CN103248500B - A kind of real-time data broadcast dispatching method as required considering data item size - Google Patents

Abstract

The invention provides a kind of real-time data broadcast dispatching method as required considering data item size, deal with data item situation not of uniform size is carried out by the thought split, and the scheduling (US-LxRxW) of scheduling (ES-LxRxW) that size data item splits and the size data item fractionation such as or not to propose, fractionation strategy and the adjustable strategies of off period of size data item such as to propose, propose the scheduling strategy of the fractionation strategy that do not wait size data item and PF and the PxBF two kinds size data item such as not.The scheduling of the scheduling that what the present invention proposed wait size data item to split and the fractionation of size data item such as not, all take into account the bandwidth sum cycle, and the fractionation benchmark size of the size data item fractionations such as the distribution for the distribution of different off periods and different data item size proposes, propose for different data item sizes and different scheduling strategies the fractionation benchmark size that the size data item such as not splits, can be good at adapting to real-time situation.

Description

A kind of real-time data broadcast dispatching method as required considering data item size

Technical field

Data broadcast of the present invention scheduling field, is specifically related to a kind of Data Broadcast under Mobile Enviroment dispatching method, especially relates to a kind ofly considering cycle data broadcast as required in real time.

Background technology

The dispatching algorithm of data broadcast is content most important, most crucial in data broadcast.Data broadcast dispatching algorithm refers in research broadcasting process, which data broadcast of this selection is gone out, and the algorithm of the priority problem of broadcast data.In general, quiescent period data broadcast, as required data broadcast and mixing broadcast three kinds is divided into.This section introduces the focus of current research respectively from quiescent period data broadcast, as required data broadcast, mixing broadcast and index four parts.

(1) static data broadcast.Major part have studied static broadcast mode, and the algorithm that wherein influence power is maximum is multi-disk schedule algorithm, which proposes data item large for access probability to be put into coil upper broadcast than faster, data item smaller for access probability is put on slow dish and broadcasts.AzerBestavros proposes the AIDA algorithm based on multi-disk schedule algorithm, and this algorithm improves promptness and serious forgiveness, allows the multipriority of better control data item simultaneously.SanjoyBaruah etc. turn for wheel the another kind that Real-Time Task System proposes based on AIDA multi-disk schedule algorithm and improve dispatching method, define the model of real-time fault tolerance broadcast disk, experimental result shows that the scheduling that the design of designated broadcast disk in this model and wheel turn Real-Time Task System is closely-related.

(2) data broadcast as required.Because static broadcast can not respond the request of user, in time so on-demand broadcasting becomes hot topic gradually.PingXuan etc. propose BoD model, and it can be broadcasted timely according to the request of user, and propose EDF algorithm and demonstrate its validity with experiment, experimental result shows, BoD successfully can meet the time-constrain of request.DemetAksoy etc. exactly define in the large-scale performance of on-demand broadcasting environment dispatching algorithm and the standard of scale, propose RxW algorithm, it shows functional in all these standards, and the demand of specifically application or a system can well adapt to; RxW is strong in the colony of user and the change of workload, because it does scheduling based on the state of current request queue instead of the access probability of estimated data's item to select.QiuFang etc. have studied traditional non-moving real-time policy and non real-time shift strategy in great detail, demonstrate traditional real time algorithm in mobile environment not necessarily or performance optimum; Propose based on pull-type real time broadcasting system, and propose an efficient dispatching algorithm ACR, it designs to mobile terminal to transmit data in time.The urgency that JianliangXu etc. consider the time and the quantity not completing request propose a kind of new dispatching algorithm SIN-α, and the one proposing this algorithm effectively realizes, the theoretical boundary of the request failure rate when asking the speed arrived to be increased to infinite is also analyzed in literary composition, experiment proves that SIN-α is better than existing algorithm in most operating load, and when high request rate close to the theoretical boundary analyzed out.JunChen etc. have studied the request of many data item in the data broadcast as required of band time restriction, propose algorithm DTIU, it combines the advantage of data item scheduling and request scheduling, the performance test results shows, this algorithm not only reduces request failure rate, in turn save the bandwidth of broadcast channel, and be better than other classic algorithm under the impact of many factors.JosephKee-YinNg etc. propose algorithm MRS and its mutation with buffer memory real-time transmission data in real time information dispatching patcher.This algorithm family considers the off period of request, the size of data object and the popularization of data in scheduling decision.Experiment proves, this algorithm can not only broadcast data on time, is better than traditional dispatching algorithm 20% in the response time simultaneously.KaiLiu have studied the request scheduling of multichannel entry under on-demand broadcasting environment, proposes algorithm DUP, solves the problem of the hungry problem of request and bandwidth availability ratio, and experiment demonstrates the superiority of algorithm.It is np problem that JingsongLv etc. to demonstrate under single channel and multichannel environment that real-time broadcast dispatches multinomial request, propose two algorithm PVC and SSA based on " profit ", for single channel and multi channel scheduling respectively, simulation result shows, compared with traditional algorithm, there is great improvement.

(3) mixing broadcast.Mixing broadcast refers to that it is by a part of hot spot data periodic broadcasting, the non-hot spot data on-demand broadcasting of another part by quiescent period broadcast and a kind of broadcast scheduling algorithm of combining of data broadcast as required.The propositions such as SunhoKim, the request of statistics collection client, is divided into hot and cold two data sets to carry out mixed scheduling by data item, the data broadcast regular according to its access probability of heat, and cold data were broadcasted according to stand-by period and the number of request that do not complete.According to average response time, the best separation of cold and hot data item of having analyzed and researched, and use Web data query to emulate the validity of the scheme that demonstrates.JungHwanOh has inquired into a kind of method of ADAPTIVE MIXED, the data that only broadcast is more popular, not too conventional data are transmitted to save the network bandwidth in the mode of program request, propose a kind of new technology, be broadcast and unicast allocated bandwidth in an integrated fashion, simulation result shows that this technology is significantly better than contemporary technology.

(4) index.The object of introduction index is the energy ezpenditure in order to reduce mobile device, because after adding index, mobile device checks that index just can know when the data item oneself needed is broadcasted, thus be switched to park mode during this period of time what wait for, save the energy, go to receive data until be just switched to listen mode time data item is come.Due to the limitation of the mobile device energy, a lot of index all in the broadcast of research Mobile data in recent years.Wherein, major part is all the index in research static broadcast, and JianliangXu proposes a kind of index with parameter----exponential index, and can optimize access delay under the condition of the boundary of given tuning period, vice versa.JianliangXu proposes a new index data structure Dtree, is used to provide the inquiry of the data be correlated with in position in mobile environment, and experimental result shows that Dtree has better overall performance than existing structure.The research of data broadcast as required of a part is all the algorithm for broadcast scheduling, and SwarupAcharya proposes algorithm MAX according to optimizing the standard of single request under worst case, and this algorithm is done well in an experiment.DemetAksoy introduces a Parameterized Algorithm, all evaluation criterions all shows good, and can adjust the time in average or the worst wait situation, also proposed simulation model in literary composition.Minority is by indexes applications to data broadcast as required.Jiun-LongHuang propose one adapt to index and data organize algorithm AIDOA, can adjust the length of index dynamically, experimental result also demonstrates the validity of this algorithm.

Consult pertinent literature, before discovery, the document of research data broadcast scheduling as required is in real time all when being consistent for data item size, the validity of research dispatching algorithm.And for the setting of data item, be all make a search when tentation data item size is fixing.But in practice, the data item size of request is inconsistent, may occur the situation of very large data item, also may occur the situation of very little data item, and the distribution of data item size also has different situations.As user request if webpage or text, so data item size probably only has a few KB; If request is video or large pictorial information, so data item just may have the size of a few MB or tens MB.

Simultaneously due to technical equipment or wireless transmission medium, bandwidth in different environments size is different.As in 2G environment, bandwidth can only reach 10Kbps, and amount of bandwidth can reach 10Mbps in 3G environment, and in 4G environment, bandwidth can reach 100Mbps.And for 3G network, from technical standpoint, 3G network is only the broadband network that last one kilometer uses wireless access, its actual bandwidth depends on signal strength signal intensity, and network capacity.

In data broadcast before, each data item only has and could serve for other mobile subscribers after the broadcast is completed.If broadcast large data item, when bandwidth is certain, so the time of this data item busy channel is long, and the number of users of service in the unit interval is reduced, and service quality reduces.Equally, if data item size is certain, when bandwidth is smaller, the time of data item busy channel is also long.

Summary of the invention

For the research for real-time data broadcast as required in prior art, all the validity of data broadcast scheduling algorithm, be all hypothesis data item size be fixing situation, but all differing of data item size is not taken into account, and do not consider the factor varied in size in bandwidth sum cycle in different environment; The present invention proposes a kind of real-time data broadcast dispatching method as required considering data item size.

Technical scheme of the present invention is 1. 1 kinds of real-time data broadcast dispatching methods as required considering data item size, it is characterized in that: comprise following process,

Mobile client submits the request of oneself to by uploading channel after, server receives request, first check this request whether in broadcast queue or prepare in queue, if, then direct request is added in broadcast queue or prepares in the request queue of corresponding data subitem in queue; If not, then request is added in the request queue of system by request time order from small to large, and be request dispatching ID;

Server takes out the request of current time from the request queue of system, searches the data item meeting request, and obtain data item and be organized into data item information, described data item information comprises the ID of data item and the content of data item;

The data item that the large young pathbreaker of server based on data item gets splits, and is added by the data subitem split in preparation queue; A data item splits into more than one data subitem, and the data item ID of these data subitems is identical, but data subitem ID is different; Prepare queue first to arrange by the ID of data item, the data subitem of same data item ID arranges by data subitem ID, and request current for data item is kept at the request queue splitting each data subitem of gained, the data subitem prepared in queue is called preparation item;

Server periodically carries out scheduling tissue by dispatching algorithm to the data prepared in queue, and dynamic data subitem of selecting adds in broadcast queue, and the data subitem in broadcast queue is broadcast, and the data subitem in broadcast queue is called broadcast item.

And employing waits size data item to split mode when the data item got being split, the fractionation of size data item fractionation mode and scheduling flow is waited to comprise the following steps,

Step a1, initialization;

Step a2, if current time is not asked, then jumps to step a7, otherwise continues to enter step a3;

Step a3, takes out the request of current time;

Step a4, obtains the data item of this request, and according to waiting fractionation strategy of size data item to be split, the off period of the data item of fractionation adjusts by the adjustable strategies according to the off period;

Step a5, process respectively for all data subitems after splitting, processing mode is as follows,

To wherein arbitrary data subitem, if at the broadcast item of broadcast queue and no broadcast, then this request is added in the request queue of this broadcast item; Otherwise, if prepare the preparation item in queue, then this request is added in the request queue of this preparation item; Otherwise, this data subitem is added preparation queue as new preparation item, and request is added in the request queue preparing new preparation item in queue;

Step a6, waits request newly to be generated, jumps to step a2;

Step a7, broadcasts according to broadcast queue;

Step a8, to judge whether just to have broadcasted in broadcast queue not empty broadcast item, otherwise directly enters step a9, that the successful request information of this broadcast item is write successful solicited message file, delete all requests of this broadcast item, delete this broadcast item, then enter step a9;

Step a9, judges whether that broadcast queue is empty, if otherwise directly enter step a10, if broadcast queue be empty, then according to etc. the dispatching algorithm of size fractionation data item generate new broadcast queue, then enter step a10;

Step a10, advance a time point, if exceeded the running time of regulation, then jumps to step a11; Otherwise, jump procedure a2;

Step a11, reads successful request message file, statistics solicited message.

And, according to waiting fractionation strategy of size data item to split in step a4, comprise following sub-step,

Step a4.1, initialization, comprises and arranges residue size remainsize=datasize, subitem sequence number subitemorder=1; Datasize is data item size;

Step a4.2, determines to split benchmark size SplitStandardSize as follows,

$\mathrm{SplitS}\tan \mathrm{dardSize}=\frac{{\mathrm{Bw}}_k\×{\mathrm{Cy}}_k}{\mathrm{SplitK}}$

Wherein, Bw _kfor the broadcast bandwidth of kth time broadcast cycle, Cy _kfor the broadcast cycle of kth time broadcast cycle; SplitK is a positive integer, represents that the data items size of broadcast in one-period is the number of the data subitem splitting benchmark size SplitStandardSize;

Step a4.3, if remainsize > is SplitStandardSize, then continues step a4.4, otherwise jumps to step a4.5;

Step a4.4, add a data subitem, data subitem size is SplitStandardSize, filling size is SplitStandardSize, sequence number is subitemorder, subitemorder=subitemorder+1, remainsize=remainsize-SplitStandardSize, jump to step a4.3;

Step a4.5, add a data subitem, data subitem size is SplitStandardSize, and filling size is remainsize, and sequence number is subitemorder, remainsize=0;

Step a4.6, returns the data subitem split.

And, in step a9 according to etc. size split the dispatching algorithm of data item and generate new broadcast queue, comprise following sub-step,

Step a9.1, enters circulation after the preparation queue before time broadcast of input kth,

Step a9.2, upgrades and prepares each data subitem d in queue _x,ycurrent request number, off period and stand-by period;

Step a9.3, calculates each data subitem d _x,ydispatching priority;

Step a9.4, will have the data subitem d of greatest priority value _x,ybe designated as d _n,mand be assigned to Bd _j;

Step a9.5, judges that in broadcast queue, all data subitem sizes add d _n,msize whether be greater than Bw _k× Cy _kif be greater than, exited circulation and enter step a9.8, otherwise continue to perform step a9.6;

Step a9.6, by Bd _jinsert broadcast queue B _kafterbody, returns step a9.2;

Step a9.7, end loop;

Step a9.8, returns kCi broadcast queue B _k.

Or, adopt when the data item got is split and do not wait size data item to split mode, do not wait the fractionation of size data item fractionation mode and scheduling flow to comprise the following steps,

Step b1, initialization;

Step b2, if current time is not asked, then jumps to step b7, otherwise continues to enter step b3;

Step b3, takes out the request of current time;

Step b4, obtains the data item of this request, and according to not waiting fractionation strategy of size data item to be split, the off period of the data item of fractionation adjusts by the adjustable strategies according to the off period;

Step b5, process respectively for all data subitems after splitting, processing mode is as follows,

To wherein arbitrary data subitem, if at the broadcast item of broadcast queue and no broadcast, then this request is added in the request queue of this broadcast item; Otherwise, if prepare the preparation item in queue, then this request is added in the request queue of this preparation item; Otherwise, this data subitem is added preparation queue as new preparation item, and request is added in the request queue preparing new preparation item in queue;

Step b6, waits request newly to be generated, jumps to step b2;

Step b7, broadcasts according to broadcast queue;

Step b8, to judge whether just to have broadcasted in broadcast queue not empty broadcast item, otherwise directly enters step b9, that the successful request information of this broadcast item is write successful solicited message file, delete all requests of this broadcast item, delete this broadcast item, then enter step b9;

Step b9, judges whether that broadcast queue is empty, if otherwise directly enter step b10, if broadcast queue be empty, then generate new broadcast queue according to the dispatching algorithm of the size fractionation data item such as not, then enter step b10;

Step b10, advance a time point, if exceeded the running time of regulation, then jumps to step b11; Otherwise, jump procedure b2;

Step b11, reads successful request message file, statistics solicited message.

And, according to not waiting fractionation strategy of size data item to split in step b4, comprise following sub-step,

Step b4.1, input data item d _i, data item d _idata item size , data item priority Weight (d _i,j), bandwidth B w _k, cycle Cy _krear initialization, comprises that to arrange remaining data item size be data item size, subitem sequence number subitemorder=1;

Step b4.2, calculates the fractionation benchmark size SplitStandardSize of this data item,

$\mathrm{SplitS}\tan \mathrm{dardSize}=\frac{\mathrm{\α}\×{\mathrm{Bw}}_k\×{\mathrm{Cy}}_k}{S_{d_i}\×\mathrm{Weight}\left(d_{i,j}\right)}$

Wherein, α is parameter;

Step b4.3, if current residual data item size is greater than SplitStandardSize, then continues step b4.4, otherwise jumps to step b4.5;

Step b4.4, add a data subitem, data subitem size is SplitStandardSize, and sequence number is subitemorder, subitemorder=subitemorder+1, and remaining data item size deducts SplitStandardSize, jumps to step b4.3; Step b4.5, adds a data subitem, and data subitem size is remaining data item size, and sequence number is subitemorder;

Step b4.6, returns the data subitem split.

And the dispatching algorithm splitting data item according to the size such as not in step b9 generates new broadcast queue, adopts the scheduling strategy that priority is maximum, comprises following sub-step,

Step 9.1.1, after input prepares queue, initialization broadcast queue is empty;

Step 9.1.2, carries out following circulation,

Step 9.1.2.1, traversal prepares queue, chooses the data subitem with greatest priority LxRxW value;

Step 9.1.2.2, if the total size of data subitem in broadcast queue adds this data subitem size still do not exceed the data subitem size that one-period can broadcast, then adds this data subitem to broadcast queue, returns step 9.1.2.1 and circulate; Otherwise, exit circulation, export broadcast queue.

Or the dispatching algorithm splitting data item according to the size such as not in step b9 generates new broadcast queue, adopts the scheduling strategy considering priority and bandwidth availability ratio, comprises following sub-step,

Step 9.2.1, after input prepares queue, initialization broadcast queue is empty;

Step 9.2.2, carries out following circulation,

Step 9.2.2.1, traversal prepare queue, choose have minimum the data subitem of value;

Step 9.2.2.2, if the total size of data subitem in broadcast queue adds this data subitem size still do not exceed the data subitem size that one-period can broadcast, then adds this data subitem to broadcast queue, returns step 9.2.2.1 and circulate; Otherwise, exit circulation and enter step 9.2.3;

Step 9.2.3, initialization residue size;

Step 9.2.4, carries out following circulation,

Step 9.2.4.1, traversal prepares queue, is chosen at data subitem size and is less than maximum data subitem under the prerequisite of residue size, if found, enter step 9.2.4.2, if can not find such data subitem, then exiting circulation and exporting broadcast queue;

Step 9.2.4.2, adds the data subitem found to broadcast queue, returns step 9.2.4.1 and circulate.

And, adjust according to the off period of the adjustable strategies of off period by the data item of fractionation in step a4 or step b4, comprise following sub-step,

(1) data item d is inputted _igo on the air in kth time broadcast cycle period h data item r off period of asking in moment , data item d _idata subitem d _{i, 1}, d _{i, 2}..., d _i,P; Initializing variable , j=1;

(2) following circulation is carried out,

(2.1) setting data subitem d _i,joff period

$\mathrm{Deadline}=\mathrm{Deadline}+S_{d_{i,j}}/{\mathrm{Bw}}_k$

(2.2) if j equals data item d _itotal number P of the data subitem split, jumps out circulation and exports the data subitem adjusting the off period, otherwise j=j+1, return (2.1) and continue circulation.

And, add up solicited message in step a11 or step b11, comprise following operation,

(1) to read in successful request message file all solicited message to dynamic array vector;

(2) ID of the successful request information in dynamic array vector according to request is sorted;

(3) the successful request information that the ID that asks in dynamic array vector is identical is read successively, if judge that the data subitem sum of the number of successful request information and the data item of request is equal, the request then judging this ID is successful request, adds up the access time of this successful request; Otherwise this request is invalidation request;

(4) read the end of dynamic array vector, then exit, export successful number of request, total access time.

The present invention can data broadcast as required in real time, can receive the request of mobile client fully, according to real-time situation, broadcast real-time for data be gone out, the request of real-time response user.Take into account in real time the differing of data item size in data broadcast as required, propose and solve problem by the method for fractionation.Fractionation can make larger data item and smaller data item to take turns to turn to dispatch, and is unlikely to allow data item size interfere with the effect of data item scheduling.Comprise two kinds of fractionation schemes:

(1) to the present invention proposes etc. the scheduling that size splits, can be good at solving data item data broadcast scheduling as required in real time not of uniform size.It has taken the size in bandwidth sum cycle into consideration, with respect to the different distributions of average off period and the distribution situation of data item mean size, combines LxRxW dispatching algorithm, proposes ES-LxRxW scheduling.The results show, ES-LxRxW scheduling is effective, and when the average off period obeys different distributions, effect even can than not considering that some algorithm of data item size is better.

(2) the present invention proposes the scheduling that the size such as not splits, can not only consider cycle and bandwidth, can also consider the priority of data item size and scheduling in fractionation, is that the real-time status that take into account in the attribute of data item self and scheduling splits.Equally, it, with respect to the distribution of different data item mean sizes, combines LxRxW dispatching algorithm, proposes US-LxRxW scheduling.Proposing in literary composition and dispatch the data item not waiting size data item to split with two kinds of scheduling strategies, is PF strategy and PxBF strategy respectively.Experiment proves, no matter US-LxRxW scheduling is that employing PF strategy or PxBF are tactful, and effect is all better than ES-LxRxW scheduling, and request failure rate is lower than ES-LxRxW, and mean access time is lower than ES-LxRxW, and average fractionation number of times is lower than ES-LxRxW.And two kinds of strategies in US-LxRxW scheduling, in request failure rate with average to split on number of times both closely, but PxBF strategy is better than PF strategy on mean access time.

Accompanying drawing explanation

Fig. 1 is the system construction drawing of the broadcast of cycle data as required of the embodiment of the present invention.

Fig. 2 is the periodic broadcasting model schematic of the embodiment of the present invention.

Fig. 3 is the fractionation strategy schematic diagram waiting size data item of the embodiment of the present invention.

Embodiment

Technical solution of the present invention is described in detail below in conjunction with drawings and Examples.

1, method provided by the present invention increases on the basis of existing technology and achieves split process, is below described:

Each process can adopt software modularity technology to be embodied as corresponding module.The system configuration of the broadcast of cycle data as required that the embodiment of the present invention proposes as shown in Figure 1.Supposing the system comprises a server (Server) and multiple mobile client, and mobile client submits oneself request to by uploading channel (UpChannel), and server is by the queue of download channels (DownChannel) broadcast data item.The system of server mainly comprises reception, obtain, split and scheduling four modules, receiver module safeguards the request queue (ReqQueue) of the request composition of a system, scheduler module safeguards for preparing a queue (ReadyQueue), safeguard a broadcast queue (BroadcastQueue) simultaneously, prepare queue to be made up of data subitem, each data subitem comprises a request queue, the request queue of a data subitem is the queue that all requests of this data subitem are formed, can be designed to when therefore specifically implementing: prepare the concrete data structure of queue for (data item ID, data item size, split size, data subitem sum, data subitem sequence number, the request queue of this data subitem).The data subitem preparing queue adds broadcast queue after scheduling, can the concrete data structure of respective design broadcast queue be (data item ID, data item size, split size, data subitem sum, data subitem ID, go on the air the time, the request queue of this data subitem).ID represents sequence number.Because data item is split as data subitem, request also can be divided in each data subitem, i.e. all data subitems of request data item that it can be asked to ask.The request queue of a data subitem is the queue that all requests of this data subitem are formed.In all request queues, the data structure that embodiment designs the information of each request concrete is (ID of request, the data item ID asked, the time of request, the off period of request).

The operation principle of method provided by the present invention comprises following process:

(1) (Accept) is received.Mobile client is by uploading the request of channel submission oneself, and each request comprises the information such as the data item information of request, the off period of request.Server receives request, first checks this request whether in broadcast queue or prepare in queue, if, then direct request is added in broadcast queue or prepares in the request queue of corresponding data subitem in queue; If not, then request is added in the request queue of system by request time order from small to large, and be request dispatching ID.This process can complete primarily of receiver module.

(2) (Fetch) is obtained.Server takes out the request of current time from the request queue of system, from the buffer memory of internet (Internet), database (Database, DB) or server, search the data item meeting request, obtains data item and is organized into data item information.The present invention does not consider that server obtains the time of data and the cache policy of server.Data item information comprises the ID of data item and the content of data item.This process can complete primarily of acquisition module.

(3) (Split) is split.The data item got to be carried out splitting (etc. size split or etc. size does not split) according to fractionation strategy by server, and is added by the data subitem split into and prepare in queue.Such data item has one or several data subitem, and the data item ID of these data subitems is identical, but data subitem ID is different.Prepare queue first by data item ID arrangement, the data subitem of same data item ID arrange by data subitem ID, and request current for data item is kept in the request queue of each data subitem of fractionation gained.This process can complete primarily of fractionation module.

(4) (Schedule) is dispatched.Server periodically carries out scheduling tissue by dispatching algorithm to the data prepared in queue, dynamic selection data subitem adds in broadcast queue, broadcast queue is sent to satellite or base station by server, and the data subitem in broadcast queue is broadcasted away by download channels by satellite or base station.This process can complete primarily of scheduler module.

Split shown in Fig. 1 and scheduling two steps, point in order to wait the fractionation (EqualSplit) of size data item and scheduling and not wait the fractionation (UnEqualSplit) of size data item and dispatch.

Mobile client (MC) conduct and server independently equipment, according to the demand of oneself, constantly produce request and be sent to server by uploading channel, constantly monitors download channels afterwards and obtain the data item of oneself asking.

During scheduling, data item is split into data subitem by server; And mobile client is after have received data subitem, assembled by data subitem, be reduced into original data item, the order of assembling can be determined according to the numbering of data subitem.

2, the data broadcast model of the embodiment of the present invention is described:

(2.1) model agreement

For the ease of research, the present invention does following hypothesis and agreement for real-time data broadcast model as required:

(1) tentation data broadcast adopts periodic broadcasting scheduling mode as required, namely server according to mobile client RQ cycle choose according to dispatching algorithm the data item that will broadcast, and to broadcast away.

(2) data processing time agreement: suppose server process request, obtain ask data item, calculating priority level and generation broadcast queue time be all 0, namely the broadcast of data item can not have interval in time because of the processing time, is all broadcast continuously.

(3) satisfiability of mobile client: suppose the data item that server broadcast mobile client is asked, the mobile client of all this data item of request all can be satisfied, the replacement problem of data item during ignoring broadcast.

(4) existence of data item: suppose that all data item of mobile client application are all exist on the server, i.e. the data item of all mobile client applications is all in the database of server, as long as dispatching algorithm goes to get.

(5) the data item identifiability broadcasted: namely mobile client can know according to identification code the data item whether it will access itself before any one data item of reception.

(6) the data item global uniqueness of broadcasting: each data item namely broadcasted has a keyword, it can identify a broadcast item uniquely, and mobile client visits the data item in data broadcast according to this keyword, this is consistent with the identifiability of the data item of broadcast.

(7) data item broadcasted position in broadcast queue is determined: define the side-play amount from certain initial time when the position of data item in data broadcast is this data item being broadcasted in broadcast channel.All data item are all that order is broadcasted, and there is not multichannel or concurrent-broadcast situation.

(8) the consistency hypothesis of data: all data item can only wait until that all data subitems of this data item are preparing could to upgrade and revise when not existing in queue, the data subitem that each request of each mobile device receives be same version data item split by.

(9) forms data request agreement: each mobile client only asks a data item, the repeatedly onrelevant between requested data item of mobile client in once asking, and user can send next request on there is no before a request msg.

(10) mobile client independence that broadcast data is accessed: the data item of all broadcast on server is visible to all mobile clients, and all mobile clients are all separate, non-interfering to the access of data item.

Because data item is split as data subitem by the present invention, below broadcast item and prepare item referring to broadcast queue and the data subitem prepared in ordered series of numbers respectively.

(2.2) off period model

In concrete enforcement, when the present invention assesses dispatching algorithm performance during the off period, specific implementation can be selected as required by those skilled in the art.Such as conventional three kinds different off period distributions, are namely evenly distributed, exponential distribution and stationary distribution.

DeadLine=ArrivalTime+Uniform (0,2*MeanSlack) (formula 1)

DeadLine=ArrivalTime+Exponential (0,2*MeanSlack) (formula 2)

DeadLine=ArrivalTime+Fixed (MeanSlack) (formula 3)

Wherein Arrivaltime is the time of advent of request, and MeanSlack is skew of average off period primary quantity, does not comprise the processing time of request.Formula 1 for being uniformly distributed the lower off period, Uniform for being uniformly distributed random function, stochastic generation one (0,2*MeanSlack] between value.Formula 2 is the exponential distribution off period, and wherein Exponential is exponential distribution random function, stochastic generation one (0,2*MeanSlack] between value.Formula 3 is the stationary distribution off period, and wherein, Fixed is stationary distribution random function, and steady state value is MeanSlack.

(2.3) the large mini Mod of data item

In concrete enforcement, when the present invention assesses dispatching algorithm performance, concrete data item size can be selected as required to calculate implementation by those skilled in the art.Such as conventional three kinds different data item size distribution, are namely uniformly distributed, exponential distribution and stationary distribution.

DataSize=Uniform (0,2*MeanSize) (formula 4)

DataSize=Exponential (0,2*MeanSize) (formula 5)

DataSize=Fixed (MeanSize) (formula 6)

Wherein, MeanSize is the skew primary quantity of data item mean size.Formula 4 is the data item size under being uniformly distributed, and Uniform is equally distributed random function, stochastic generation one (0,2*MeanSize] between value.Formula 5 is the data item size under exponential distribution, Exponential exponential distribution random function, stochastic generation one (0,2*MeanSize] between value.Formula 6 is the data item size of stationary distribution, and wherein Fixed is the random function of stationary distribution, and steady state value is MeanSize, and namely data item size is fixing special circumstances.

(2.4) broadcast model definition

Figure 2 shows that the broadcast scheduling cycle after data item fractionation, if carried out kth-1 scheduling, kth time scheduling, kth+1 scheduling ..., the broadcast cycle of time broadcast scheduling of kth wherein, 1,2,3,4 ..., h ..., M is a corresponding data item respectively.For studying conveniently, the present invention does following hypothesis and definition, as shown in table 1.

The hypothesis of table 1 model and definition

3, appraisement system

In real time as required in Radio Data System, the quality of the stand-by period measure algorithm should only do not considered with traditional algorithm, and when the meeting consumers' demand in real time of algorithm, the service quality of user should be taken into account.And the real-time demand meeting user, be not namely the data item that simple broadcasting user is asked, the off period of user's request must be met.Meanwhile, the efficiency of algorithm is also the response time being directly connected to broadcast, and therefore the efficiency of algorithm also should be considered.

The present invention evaluates from request failure rate, mean access time and average number of times three aspects that split dispatching algorithm.Request failure rate is used to the situation that is satisfied evaluating user's request, and the access time is used to the service quality reflecting user, and the average number of times that splits is used to evaluate the efficiency that splits.

(1) failure rate (LossRate, LR) is asked.

Request failure rate refers to the ratio that the request of inefficacy is shared in all requests can reflect the validity of scheduling, the situation about being satisfied of namely all requests.Wherein, total number of request refers to the number of the request of total inefficacy and the number sum of total successful request.

If in the processing procedure of server, off period of certain request to but the data item of its request be not broadcasted, so this data item broadcast is lost efficacy to this request.On the contrary, if data item can be broadcasted before the off period of request, so this request is effective.After data item splits into data subitem, for each mobile client, only have him to broadcast by all data subitems of requested data item, this request could be satisfied, and namely this request is successful request; On the other hand, if the off period of any one data subitem wherein of his request msg misses, so the broadcast of this data item just fails, and namely this request is invalidation request.

(2) mean access time (AverageAccessTime, AAT).

Mean access time refers to the mean value of the access time of all successful request, and namely the summation of the access time of all successful request is divided by total successful request number.Access time refers to the time receiving data and continue between this from mobile client issues a request to mobile client.

(3) number of times (AverageSplitTimes, AST) is on average split.

The present invention it is considered that to data item etc. the fractionation of size and the fractionation of the size such as not, will take server CPU during fractionation, the number of times therefore split is more, and server spends the time on splitting can be more, and the efficiency of fractionation can be lower.Therefore the present invention's fractionation number of times carrys out the speed of service of measure algorithm, i.e. the efficiency of algorithm.The fractionation number of times of data item is defined as the number of the data subitem that each data item splits into by the present invention, so on average splits the result of total fractionation number of times divided by total number of the data item split that number of times is data item.

4, to embodiments provides etc. size data item split and not etc. size data item to split two kinds of modes selective, the specific implementation that reciprocity size data item splits below is described.

(4.1) fractionation that splits of size data item and the scheduling principle such as

In actual applications, the data item size of user's request is not of uniform size, has larger data item, also has smaller data item.If the algorithm before just simply applying mechanically, and data item do not split and process, following inconvenience may be with.(1) bandwidth is inadequate.Simultaneously due to technical equipment or wireless transmission medium, bandwidth in different environments size is different, and bandwidth in a lot of situation therefore may be caused in practice to be inadequate.When bandwidth is inadequate, if cycle size is fixing, then large data item may be caused cannot not broadcast completely in one-period.(2) efficiency is not high.In data broadcast before, each data item only has and could serve for other mobile subscribers after the broadcast is completed.When bandwidth is certain, if broadcast large data item, so the time of this data item busy channel is long, and the number of users of service in the unit interval is reduced, and service quality reduces.Equally, if data item size is certain, when bandwidth is smaller, the time of data item busy channel is also by long.(3) not be applicable to very much for periodic broadcasting.Periodic broadcasting is that server is periodically chosen data item to broadcast, more extreme situation, when the bandwidth sum cycle is all determined, if certain data item size has exceeded in a broadcast cycle the maximum data item size that can broadcast, so this data item then can not have been broadcasted within this cycle, and this has just runed counter to the theory of periodic broadcasting.

The present invention proposes the data subitem of the sizes such as data item splits into.After the size such as data item is split, relative to the fractionation of the size such as not, based on what time considering as follows.(1) mobile client conveniently can predict the time that next data item arrives.In practice, when amount of bandwidth is certain, if data item such as is at the size, then mobile client is when intercepting download channels, can calculate the arrival time of next data item.When data item arrives, first judge the data of whether oneself asking, if so, then can continue to intercept, and obtain the data oneself needed; If not, then can be switched to park mode, when next data item arrives, then be switched to enable mode and intercept channel, and continue to intercept.Therefore, etc. data item size time, the present invention just well can predict the time that next data item arrives according to existing condition, thus saves the energy of mobile client.(2) can be good at using before achievement.The algorithm of data broadcast scheduling as required in real time is before all study when sizes such as data item, and has occurred some good algorithms, as EDF, MRF, RxW, SIN-α, LxRxW of the prior art.If by sizes such as data item split into, then the data broadcast dispatching algorithm before can well using for reference, dispatches data item.

(4.2) fractionation that splits of size data item and the scheduling flow such as

The fractionation split Deng size data item and scheduling are on the basis of real-time digital broadcasting dispatch system as required, first data item are carried out etc. to the fractionation of size, then carry out dispatching and broadcasting.When splitting etc. size, pre-set SplitK value, and then according to etc. size data item split fractionation strategy split.

The fractionation split Deng size data item and scheduling (EqualSplitandLxRxW, ES-LxRxW), concrete flow process is as process ES-LxRxW.

Process: ES-LxRxW

(1) initialization, generally comprises request queue, prepares the initial work such as queue, broadcast queue, time point, when specifically implementing, can design concrete operations by those skilled in the art;

(2) if current time is not asked, then jump to (7), otherwise continue to enter (3);

(3) this request is taken out, i.e. the request of current time;

(4) obtain the data item of this request, according to waiting fractionation strategy of size data item to be split, the off period of the data item of fractionation adjusts by the adjustable strategies according to the off period;

(5) for all data subitems after splitting, process respectively: if at the broadcast item of broadcast queue and no broadcast, then this request is added in the request queue of this broadcast item; Otherwise, if prepare the preparation item in queue, then this request is added in the request queue of this preparation item; Otherwise, generate new preparation item and add preparation queue (add by this data subitem and prepare queue), and request is added in the request queue preparing this new preparation item in queue;

(6) request newly to be generated such as; Jump to (2);

(7) broadcast according to broadcast queue;

(8) not empty broadcast item is judged whether just to have broadcasted, if otherwise directly enter (9), if just broadcasted not empty broadcast item, then the successful request information of this broadcast item is write successful solicited message file (solicited message file can be successfully established in advance), delete all requests of this broadcast item, delete this broadcast item, then enter (9);

(9) judge whether that broadcast queue is empty, if otherwise directly enter (10), if broadcast queue be empty, then according to etc. the dispatching algorithm of size fractionation data item generate new broadcast queue, then enter (10);

(10) advance a time point, if exceeded the running time of regulation, then jumps to (11); Otherwise, redirect (2);

(11) successful request message file is read, statistics solicited message (tactful SSRM).

Have the fractionation of data item in the present invention, dispatch for data subitem, therefore when adding request, each data subitem can add request to, therefore in system loss of timing with when successfully asking, simply can not calculate by method before.

For invalidation request, miss when the off period of certain data subitem, then the request of this ID was lost efficacy.The present invention, for the request of losing efficacy, dispatches during scheduling, while add up, and when encountering the request of inefficacy, namely the off period of this request misses, and can delete this solicited message, finds simultaneously and deletes the request of same request ID in the preparation item preparing in queue other.In reality, when certain request of certain data subitem misses the off period, namely the off period of this request of this data subitem is less than current time, so can infer, the data subitem that the sub-sequence number of the data item that this request is asked is less than the sub-sequence number of current data subitem is broadcasted successfully, otherwise the off period of this request early will be missed.Therefore, as long as delete the request with request ID in the sub-sequence number data subitem larger than the sub-sequence number of current data item here.

During statistics successful request information, whether the data subitem that judge each request successful broadcast within its off period, if all successful broadcast, then this request is successful request; If any one data subitem misses the off period, then this request is still invalidation request.

Strategy (the StatisticSuccessRequestMessagestrategy of statistics successful request information, SSRM), this strategy only statistics number of successful request and summation of access time, because invalidation request is just being added up, so the request of inefficacy all neglected in this strategy before.

Vector represents container or dynamic array, and be the Essential Terms in existing C++ STL, the embodiment of the present invention adopts dynamic array.Sum for the data subitem of the number of solicited message and the data item of request is equal situation, namely the request of this ID is the situation of successful request, when adding up its access time, the present invention takes the time of first data subitem that family receives for its time of reception, instead of subitem sequence number is the time of reception of the data subitem of 1.Because, can be known by explanation above, data subitem is not necessarily broadcasted from small to large according to its subitem sequence number, and to be user calculate by this time point starting receiving data item the access time being related to time of reception, therefore, the timing definition of first data subitem that the present invention receives mobile client is the time of mobile client receiving data item, and the access time of mobile client is duration between request time to time of reception.

(4.3) the fractionation strategy of size data item such as

Cutting and packaging Problems (CuttingandPacking) are classical problems, and it is the combinatorial problem of difficulty, under the background of much reality, have application, many generally in industry and service industry's appearance.In general, it is split into " small articles " " large objects ", in this way waste minimized.Problems more typically has: the cutting paper logs of paper industry become the large plank of the cutting of narrow volume, furniture industry be divided into less rectangular slab, in fashion industry from the irregular assembly of fabric warrping cutting profile etc.

The present invention uses the thought of cutting and packaging Problems, and adopt the mode split will be assembled into one-period and broadcast away, the data of fractionation after receiving data, are assembled into original data item by mobile client again.

For the fractionation of data item, the present invention such as to be divided at the fractionation of size data item and the fractionation of the size data item such as not, and this chapter is by the fractionation of size data items such as first introducing.

Deng the fractionation of size data item, be the data subitem all data item being split into equal size, then carry out schedule broadcast.As long as namely define a size split, then all data item all will be split into the equirotal data subitem of fractionation of this definition.

The present invention defines the size of fractionation for splitting benchmark size (SplitStandardSize), waits the fractionation of size data item all to split into the data subitem splitting benchmark size by all data item.Split benchmark size because data item size may be less than, therefore the present invention defines the data item size that the filling size of data subitem is data subitem actual load.Concrete waits the fractionation strategy (EqualSplit, ES) of size data item as follows.

As shown in Figure 3, if it is 3M that definition splits benchmark, so data item size is the data item d of 8M _i3 data subitems will be split into: data subitem d _{i, 1}, data subitem d _{i, 2}, data subitem d _{i, 3}, the sub-item size of data is all 3M.Wherein, data subitem d _{i, 1}all filled by data item, data subitem d _{i, 2}all filled by data item, and data subitem d _{i, 3}only be filled with the data item of 2M, also have 1M be in order to keep data item etc. size, and reserved out do not fill up region, data item size is 3M, and filling size is 2M.For reserved region out, it can not play effect for request, but it can take the bandwidth of broadcast.

In order to study conveniently, the present invention does following hypothesis for waiting fractionation strategy of size data item and illustrates:

(1) the data subitem after data item fractionation does not need according to sequencing broadcast, i.e. the data subitem broadcast order not necessarily d of data item di _{i, 1}, d _{i, 2}, d _{i, 3}.

(2) for each request, only have this request within the off period, receive all data subitems and be only successful request.If any one data subitem has wherein exceeded the off period, then this request had been lost efficacy.

(3) the invalidation request number of data item and successful request number are all for original data item, the invalidation request number of data subitem and the not double counting of successful request number.

(4) deadline of data subitem broadcast is decided by the data item of its actual filling, i.e. data subitem d _{i, 3}the broadcast deadline refer to broadcast deadline of the data item of the 2M of filling.

(4.4) adjustable strategies of off period

After data item splits, according to before to the definition of off period, as long as data item is gone on the air, the off period of then data item just meets.Data item d as mentioned above _isplit into data subitem d _{i, 1}, d _{i, 2}, d _{i, 3}, according to definition before, as long as data item d _igo on the air, i.e. data subitem d _{i, 1}go on the air, data item off period just meet and request.But, if server is broadcasting data subitem d _{i, 1}afterwards, and always not broadcast data subitem d _{i, 2}, d _{i, 3}, so this request is just slowly not being met.Obviously, if the time waited for is excessively of a specified duration, it is off period that can exceed request.Therefore, after data item splits into data subitem, the off period will change through corresponding.Here, the present invention will be divided in corresponding data subitem the off period.

In the fixed-size situation of data item, when bandwidth is given, the off period of request determines (deadline of the data item of certain request of going on the air), and the off period (completing the deadline of these data of broadcast) of so having asked is determined.Same, when bandwidth is given, when waiting size to split, the airtime of data subitem is determined, the time that so data item broadcast completes is determined.If determine the off period of request, then the request that determines completes the off period.According to request off period deadline, can push away broadcast this request complete between off period finishline meet pass be .Request according to current data subitem completes the off period, just can push away to obtain request off period of next data subitem.

Second off period to last data subitem is defined the new off period, i.e. second data subitem d according to the time of transmission by the present invention _{i, 2}off period be first data subitem d _{i, 1}off period add time of first data subitem broadcast, the 3rd data subitem d _{i, 3}off period be second data subitem d _{i, 2}off period add time of second data subitem broadcast, the rest may be inferred.Concrete off period adjustable strategies (DeadlineAdjust, DA) is as follows:

According to the setting rule of table 1, represent data item d _igo on the air in kth time broadcast cycle period h data item r off period of asking in moment.Illustrate: if data item d _iany one data subitem exceeded the off period and be not broadcasted, then data item d _ilost efficacy, and all data item d broadcasting He do not broadcast _idata subitem all lost efficacy.

(4.5) scheduling strategy that splits of size data item such as

In the scheduling of real-time on-demand broadcasting, in the broadcast scheduling algorithm not considering data item size, be all that the data item size of hypothesis is fixing situation, the model after splitting with the size such as of the present invention is identical.Do not consider that data item size has following several typical dispatching algorithm.Do not split data subitem in these algorithms, compare for the purpose of optimization for convenience, data item is designated as d _i,j.

(1) off period is used for representing the urgency of data item by EDF, gets off period data item the earliest and broadcasts at first.Tentation data item is d _i,j, a current kth broadcast cycle dispatches h data item, d _i,jpriority calculating formula is .Be worth less data item priority higher.The shortcoming of algorithm only considered off period situation, the few data item of the request data item more than request may be caused preferentially to broadcast, cause failure rate low not.

(2) MRF is using number of request as the urgency of data item and importance, gets the data item having largest request number and broadcasts at first.Tentation data item is d _i,j, a current kth broadcast cycle dispatches h data item, d _i,jpriority computing formula is .Be worth less data item priority higher.The shortcoming of algorithm is the request situation that only considered data item, if be applied in real-time on-demand broadcasting, to the not clear sense of the urgency of data item request, therefore failure rate may be very large.

(3) RxW is using number of request and the stand-by period urgency as data item, gets to have number of request and the maximum data item of maximum wait time product is broadcasted at first.Tentation data item is d _i,j, a current kth broadcast cycle dispatches h data item, d _i,jpriority calculating formula is .Be worth less data item priority higher.Although the shortcoming of algorithm to reduce mean access time and average tuning period, can not for real-time on-demand broadcasting problem, the same with MRF algorithm, failure rate may be very large.

(4) SIN-α is using the urgency of SIN value (the residue slack time of request to be solved) as data item, gets the data item having minimum SIN value and broadcasts at first.Tentation data item is d _i,j, a current kth broadcast cycle dispatches h data item, d _i,jpriority calculating formula is .Be worth less data item priority higher, in algorithm, α is the weight weighing off period and number of request, and during experiment display α=1, the effect of calculating formula is close with optimal situation.The shortcoming of algorithm reflects the overall urgency of data item with the off period, perfect not, and do not consider stand-by period situation.

(5) LxRxW is using by failure number, number of request and the stand-by period urgency as data item, and the data item of getting by the achievement of invalidation request number, number of request and maximum wait time three is maximum is broadcasted at first.Tentation data item is d _i,j, a current kth broadcast cycle dispatches h data item, d _i,jpriority calculating formula is , it is higher to be worth less data item priority.The algorithm urgency reflecting the off period by failure number, well combines number of request and maximum wait time again, reflects the urgency of data item entirety.

LxRxW dispatching algorithm combine data item by invalidation request number, current request number and the maximum wait time main policies factor as scheduling, wherein more can the overall urgency of representative data item than the off period of data item by invalidation request number.

In order to the priority of quantitatively evaluating data item, in each broadcast cycle, before determining the data item of each medium wave, data item all can calculate a LxRxW value.Kth time broadcast cycle, during h medium wave, data item d _i,jlxRxW value computing formula as shown in Equation 7.

$\mathrm{Weight}\left(d_{i,j}\right)=\frac{1}{L_{d_{i,j},k,h}\×R_{d_{i,j},k,h}\×W_{d_{i,j},k,h,r}}$ (formula 7)

Wherein, the data item priority with minimum Weight value is the highest, is broadcasted at first.But in Weight value calculates may be 0, now just use represent data item d _i,jurgency, be therefore revised as formula 8

$\mathrm{Weight}\left(d_{i,j}\right)=\{\begin{array}{c}\frac{1}{L_{d_{i,j},k,h}\×R_{d_{i,j},k,h}\×W_{d_{i,j},k,h,r}}\mathrm{if}(L_{d_{i,j},k,h}\≠0)\\ \frac{{\mathrm{StartD}}_{d_{i,j},k,h,r}}{R_{d_{i,j},k,h}}\mathrm{if}(L_{d_{i,j},k,h}=0)\end{array}$ (formula 8)

For the sake of ease of implementation, the present invention well combines waiting fractionation of size data item with LxRxW dispatching algorithm, first carries out broadcast scheduling by LxRxW again according to ES fractionation, and concrete dispatching algorithm is as process Schedule, but wherein d _i,jrepresent and split the data obtained subitem.

(4.6) the value strategy of benchmark size is split

Deng the fractionation of size data item, the number of fractionation is decided by fractionation benchmark size SplitStandardSize completely.And split benchmark size, should consider in conjunction with the actual conditions of amount of bandwidth and cycle size.

(1) amount of bandwidth (Bandwidth).Bandwidth refers to server transmitted per unit time data item size.Bandwidth is larger, and in the unit interval, the data item of broadcast is more, and the throughput of server is larger, and the time of onesize data item broadcast is less, and the load of other condition one timing systems diminishes.

(2) cycle size (Cycle).Cycle refers to the time size that server is once broadcasted.Cycle is larger, and the data item broadcasted when bandwidth is certain is more.

The size that data item splits should be relevant to the data item size that each cycle can broadcast, namely with the product (Bw in bandwidth sum cycle _k× Cy _k) size be correlated with.And in one-period, the data item number of broadcast should be integer, so the fractionation benchmark size of data item should be can be divided exactly by the product in bandwidth and cycle.Because the data item size of broadcast in one-period, if not an integral multiple data item size, so can there is waste in bandwidth, also can waste the speech amplifier meeting of data item accordingly, thus make failure rate increase.Therefore, the size that data item splits should be Bw _k× Cy _k/ SplitK, wherein SplitK is a positive integer, represents that the data items size of broadcast in one-period is the number of the data subitem splitting benchmark size SplitStandardSize.

$\mathrm{SplitS}\tan \mathrm{dardSize}=\frac{{\mathrm{Bw}}_k\×{\mathrm{Cy}}_k}{\mathrm{SplitK}}(\mathrm{SplitK}\∈N^{+})$ (formula 9)

Wherein, N ⁺represent positive integer, for the problem that the size such as data item splits, data item splits into much portions and is only best actually, only has and could illustrate by experiment.And for fractionation scheme, need an evaluation criterion, evaluate the quality that this splits scheme.

Because the present invention considers from service quality, meet user, the best demand meeting user, so the present invention evaluates from request failure rate and mean access time two aspects dispatching algorithm namely at most.In addition, because the fractionation of data item will occupy the regular hour, the efficiency therefore split also is a key point that should be noted that.The present invention adopts the average number of times that splits to weigh the efficiency of fractionation, and average fractionation number of times is less, and the number of times of fractionation is fewer, and the efficiency of dispatching algorithm is higher.

Concrete enforcement can obtain the selected value roughly of SplitK from experimental result, the summary sheet of the suggestion value of SplitK value when table 2 is various combinations of off period distribution and data item size distribution.

The best SplitK of table 2 gathers

5, in the specific implementation that not etc. size data item does not split, except the fractionation strategy of size and the scheduling strategy of data item such as not, other split the same with item such as size data such as grade, the adjustable strategies of such as off period.Below the fractionation strategy of size and the scheduling strategy of data item such as are not specifically described:

(5.1) fractionation not waiting size data item to split and scheduling principle

Although split the achievement before can be good at predicting the airtime of next data item and well applying etc. the data item of size, the scheduling waiting size data item to split has some deficiency following.

(1) utilance of bandwidth is not taken into full account.Deng the fractionation of size data item, be not filled owing to there being a lot of data subitem, this waste all can had living space after making the fractionation of most of data item, thus the bandwidth of server is not utilized completely, thus the stand-by period of a part of user can become large relatively, the service quality of server can corresponding step-down.

(2) server efficiency is low.When data item size is larger, the number of times of fractionation is many, and to split data item be the CPU needing to take server, and therefore server can cause efficiency very low because of the fractionation of data item.

(3) impact of data item size on algorithm is not considered.If data item size is too little, the phenomenon of a lot of space waste may be there is, the utilance of bandwidth above-mentioned may be caused very low.On the contrary, if data item is too large, then the efficiency step-down of server may be made.Therefore, a better method should be chosen, make both compromise.

(4) the data item urgent impact on splitting in time is not considered.Temporal urgency is directly correlated with service quality, when therefore splitting, it also should be taken into account.

This chapter proposes data subitem data item being split into the size such as not.Relative to the fractionation waiting size data item, the fractionation of size data item is not waited to have following advantage.

(1) can be good at the utilance of the bandwidth in the lifting cycle.To the fractionation of the size such as or not data item, the bandwidth that can take into account each cycle can make full use of.

(2) can be good at considering that data item size is on the impact of algorithm.The data item that data item size is larger, if broadcasted always, the bandwidth resources taken are many, and namely in bandwidth one timing, the time of broadcast is long, and the number of users of unit interval service reduces.Therefore, the data item that data item size is larger, repeatedly should split, repeatedly broadcasts.

(3) can adjust the real-time condition of data item dynamically.During fractionation, in conjunction with the dispatching priority of data item, the real-time by invalidation request number, number of request and maximum wait time of data item can be considered, make fractionation directly can follow service quality correlation.

(5.2) fractionation not waiting size data item to split and scheduling flow

Do not wait the fractionation of size data item, as the term suggests split according to different fractionation benchmark by different data item, the fractionation benchmark size defined in also namely waiting size data item to split can be different according to different data item.

For the equirotal data item of data item and etc. size split data item, data item itself is of a size, and be just in time the complete several data item of broadcast in the cycle, therefore the situation before is all worth the priority considering data item, and the factor of data item size is not taken into account when dispatching.And for the data item that the size such as not splits, the data subitem size that each data item splits into not etc., therefore, when dispatching, not only will not considered the priority of data item, also must consider data item size.

The present invention is divided into two parts by not waiting scheduling of size data item, namely first after getting data item, data item is carried out to the fractionation of the size such as not, and is added to by the data subitem split into and prepare in queue.Afterwards, at scheduling phase, the data item split is carried out schedule broadcast.

The scheduling not waiting size data item to split is on the basis of digital broadcasting dispatch system as required in real time, first data item is carried out to the fractionation of the size such as not, then carries out dispatching and broadcasting.Do not wait the fractionation of size data item, the data item of size such as following split similar, is also first define one to split benchmark size.Unlike, the benchmark split Deng size data item calculates according to SplitK value at the very start, but do not wait the fractionation benchmark size of size data item can the difference of factor data item and different, the fractionation benchmark size of each data item calculates in real time before fractionation.

The fractionation not waiting size data item to split and scheduling (UnequalSplitandLxRxW, US-LxRxW), concrete flow process and the fractionation waiting size data item to split and scheduling flow similar, as process US-LxRxW.

Process: US-LxRxW

(1) initialization, generally comprise request queue, prepare the initial work such as queue, broadcast queue, time point, when specifically implementing, can design idiographic flow by those skilled in the art operates; ;

(2) if current time is not asked, then jump to (7), otherwise continue to enter (3);

(3) this request is taken out;

(4) obtain the data item of request, according to not waiting fractionation strategy of size data item to be split, the off period of the data item of fractionation adjusts by the adjustable strategies according to the off period, and adjustable strategies is as DA;

(5) for all data subitems after splitting, if in broadcast queue and no broadcast, then this request is added in the request queue of this broadcast item; Otherwise, if in preparation queue, then this request is added in the request queue of this preparation item; Otherwise, generate new item and add preparation queue to, and request is added in the request queue preparing this new item in queue;

(6) new request is generated and the queue that joins request; Jump to (2);

(7) broadcast;

(8) not empty broadcast item is judged whether just broadcast, if otherwise directly enter (9), not empty broadcast item has been broadcasted if firm, then by the successful request information writing in files of this broadcast item, delete all requests of this broadcast item, delete this broadcast item, then enter (9);

(9) judge whether that broadcast queue is empty, if otherwise directly enter (10), if broadcast queue be empty, then generate new broadcast queue according to the dispatching algorithm of the size fractionation data item such as not, then enter (10);

(10) advance a time point, if exceeded the running time of regulation, then jumps to (11); Otherwise, jump to (2);

(11) successful request message file is read, statistics solicited message.Statistics solicited message specific implementation is shown in tactful SSRM.

(5.3) the fractionation strategy not waiting size data item to split

Data item size is carried out to the fractionation of the size such as not, the most directly evaluating is exactly the service quality at every turn broadcasted.And ask failure rate and mean access time to be the standard evaluating service quality.

(1) cycle size and amount of bandwidth.The data item size that the product representation one-period of cycle and bandwidth can be broadcasted.Seizing the opportunity of cycle and bandwidth is larger, the data item size that one-period can be broadcasted is larger, and the throughput of system is larger, and the size that data item splits can be larger, thus while the request failure rate ensureing data item, make the fractionation number of times of data item few as much as possible.

(2) data item size.Here the data item size do not split referred to, it is the attribute of data item self, and splitting scheduling should consider.Different data item, varies in size, and the size of fractionation also should be different.Data item size is larger, and the fractionation number of times that data item is possible is more, and the failure rate of data item can increase, and the efficiency of server can reduce.

(3) time urgency degree.The urgent degree of time is directly connected to total failure rate and mean access time, and it should be directly reflected into and split in benchmark size.The dispatching priority that the present invention considers adopts LxRxW algorithm, and it has considered by invalidation request number, number of request and maximum wait time three.Dispatching priority size is larger, and namely the value of LxRxW is larger, represent the off period closer to, data item should preferentially be broadcasted, and therefore this data item can not be torn open too little.Because if that tears open is too little, then the data subitem split into can be a lot, thus failure rate is increased, and the efficiency of server reduces.

Identical with waiting fractionation of size data item, the present invention defines and splits benchmark size is SplitStandardSize.During fractionation, first calculate real-time SplitStandardSize, then data item is split into data subitem.

(1) by the analysis splitting the factor that size is correlated with, data item d is found _idata item size the data item size can broadcasted with one-period is larger, namely be worth larger, the pressure of broadcast to bandwidth of data item is larger, and the unit interval number of users of service is fewer, and the size namely split should be less, and the number of users of serving just can be made to obtain equilibrium.Namely size with fractionation size be inversely proportional to.

(2) dispatching priority of data item value larger, represent that off period of data item is more close, number of request is more, the stand-by period is longer, data item more should preferentially be broadcasted, and therefore the fractionation size of data item should be larger.Namely size with fractionation size be directly proportional.

Therefore, be formula 10 according to 2 formula that can obtain the fractionation size of data item above.

$\mathrm{SplitS}\tan \mathrm{dardSize}=\mathrm{\α}\×\frac{{\mathrm{Bw}}_k\×{\mathrm{Cy}}_k\×L_{d_{i,j},k,h}\×R_{d_{i,j},k,h}\×W_{d_{i,j},k,h,r}}{S_{d_i}}$ (formula 10)

Wherein, the α in formula is parameter to be set, and when specifically implementing, value can be selected by the simulation experiment result and preset.Above formula can be modified to formula 11 by the amendment type of LxRxW in formula 8.

$\mathrm{SplitS}\tan \mathrm{dardSize}=\frac{\mathrm{\α}\×{\mathrm{Bw}}_k\×{\mathrm{Cy}}_k}{S_{d_i}\×\mathrm{Weight}\left(d_{i,j}\right)}$ (formula 11)

What split benchmark definition is the size that split of each data item.So in reality, how each block should split, and specifically splits much, present invention employs following strategy.

The fractionation strategy (UnequalsizeSplit, US) of size such as not.According to the computing formula of the fractionation benchmark size of size such as not, each data item is according to its actual conditions, calculate after splitting benchmark size, the data subitem splitting benchmark size is constantly split out from data item, until data item size is less than split benchmark size, using the data item that remains separately as a data subitem, data item size is residue size.Concrete does not wait the fractionation strategy of size data item as follows.

For the request in real system, after obtaining the data item of request, first judge the data item of asking whether in broadcast queue or prepare in queue, if, then take out the fractionation size of this data item, if do not existed, then recalculate weight value, and calculate the SplitStandardSize of this data item, by it according to this standard resolution, and upgrade the off period of each data subitem.

Each data subitem of data item is judged successively: judge this data subitem whether in broadcast queue, if, then this request is added its request queue; Otherwise judge whether this data subitem is preparing in queue, if, then this request is added in its request queue; Otherwise add this data subitem as new item to preparation queue.Like this until all data subitems of this data item all judge that complete this data item that namely completes obtains, splits and adds.

Table 3 does not wait size data item to split the example of strategy

As shown in table 3 is an example not waiting size data item fractionation strategy, d ₁fractionation benchmark size be 3M, then size is the d of 8M ₁split into the d that size is 3M _1,1, size is the d of 3M _1,2be the d of 2M with size _1,3three data subitems; d ₂fractionation benchmark size be 3.3M, then size is the d of 5M ₂split into the d that size is 3.3M _2,1be the d of 1.7M with size _2,2two data subitems.

(5.4) scheduling strategy not waiting size data item to split

In the periodic broadcasting scheduling considering data item size, data item size differs, if do not adopt the size data item that waits described in last chapter to split, there will be in each cycle when dispatching has bandwidth waste.Therefore, the present invention is when dispatching, can be selected as required to adopt concrete scheduling strategy by those skilled in the art, such as these two kinds different scheduling strategies: the first is the scheduling strategy that preferential broadcast meets the highest data item of dispatching priority LxRxW, and the second is the scheduling strategy considering bandwidth availability ratio and dispatching priority.

(1) scheduling strategy (PriorityFirst, PF) that priority is maximum.Dispatching priority representative be the urgency of time.Each cycle is when starting to dispatch, server is constantly chosen the data subitem preparing to have in queue dispatching priority LxRxW and is joined broadcast queue, until the sub-item size summation of data has exceeded the data item size that this cycle can broadcast and just stops in broadcast queue.The data subitem that the target of this strategy to be data subitem in the broadcast queue making at every turn to choose be all priority is maximum, thus meet the request of more user as much as possible, make service quality higher.

(2) scheduling strategy that priority is maximum can data item that preferentially broadcast scheduling time urgency is the highest, but has the waste of bandwidth.And if make bandwidth availability ratio maximum just merely, although take full advantage of bandwidth, do not consider completely for the time factor being directly connected to scheduling validity, therefore it is not an effective strategy.The priority of scheduling and the utilance of bandwidth consider by the present invention, propose the scheduling strategy PxBF(PriorityandBandwidthFirst, the PxBF that consider priority and bandwidth availability ratio).

The present invention uses for reference the thought of 0-1 knapsack algorithm, the dispatching priority (LxRxW) of data item is mapped as the value of article, data item size (amount of bandwidth namely taken) is mapped as the weight of article, the data item size that one-period can be broadcasted is the total weight that knapsack can bear, target is under total weight is no more than the prerequisite of the total weight that knapsack can bear, and makes the total value of got data item maximum.When to fetch data, to each data item d _i,jonly have 2 kinds of selections, namely load knapsack (1) or do not load knapsack (0).Can not by data item d _i,jload knapsack repeatedly, will can not only load the data item d of part _i,j.The typical algorithm solving 0-1 knapsack problem has dynamic programming algorithm and greedy algorithm.It is the optimal solution solving 0-1 knapsack problem that Dynamic Programming solves scheme, and it represents whether choosing of article with two-dimensional array.And the scale of data item of the present invention is larger, if solved with Dynamic Programming, then space complexity is very high, and therefore the present invention takes greedy algorithm.Calculate the dispatching priority of each data item and the ratio of data item size, then press the descending sort of this ratio, from Section 1, fill knapsack, be then Section 2, the like, much more as much as possible to put, until fill knapsack.

Concrete algorithm, when each cycle starts to dispatch, chooses maximum data item adds broadcast queue to, until data item size summation has exceeded the data item size that this cycle can broadcast and just stops in broadcast queue.If there is remaining bandwidth, the present invention adopt bandwidth availability ratio maximum supply method, namely select the maximum data item that can broadcast.From formula 3-1, $L_{d_{i,j},k,h}\×R_{d_{i,j},k,h}\×W_{d_{i,j},k,h,r}/S_{d_{i,j}}=1/S_{d_i}\×\mathrm{Weight}\left(d_{i,j}\right)$ , namely choose minimum data item.Specific strategy is described below:

Gathering of optimum a-value can be obtained as shown in table 4 by experiment.

Table 4 optimum a-value gathers

Specific embodiment described in the present invention is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various amendment or supplement or adopt similar mode to substitute to described specific embodiment, but can't depart from spirit of the present invention or surmount the scope that appended claims defines.

Claims (8)

1. consider a real-time data broadcast dispatching method as required for data item size, it is characterized in that: comprise following process,

Mobile client submits the request of oneself to by uploading channel after, server receives request, first check this request whether in broadcast queue or prepare in queue, if, then direct request is added in broadcast queue or prepares in the request queue of corresponding data subitem in queue; If not, then request is added in the request queue of system by request time order from small to large, and be request dispatching ID;

Server takes out the request of current time from the request queue of system, searches the data item meeting request, and obtain data item and be organized into data item information, described data item information comprises the ID of data item and the content of data item;

The data item that the large young pathbreaker of server based on data item gets splits, and is added by the data subitem split in preparation queue; A data item splits into more than one data subitem, and the data item ID of these data subitems is identical, but data subitem ID is different; Prepare queue first to arrange by the ID of data item, the data subitem of same data item ID arranges by data subitem ID, and request current for data item is kept at the request queue splitting each data subitem of gained, the data subitem prepared in queue is called preparation item;

Server periodically carries out scheduling tissue by dispatching algorithm to the data prepared in queue, and dynamic data subitem of selecting adds in broadcast queue, and the data subitem in broadcast queue is broadcast, and the data subitem in broadcast queue is called broadcast item; Wherein,

When the data item got being split, employing waits size data item to split mode, waits the fractionation of size data item fractionation mode and scheduling flow to comprise the following steps,

Step a1, initialization;

Step a2, if current time is not asked, then jumps to step a7, otherwise continues to enter step a3;

Step a3, takes out the request of current time;

Step a4, obtains the data item of this request, and according to waiting fractionation strategy of size data item to be split, the off period of the data item of fractionation adjusts by the adjustable strategies according to the off period;

Step a5, process respectively for all data subitems after splitting, processing mode is as follows,

To wherein arbitrary data subitem, if at the broadcast item of broadcast queue and no broadcast, then this request is added in the request queue of this broadcast item; Otherwise, if prepare the preparation item in queue, then this request is added in the request queue of this preparation item; Otherwise, this data subitem is added preparation queue as new preparation item, and request is added in the request queue preparing new preparation item in queue;

Step a6, waits request newly to be generated, jumps to step a2;

Step a7, broadcasts according to broadcast queue;

Step a8, to judge whether just to have broadcasted in broadcast queue not empty broadcast item, no, then directly enter step a9, be, then the successful request information of this broadcast item is write successful solicited message file, delete all requests of this broadcast item, delete this broadcast item, then enter step a9;

Step a9, judges whether that broadcast queue is empty, if not, then directly enters step a10, if broadcast queue be empty, then according to etc. the dispatching algorithm of size fractionation data item generate new broadcast queue, then enter step a10;

Step a10, advance a time point, if exceeded the running time of regulation, then jumps to step a11; Otherwise, jump procedure a2;

Step a11, reads successful request message file, statistics solicited message;

Or, adopt when the data item got is split and do not wait size data item to split mode, do not wait the fractionation of size data item fractionation mode and scheduling flow to comprise the following steps,

Step b1, initialization;

Step b2, if current time is not asked, then jumps to step b7, otherwise continues to enter step b3;

Step b3, takes out the request of current time;

Step b4, obtains the data item of this request, and according to not waiting fractionation strategy of size data item to be split, the off period of the data item of fractionation adjusts by the adjustable strategies according to the off period;

Step b5, process respectively for all data subitems after splitting, processing mode is as follows,

To wherein arbitrary data subitem, if at the broadcast item of broadcast queue and no broadcast, then this request is added in the request queue of this broadcast item; Otherwise, if prepare the preparation item in queue, then this request is added in the request queue of this preparation item; Otherwise, this data subitem is added preparation queue as new preparation item, and request is added in the request queue preparing new preparation item in queue;

Step b6, waits request newly to be generated, jumps to step b2;

Step b7, broadcasts according to broadcast queue;

Step b8, to judge whether just to have broadcasted in broadcast queue not empty broadcast item, no, then directly enter step b9, be, then the successful request information of this broadcast item is write successful solicited message file, delete all requests of this broadcast item, delete this broadcast item, then enter step b9;

Step b9, judges whether that broadcast queue is empty, if not, then directly enters step b10, if broadcast queue be empty, then generates new broadcast queue according to the dispatching algorithm of the size fractionation data item such as not, then enters step b10;

Step b10, advance a time point, if exceeded the running time of regulation, then jumps to step b11; Otherwise, jump procedure b2;

Step b11, reads successful request message file, statistics solicited message.

2. consider the real-time data broadcast dispatching method as required of data item size according to claim 1, its feature exists: according to waiting fractionation strategy of size data item to split in step a4, comprise following sub-step,

Step a4.1, initialization, comprises and arranges residue size remainsize=datasize, subitem sequence number subitemorder=1; Datasize is data item size;

Step a4.2, determines to split benchmark size SplitStandardSize as follows,

$SplitS\tan dardSize=\frac{{\mathrm{Bw}}_k\×{\mathrm{Cy}}_k}{SplitK}$

Wherein, Bw _kfor the broadcast bandwidth of kth time broadcast cycle, Cy _kfor the broadcast cycle of kth time broadcast cycle; SplitK is a positive integer, represents that the data items size of broadcast in one-period is the number of the data subitem splitting benchmark size SplitStandardSize;

Step a4.3, if remainsize > is SplitStandardSize, then continues step a4.4, otherwise jumps to step a4.5;

Step a4.4, add a data subitem, data subitem size is SplitStandardSize, filling size is SplitStandardSize, sequence number is subitemorder, subitemorder=subitemorder+1, remainsize=remainsize-SplitStandardSize, jump to step a4.3;

Step a4.5, add a data subitem, data subitem size is SplitStandardSize, and filling size is remainsize, and sequence number is subitemorder, remainsize=0;

Step a4.6, returns the data subitem split.

3. consider the data broadcast dispatching method as required in real time of data item size according to claim 2, it is characterized in that: in step a9 according to etc. the size dispatching algorithm that splits data item generate new broadcast queue, comprise following sub-step,

Step a9.1, enters circulation after the preparation queue before time broadcast of input kth,

Step a9.2, upgrades and prepares each data subitem d in queue _x,ycurrent request number, off period and stand-by period;

Step a9.3, calculates each data subitem d _x,ydispatching priority;

Step a9.4, will have the data subitem d of greatest priority value _x,ybe designated as d _n,mand be assigned to Bd _j;

Step a9.5, judges that in broadcast queue, all data subitem sizes add d _n,msize whether be greater than Bw _k× Cy _kif be greater than, exited circulation and enter step a9.8, otherwise continue to perform step a9.6;

Step a9.6, by Bd _jinsert broadcast queue B _kafterbody, returns step a9.2;

Step a9.7, end loop;

Step a9.8, returns kCi broadcast queue B _k.

4. considering the real-time data broadcast dispatching method as required of data item size according to claim 1, it is characterized in that: according to not waiting fractionation strategy of size data item to split in step b4, comprising following sub-step,

Step b4.1, input data item d _i, data item d _idata item size data subitem d _i,jpriority Weight (d _i,j), bandwidth B w _k, cycle Cy _krear initialization, comprises that to arrange remaining data item size be data item size, subitem sequence number subitemorder=1;

Step b4.2, calculates the fractionation benchmark size SplitStandardSize of this data item,

$SplitS\tan dardSize=\frac{\mathrm{\α}\×{\mathrm{Bw}}_k\×{\mathrm{Cy}}_k}{S_{d_i}\×Weight\left(d_{i,j}\right)}$

Wherein, α is parameter;

Step b4.3, if current residual data item size is greater than SplitStandardSize, then continues step b4.4, otherwise jumps to step b4.5;

Step b4.4, add a data subitem, data subitem size is SplitStandardSize, and sequence number is subitemorder, subitemorder=subitemorder+1, and remaining data item size deducts SplitStandardSize, jumps to step b4.3;

Step b4.5, adds a data subitem, and data subitem size is remaining data item size, and sequence number is subitemorder;

Step b4.6, returns the data subitem split.

5. consider the real-time data broadcast dispatching method as required of data item size according to claim 4, it is characterized in that: the dispatching algorithm splitting data item according to the size such as not in step b9 generates new broadcast queue, adopt the scheduling strategy that priority is maximum, comprise following sub-step

Step 9.1.1, after input prepares queue, initialization broadcast queue is empty;

Step 9.1.2, carries out following circulation,

Step 9.1.2.1, traversal prepares queue, chooses and has greatest priority value Weight(d _i,j) data subitem;

Step 9.1.2.2, if the total size of data subitem in broadcast queue adds this data subitem size still do not exceed the data subitem size that one-period can broadcast, then adds this data subitem to broadcast queue, returns step 9.1.2.1 and circulate; Otherwise, exit circulation, export broadcast queue.

6. consider the real-time data broadcast dispatching method as required of data item size according to claim 4, it is characterized in that: the dispatching algorithm splitting data item according to the size such as not in step b9 generates new broadcast queue, adopt the scheduling strategy considering priority and bandwidth availability ratio, comprise following sub-step

Step 9.2.1, after input prepares queue, initialization broadcast queue is empty;

Step 9.2.2, carries out following circulation,

Step 9.2.2.1, traversal prepare queue, choose have minimum the data subitem of value;

Step 9.2.2.2, if the total size of data subitem in broadcast queue adds this data subitem size still do not exceed the data subitem size that one-period can broadcast, then adds this data subitem to broadcast queue, returns step 9.2.2.1 and circulate; Otherwise, exit circulation and enter step 9.2.3;

Step 9.2.3, initialization residue size;

Step 9.2.4, carries out following circulation,

Step 9.2.4.1, traversal prepares queue, is chosen at data subitem size and is less than maximum data subitem under the prerequisite of residue size, if found, enter step 9.2.4.2, if can not find such data subitem, then exiting circulation and exporting broadcast queue;

Step 9.2.4.2, adds the data subitem found to broadcast queue, returns step 9.2.4.1 and circulate.

7. consider the real-time data broadcast dispatching method as required of data item size according to claim 1, it is characterized in that: adjust according to the off period of the adjustable strategies of off period by the data item of fractionation in step a4 or step b4, comprise following sub-step,

(1) data item d is inputted _igo on the air in kth time broadcast cycle period h data item r off period of asking in moment data item d _idata subitem d _{i, 1}, d _{i, 2}..., d _i,p; Initializing variable j=1;

(2) following circulation is carried out,

(2.1) setting data subitem d _i,joff period

$Deadline=Deadline+S_{d_{i,j}}/{\mathrm{Bw}}_k$

Wherein, Bw _kfor the broadcast bandwidth of kth time broadcast cycle, it is the data item size of a jth data subitem of i-th data item;

(2.2) if j equals data item d _itotal number P of the data subitem split, jumps out circulation and exports the data subitem adjusting the off period, otherwise j=j+1, return (2.1) and continue circulation.

8. consider the real-time data broadcast dispatching method as required of data item size according to claim 1, it is characterized in that: in step a11 or step b11, add up solicited message, comprise following operation,

(1) to read in successful request message file all solicited message to dynamic array vector;

(2) ID of the successful request information in dynamic array vector according to request is sorted;

(3) the successful request information that the ID that asks in dynamic array vector is identical is read successively, if judge that the data subitem sum of the number of successful request information and the data item of request is equal, the request then judging this ID is successful request, adds up the access time of this successful request; Otherwise this request is invalidation request;

(4) read the end of dynamic array vector, then exit, export successful number of request, total access time.