iterative transition process, update state_table the most accordingly Nth row, if page h is written over, then the number of Nth row in state_table, h row is set to 1, if page h is not changed Write, then Nth row, the number of h row in state_table are set to 0.

In step S102, updating page information to be transmitted, page to be transmitted includes:

(1) page being written over during last round of iterative migration；

(2) page that shutdown migration phase is transmitted it is put into.

In step S103, institute's available for transmission page is carried out page and rewrites probabilistic forecasting, enter the most according to the following steps OK:

(1) according to the nearest n times internal storage state change information of page i to be transmitted, the time of page i to be transmitted is calculated Page rewrites probability P_it；

(2) according to the last internal storage state change information of the adjacent memory page of page i to be transmitted, calculate to be transmitted The space page of page i rewrites probability P_is；

(3) according to P_itAnd P_is, the page calculating page i to be transmitted rewrites probability P_i:

P_i=ω_tP_it+ω_sP_is；

Wherein ω_tIt is that time memory page rewrites probability right value, ω_sIt is that space page rewrites probability right value, ω_s+ ω_t=1；

(4) traversal institute available for transmission page, the page calculating institute's available for transmission page rewrites probability.

As in figure 2 it is shown, according to the nearest n times internal storage state change information of page i to be transmitted, calculate page to be transmitted The time memory page of i rewrites probability P_it, specifically sequentially include the following steps:

(1) from state_table, obtain the nearest n times internal storage state change information of page i to be transmitted, constitute to be passed The internal storage state change information vector A of defeated page i_i=(a_1i, a_2i..., a_Ni)^T, A_iThe i-th row equal to state_table；

(2) if A_i≠(1,1,...,1)^T _1×NAnd A_i≠(0,0,...,0)^T _1×N, pass through A_iCalculate the power of different prediction step Weight values, it was predicted that the weighted value of step-length k is:

$r_{ik}=\frac{\underset{l=1}{\overset{N-k}{\&Sigma;}}(a_{li}-{\stackrel{\&OverBar;}{A}}_i)\&CenterDot;(a_{(l+k)i}-{\stackrel{\&OverBar;}{A}}_i)}{\underset{l=1}{\overset{N}{\&Sigma;}}{(a_{li}-{\stackrel{\&OverBar;}{A}}_i)}^2},(k=1,2,\mathrm{......},j;j<N);$

WhereinIt is the meansigma methods of the internal storage state change information vector of page i to be transmitted,J is maximum predicted step-length；

(3) weighted value of different prediction step is standardized, obtain the weight coefficient of different prediction step, it was predicted that step The weight coefficient of long k is:

$W_{ik}=\frac{\left|r_{ik}\right|}{\underset{l=1}{\overset{j}{\&Sigma;}}\left|r_{i1}\right|};$

(4) if A_i=(1,1 ..., 1)_1×NOr A_i=(0,0 ..., 0)_1×N, the most directly calculate different prediction step Weight coefficient, it was predicted that the weight coefficient of step-length k is:

$W_{ik}=\frac{1}{j};$

(5) A is passed through_iCalculate the transition probability matrix of different prediction step, it was predicted that the transition probability matrix of step-length k is:

$M_i^{\left(k\right)}=\left[\begin{array}{cc}{p}_{00}^{\left(k\right)}& p_{01}^{\left(k\right)}\\ p_{10}^{\left(k\right)}& p_{11}^{\left(k\right)}\end{array}\right];$

WhereinM=0 or 1, n=0 or 1,

(6) by the weight coefficient of different prediction step, the transition probability matrix of different prediction step and internal memory to be transmitted Nearest j the internal storage state change information of page i calculates the page status predication vector P of time_itv:

$P_{itv}=W_{i1}\&CenterDot;(1-a_{Ni},a_{Ni})\&CenterDot;M_i^{\left(1\right)}+W_{i2}\&CenterDot;(1-a_{(N-1)i},a_{(N-1)i})\&CenterDot;M_i^{\left(2\right)}+\mathrm{...}+W_{ij}\&CenterDot;(1-a_{(N-j+1)i},a_{(N-j+1)i})\&CenterDot;M_i^{\left(j\right)};$

(7)P_it=P_itv(2), P_itv(2) P is represented_itvSecond element.

As it is shown on figure 3, the last internal storage state change information of the adjacent memory page according to page i to be transmitted, meter The space page calculating page i to be transmitted rewrites probability P_is, specifically sequentially include the following steps:

(1) from state_table, obtain the adjacent memory page i-2 of page i to be transmitted, adjacent memory page i-1, adjacent Page i+1, the last internal storage state change information a of adjacent memory page i+2_N(i-2)、a_N(i-1)、a_N(i+1)、a_N(i+2), a_N(i-2)Nth row equal to state_table, the i-th-2 row, a_N(i-1)Nth row equal to state_table, the i-th-1 row, a_N(i+1) Nth row equal to state_table, i+1 row, a_N(i+2)Nth row equal to state_table, the i-th+2 row；

(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-2a_N(i-2)+α_i-1a_N(i-1)+α_i+1a_N(i+1)+α_i+2a_N(i+2)；

Wherein α_i-2For the weight coefficient of page i-2, α_i-1Weight coefficient, α for page i-1_i+1For page i+1 Weight coefficient, α_i+2For the weight coefficient of page i+2, α_i-1=α_i+1, α_i-2=α_i+2, α_i-1=2 α_i-2, α_i-2+α_i-1+α_i+1+ α_i+2=1；

(3) when being edge page for page i to be transmitted, P_isCalculating parameter be handled as follows:

When i=1 when, a_N(i-2)=0, a_N(i-1)=0, α_i-2=0, α_i-1=0, α_i+1=2 α_i+2, α_i+1+α_i+2=1；

When i=2 when, a_N(i-2)=0, α_i-2=0, α_i-1=α_i+1, α_i+1=2 α_i+2, α_i-1+α_i+1+α_i+2=1；

When i=M-1 when, a_N(i+2)=0, α_i+2=0, α_i-1=α_i+1, α_i-1=2 α_i-2, α_i-1+α_i+1+α_i+2=1；

When i=M when, a_N(i+2)=0, a_N(i+1)=0, α_i+2=0, α_i+1=0, α_i-1=2 α_i-2, α_i-1+α_i-2=1.

Shutdown transition condition in step S105, particularly as follows: meet following one of which:

(1) iterative migration number of times is more than 30 times.

(2) it is put into and shuts down the internal memory number of pages that is written over during the internal memory number of pages of migration phase transmission and epicycle iterative migration Sum is less than page 50.

Enter and shut down migration phase, complete carrier wave emigration, perform the most according to the following steps:

(1) source virtual base station is closed.

(2) last is taken turns the page being written over during iterative migration and is put into the page biography of shutdown stage transmission Deliver to purpose virtual base station, CPU state is sent to purpose virtual base station simultaneously.

(3) starting purpose virtual base station, carrier wave emigration completes.

Below in conjunction with emulation, the application effect of the present invention is explained in detail.

In order to test the migration performance of the present invention, parameter is provided that the virtual base station memory size of carrier wave is 2048MB； Migrate and carry a width of 1000Mb/s；Collection period is 2s；Rewriting probability threshold value is 0.7；N=20；Maximum predicted step-length is 4；Time Between page rewrite probability right value be 0.5；It is 0.5 that space page rewrites probability right value.Select tradition pre-copy iteration Carrier wave emigration method method as a comparison, carry out the emulation of 30 Monte Carlo Experiments, obtain carrier wave emigration as shown in Figure 4 The shutdown migration time of gross migration time and carrier wave emigration as shown in Figure 5.

From Fig. 4 and Fig. 5, present invention multipotency reduces by gross migration time and the transmitted data amount, multipotency of about 90% Reduce by the shutdown migration time of about 80%, improve the migration performance of carrier wave emigration.

The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all essences in the present invention Any amendment, equivalent and the improvement etc. made within god and principle, should be included within the scope of the present invention.

Claims (4)

1. the carrier wave emigration method rewriting probabilistic forecasting based on page, it is characterised in that described based on page rewriting The carrier wave emigration method of probabilistic forecasting comprises the following steps:

Step one, receives carrier wave emigration signal, enters the premigrates stage, collects all pages in the virtual base station that carrier wave is corresponding Nearest n times internal storage state change information；

Step 2, enters iterative migration stage, first run iteration, then update the nearest n times internal storage state change letter of all pages Breath, is then sent to destination by the full memory page of virtual base station, proceeds to step 5, be not first run iteration, then update all The nearest n times internal storage state change information of page, and page information to be transmitted, proceed to step 3；

Step 2, enters the iterative migration stage, if first run iteration, then updates the nearest n times internal storage state of all pages Change information, is then sent to destination by the full memory page of virtual base station, proceeds to step 5, if not first run iteration, Update the nearest n times internal storage state change information of all pages, and page information to be transmitted, proceed to step 3；

Step 3, it was predicted that the page of institute's available for transmission page rewrites probability；

Step 4, page is rewritten probability exceed rewrite probability threshold value page to be transmitted be put into shutdown migration phase enter Row transmission, rewrites page probability and is put into epicycle iteration passes less than the page to be transmitted rewriting probability threshold value Defeated；

Step 5, it may be judged whether meet and shut down transition condition, be unsatisfactory for, proceed to step 2, meets, and enters and shuts down migration phase, Complete carrier wave emigration.

2. rewrite the carrier wave emigration method of probabilistic forecasting as claimed in claim 1 based on page, it is characterised in that to all Page to be transmitted carries out page and rewrites probabilistic forecasting and specifically include following steps:

The first step, according to the nearest n times internal storage state change information of page i to be transmitted, calculates the time of page i to be transmitted Page rewrites probability P_it；

Second step, according to the last internal storage state change information of the adjacent memory page of page i to be transmitted, calculates to be transmitted The space page of page i rewrites probability P_is；

3rd step, according to P_itAnd P_is, the page calculating page i to be transmitted rewrites probability P_i:

P_i=ω_tP_it+ω_sP_is；

Wherein ω_tIt is that time memory page rewrites probability right value, ω_sIt is that space page rewrites probability right value, ω_s+ω_t=1；

4th step, traversal institute available for transmission page, the page calculating institute's available for transmission page rewrites probability.

3. rewrite as claimed in claim 2 the carrier wave emigration method of probabilistic forecasting based on page, it is characterised in that described the One step specifically sequentially includes the following steps:

(1) from page status information table state_table, obtain the nearest n times internal storage state change of page i to be transmitted Information, constitutes the internal storage state change information vector A of page i to be transmitted_i=(a_1i, a_2i..., a_Ni)^T, A_iEqual to state_ I-th row of table；

(2) if A_i≠(1,1,...,1)^T _1×NAnd A_i≠(0,0,...,0)^T _1×N, pass through A_iCalculate the weight of different prediction step Value, it was predicted that the weighted value of step-length k is:

$r_{ik}=\frac{\underset{l=1}{\overset{N-k}{\&Sigma;}}\left(a_{li}-{\stackrel{\&OverBar;}{A}}_i\right)\&CenterDot;\left(a_{\left(l+k\right)i}-{\stackrel{\&OverBar;}{A}}_i\right)}{\underset{l=1}{\overset{N}{\&Sigma;}}{\left(a_{li}-{\stackrel{\&OverBar;}{A}}_i\right)}^2},\left(k=1,2,\mathrm{......},j;j<N\right);$

WhereinIt is the meansigma methods of the internal storage state change information vector of page i to be transmitted,J is Maximum predicted step-length；

(3) weighted value of different prediction step is standardized, obtain the weight coefficient of different prediction step, it was predicted that step-length k Weight coefficient be:

$W_{ik}=\frac{\left|r_{ik}\right|}{\underset{l=1}{\overset{j}{\&Sigma;}}\left|r_{il}\right|};$

(4) if A_i=(1,1 ..., 1)_1×NOr A_i=(0,0 ..., 0)_1×N, the most directly calculate the weight of different prediction step Coefficient, it was predicted that the weight coefficient of step-length k is:

$W_{ik}=\frac{1}{j};$

(5) A is passed through_iCalculate the transition probability matrix of different prediction step, it was predicted that the transition probability matrix of step-length k is:

$M_i^{\left(k\right)}=\left[\begin{array}{cc}{p}_{00}^{\left(k\right)}& p_{01}^{\left(k\right)}\\ p_{10}^{\left(k\right)}& p_{11}^{\left(k\right)}\end{array}\right];$

WhereinM=0 or 1, n=0 or 1,

(6) by the weight coefficient of different prediction step, the transition probability matrix of different prediction step and page i to be transmitted Nearest j internal storage state change information calculates the page status predication vector P of time_itv:

$P_{itv}=W_{i1}\&CenterDot;(1-a_{Ni},a_{Ni})\&CenterDot;M_i^{\left(1\right)}+W_{i2}\&CenterDot;(1-a_{\left(N-1\right)i},a_{(N-1)i})\&CenterDot;M_i^{\left(2\right)}+\mathrm{...}+W_{ij}\&CenterDot;(1-a_{(N-j+1)i},a_{(N-j+1)i})\&CenterDot;M_i^{\left(j\right)};$

(7)P_it=P_itv(2), P_itv(2) P is represented_itvSecond element.

4. rewrite the carrier wave emigration method of probabilistic forecasting as claimed in claim 2 based on page, it is characterised in that described According to the last internal storage state change information of the adjacent memory page of page i to be transmitted, calculate the space of page i to be transmitted Page rewrites probability P_is, specifically sequentially include the following steps:

(1) from state_table, obtain the adjacent memory page i-2 of page i to be transmitted, adjacent memory page i-1, adjacent memory Page i+1, the last internal storage state change information a of adjacent memory page i+2_N(i-2)、a_N(i-1)、a_N(i+1)、a_N(i+2), a_N(i-2)Deng In Nth row, i-th-2 row of state_table, a_N(i-1)Nth row equal to state_table, the i-th-1 row, a_N(i+1)It is equal to The Nth row of state_table, i+1 arrange, a_N(i+2)Nth row equal to state_table, the i-th+2 row；

(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-2a_N(i-2)+α_i-1a_N(i-1)+α_i+1a_N(i+1)+α_i+2a_N(i+2)；

Wherein α_i-2For the weight coefficient of page i-2, α_i-1Weight coefficient, α for page i-1_i+1Power for page i+1 Weight coefficient, α_i+2For the weight coefficient of page i+2, α_i-1=α_i+1, α_i-2=α_i+2, α_i-1=2 α_i-2, α_i-2+α_i-1+α_i+1+α_i+2= 1；

(3) when being edge page for page i to be transmitted, P_isCalculating parameter be handled as follows:

As i=1, a_N(i-2)=0, a_N(i-1)=0, α_i-2=0, α_i-1=0, α_i+1=2 α_i+2, α_i+1+α_i+2=1；

As i=2, a_N(i-2)=0, α_i-2=0, α_i-1=α_i+1, α_i+1=2 α_i+2, α_i-1+α_i+1+α_i+2=1；

As i=M-1, a_N(i+2)=0, α_i+2=0, α_i-1=α_i+1, α_i-1=2 α_i-2, α_i-1+α_i+1+α_i+2=1；

As i=M, a_N(i+2)=0, a_N(i+1)=0, α_i+2=0, α_i+1=0, α_i-1=2 α_i-2, α_i-1+α_i-2=1.