CN101645911B - Dynamic cluster-based news service method - Google Patents
Dynamic cluster-based news service method Download PDFInfo
- Publication number
- CN101645911B CN101645911B CN 200810117824 CN200810117824A CN101645911B CN 101645911 B CN101645911 B CN 101645911B CN 200810117824 CN200810117824 CN 200810117824 CN 200810117824 A CN200810117824 A CN 200810117824A CN 101645911 B CN101645911 B CN 101645911B
- Authority
- CN
- China
- Prior art keywords
- data
- news
- computing
- node
- cluster
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 69
- 230000007246 mechanism Effects 0.000 claims abstract description 26
- 230000003993 interaction Effects 0.000 claims abstract description 7
- 230000026676 system process Effects 0.000 claims abstract 2
- 230000005540 biological transmission Effects 0.000 claims description 25
- 238000005516 engineering process Methods 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 12
- 238000012544 monitoring process Methods 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 6
- 230000007774 longterm Effects 0.000 claims description 5
- 230000001360 synchronised effect Effects 0.000 claims description 2
- 230000000737 periodic effect Effects 0.000 claims 1
- 230000002452 interceptive effect Effects 0.000 abstract description 5
- 238000011161 development Methods 0.000 abstract description 3
- 230000032683 aging Effects 0.000 description 22
- 230000006399 behavior Effects 0.000 description 13
- 238000012546 transfer Methods 0.000 description 12
- 230000000694 effects Effects 0.000 description 9
- 238000004364 calculation method Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 6
- 230000015572 biosynthetic process Effects 0.000 description 5
- 238000011330 nucleic acid test Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 3
- 230000007812 deficiency Effects 0.000 description 3
- 238000012217 deletion Methods 0.000 description 3
- 230000037430 deletion Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 241000337544 Limnoriidae Species 0.000 description 1
- 210000003484 anatomy Anatomy 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000003139 buffering effect Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000001747 exhibiting effect Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 235000014594 pastries Nutrition 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000010076 replication Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Landscapes
- Information Transfer Between Computers (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
本发明公开了一种基于动态集群的新闻服务方法，属于互联网应用开发领域。本发明首先在计算节点上建立数据接收和请求机制，然后在任何两个节点之间建立直接交互，之后节点可以寻找与其计算任务潜在可配合节点并建立连接，然后根据流网协议建立动态集群，根据新闻服务的需要新闻系统对动态集群内的数据进行相应处理。本发明的新闻服务方法，可以及时准确的把握新闻的变化，从而及时有效的提供各种新闻服务，比如对新闻进行有效的排序、提供交互功能等。
The invention discloses a dynamic cluster-based news service method, which belongs to the field of Internet application development. The present invention first establishes a data receiving and requesting mechanism on a computing node, and then establishes direct interaction between any two nodes, and then the node can find and establish a connection with a potentially compatible node for its computing task, and then establish a dynamic cluster according to the flow network protocol, The news system processes the data in the dynamic cluster according to the needs of the news service. The news service method of the present invention can timely and accurately grasp news changes, thereby providing various news services in a timely and effective manner, such as effectively sorting news, providing interactive functions, and the like.
Description
Technical field
The present invention relates to a kind of news service method, belong to the application of distributed computing technology in internet environment based on dynamic cluster.The dynamic cluster technology is a kind of the Internet computing environment based on dynamic change; Wherein relate generally to distributed resource tissue [1], data broadcasting [2], consistency maintenance [3], event-driven mechanism [4] and data backup multiple internet-based present technique such as [5], belong to field of Internet application and development.
Background technology
Information system is one of multinomial basic application in the Internet.As a kind of data resource to time-sensitive, the application system relevant with it must possess immediately, timeliness, quicksort, personalization, characteristic such as extensive even mutual.The current internet information system is because the restriction of itself basic structure can't reach above-mentioned requirements.
News at first requires instantaneity.Based on the information system of current internet basic fundamental [6] on instantaneity to be implemented in technical be imperfection.Instantaneity at first requires to exist between Internet Server [7,8] and its clients corresponding event-driven mechanism.In current internet system, this event-driven mechanism is non-existent, must simulate through voting mechanism [9].But when system scale increased, this voting mechanism will cause excess pressure to system.So, in the current internet information system, have only and adopted voting mechanism in the limited webpage to strengthen the instantaneity of this page.Yet, in most of related news issue page the instantaneity characteristic is not provided, in order to avoid waste great amount of calculation resource.Address this problem, need an effective event-driven mechanism to support.
Assurance to effect of time for news is not enough.Any news data all possesses ageing.But this ageing assurance is to draw through the judgement to the news life cycle.News ageing roughly can be divided into permanently effective, specific effectively and effectively instant.Permanently effective some news data that refer to have high value, and can in the long-term period, keep.Specific some news data that effectively refer to only have value in special time period, exceeding its value of this time period can just can lose gradually.The instant news validity in blink after issue that effectively refers to; Nearly all news data all possesses instant validity.This ageing comprehensive assurance needs the support of great amount of calculation resource to news data, and this becomes current information system and obtains ageing bottleneck.Limited artificial participation of the ageing basic dependence of current information system solved.
Quicksort is also significant to information system.Though news possesses instantaneity, in a period, a plurality of news data can appear.Satisfying the news instantaneity simultaneously, as time passes, should be worth news itself and make rapid judgement, coming out prior data arrangement to avoid significant data to be flooded by a large amount of news data.Current information system does not provide the ordering service, and highlight makes it appear at the remarkable position of the page through the mode of manual intervention fully.Manual intervention is unfavorable for timely adjustment is made in the variation of news data; Because news data is numerous and jumbled, excessive to the demand of manual work; At last, artificial ordering also can't be avoided the objective evaluation to news data.Restricting current information system adopts the subject matter of ordering strategy to be there is not suitable order standard.The most effective order standard is web page interlinkage degree of reusing [10] in the current internet data.But the acquisition of this degree of reusing needs related data that enough life periods are arranged on the internet, and only in this way the related news data just might be reused.But this strategy speciality lucky and news data contradicts.Have a kind of method also can adopt, the number of times of promptly being read according to news is assessed its value.The effective prerequisite of this sort method is in the finite time that news occurs, and has abundant user to see this news.Frequency of reading is just meaningful in this case.But because current information system lacks instantaneity, synchronization can not guarantee that abundant user sees latest data, causes the error of data acquisition.Secondly, height is seen latest data concomitantly, can cause that also information system itself bears excessive pressure.At last, current information system also causes still valuable news to be ignored by system to the deficiency of ageing assurance, makes the user must not see related news, and this has also caused the defective in the ordering.
The scale of information system requires enough big.This also is a great deficiency of current information system.In order to deal with possible potential user, current information system has to increase the input of computational resource; Simultaneously, must be cost also to have reduced service quality, be main like information system with the lightweight data.These modes all cause the cost performance of system to descend.
At last, for press service, good interaction capabilities also is necessary.But under the situation that computational resource is restricted, information system is main with ISP's distributing data mostly, and for the user interactive service is not provided.
Summary of the invention
The objective of the invention is to through the dynamic cluster The Application of Technology make that the internet news system reaches immediately, timeliness, quicksort, personalization, extensive and strong mutual target.
A kind of news service method based on dynamic cluster the steps include:
1) on computing node, sets up Data Receiving and request mechanism;
2) between any two nodes, set up directly alternately;
3) node is sought with its calculation task is potential and is cooperated node and connect;
4) information system is set up dynamic cluster according to the drift net agreement;
5) according to the needs information system of press service the data in the dynamic cluster are carried out handled.
Further, realize said Data Receiving and the request mechanism on computing node, set up through tcp/ip program in the said method.
Further, in the said method through the NAT crossing technology to NAT pass through realize setting up between the said node directly mutual.
Further, realize maximum use through the effective thread pool technology of employing on said node in the said method to each node computational resource.
Further, adopt route technology based on community network to realize that said node is sought in the said method and its calculation task is potential cooperates node and connect.
Further, said dynamic cluster is under the running status, shares set based on the computational resource that common calculating purpose computing node forms, and its characteristic comprises: periodicity life, common purpose, independent assortment, highly dynamic, mesh topology, highly heterogeneous.
Further, the said information system method of setting up dynamic cluster according to the drift net agreement is:
1) employing is sought computational resource based on the routing algorithm of community network;
2) adopt related protocol to carry out transfer of data according to the data characteristics of transmitting between node; Said data characteristics comprise lightweight data, lightweight data relevant with sequential, with sequential associated weight grade data, with the irrelevant heavyweight data of sequential; Take message based reliable multicast algorithm to transmit for said lightweight data; For said and sequential associated weight grade data, with data be divided into the approaching data of sequential and sequential require not urgent to data transmit, take synchronous or asynchronous transport method is transmitted for the approaching data of said sequential.
Further, require not urgent data for said sequential in the said method and be with method that the irrelevant heavyweight data of sequential are taked:
1) computing node and a plurality of potential node are asynchronous concurrent mutual;
2) computing node is sought other potential node simultaneously as both candidate nodes, when the both candidate nodes potential ability is higher than current certain mutual potential node, initiatively poor performance or contribute few node initiatively to replace;
3) each node is monitored with the behavior of assessment node in cluster the node of getting in touch in reciprocal process, rewards the contributor, punishes a node that work is asked for and sets up proportional relation in the node contribution with between asking for;
The method of 4) getting in touch through active of layout central server and computing node in the system guarantees fairness.
Further, when being multiple heterogeneous data for said data in the said method, take lightweight high priority data, the heavyweight data relevant to take second place and transmit with the irrelevant last principle of heavyweight data of sequential with sequential.
Further, said needs information system according to press service is carried out handled to the data in the dynamic cluster, and it comprises:
1) news value assessment; Its method is for to sort to news data according to user behavior and dynamic cluster scale, and said user behavior comprises: the news data to request receives is selected to read, ignore, reject or deletion;
2) news timeliness assessment; Its method is for passing through the ageing of monitoring dynamic cluster LCA news data, said ageing comprising: long-term period, special period and instant period;
3) enhanced system capability; Guarantee the strong retractility of news data delivery system through making full use of computational resource, its method is: each node is contributed self computational resource according to the drift net agreement in the dynamic cluster when obtaining computational resource.
Main contents of the present invention are following:
(1) sets up instant internet news system.Internet news method of servicing of the present invention at first possesses instant characteristic.The issue of any news will be transferred to the user with instant mode.Traditional information system can only be on Internet Server stored data, when user capture, could obtain related data.Any variation of news data or the generation of latest data all can only rely on user's visit: instantaneity is decided by the user capture frequency fully.No matter the present invention is that new data issue or legacy data have more accomplished that instantaneity and user capture frequency have no relation in new capital.
(2) foundation possesses ageing internet news system.Internet news method of servicing of the present invention also possesses ageing.Can reach the purpose that its life cycle is monitored through management, thereby effectively bring into play the value of news itself all news data.The current internet information system to news data ageing assurance mainly accomplish through manual work, this mainly be do not have appropriately to weigh the standard of news data life cycle or obtain the computational resource that index of correlation consumes excessive.The present invention can effectively address these problems, and handles the ageing of news data respectively with long-term, any specific time period and instant three kinds of situation, reaches the purpose to the reliable maintenance of news data validity.
(3) set up the effectively internet news system of ordering.The news ordering is the necessary algorithm that embodies news value.Because the ordering of news data can't cause the current internet sequencing problem to be difficult for solving by means of the multiplexing method of general internet data hyperlink.In addition, the non-instantaneity of news data also causes the deviation that when collecting frequency of reading, occurs.The present invention relies on the support of the instant ability of system, to the reliable collection of user's frequency of reading and corelation behaviour and combine suitable methods of exhibiting, reached the effective purpose of ordering of news data.
(4) set up rational exhibition method.Concerning the internet news system, exhibition method is read news, holds User Status and gathered user's frequency of reading the user all has important function.Rational news exhibition method should be embodied in reflection to effect of time for news, read on evaluation system auxiliary to the fair consideration of showing of data and to data.The current internet information system directly causes on data display, can't providing to the user the ageing data of news owing to not ageing to news data assurance ability.Fair show that referring to all news data all has an opportunity to be read or receive by the user when the news life cycle is initial, this be news data must accomplish.But because the current internet display technique seriously is limited by the computing capability of Internet Server, causes and show that the interface is narrow and small and static, can only show, weakened the fairness that news data is observed finite data.Problem in the displaying directly can have influence on the evaluation to the news data quality.In estimating the news Quality Process, user's behavior (mainly being presented as reading in the press service) also can be affected, and can't effectively make appropriate judge according to user behavior to data.The present invention has made the change of essence on above-mentioned 3, having reached to provide fair chance and directly to contribute to the assessment to news data again for the user provides friendly bandwagon effect for all news data.In the middle of displaying of the present invention, news can be listed respectively with ageing difference, and all data do renewal with the dynamic rolling mode, guarantees in its life cycle, to have the chance of basic equality to show the user.
(5) set up large-scale internet news system.High-quality internet news system entails is large-scale.But it is the increase that resorts to computational resource that the current internet information system realizes large-scale mode.This method itself has high input, and still need under the prerequisite of reduction service quality, could accomplish.Internet news method of servicing of the present invention can guarantee that system provides service under unconfined scale.
(6) set up strong mutual internet news system.Strong also is the characteristic that information system should possess alternately.In the internet news method of servicing of the present invention, the user can also carry out instant extensive comment to news except can obtaining the related news service, promptly so-called mutual.The understanding that this is worth news itself and enlarge the facility that its effect is brought.
Good effect of the present invention is:
Through news service method of the present invention, the variation of assurance news that can be is promptly and accurately effectively sorted to news, and interactive function is provided simultaneously, thereby guarantees ageing, the value that embodies highlight of news.
Description of drawings
Fig. 1. the topological structure at once of a dynamic cluster;
Fig. 2. method flow diagram of the present invention.
Embodiment
Flow process of the present invention is as shown in Figure 2, combines embodiment to describe the present invention in detail at present:
The first step, the calculating of setting up dynamic cluster is basic.Set up dynamic cluster and calculate the basis, at first need make each node when having request msg and service ability, also possess the ability of other node requests of response.Achieve this end not difficultly, utilize basic network technology [11] just can accomplish.Difficulty is to break through the NAT obstacle that exists on the Internet, can adopt the NAT crossing technology to solve, and its method can be with reference to [12].The solution of this problem can be so that can be directly mutual each other between any two nodes in the Internet.This is to produce the prerequisite that large-scale internet calculates.Secondly, such node also need improve its throughput except can asking and respond, and promptly under the enabled condition of node computational resource, can establish a communications link with a plurality of nodes concomitantly, and can in such connection, continue the transmission data.The important technology here is exactly the realization [13] of thread pool.At last, state target in realization after, also having an important process is exactly to give node to seek the ability [14] of cooperative nodes.In a large-scale internet system, any one node can only be connected and cooperate calculating with limited node.Should be that a node searches out optimal node in numerous on the internet computing nodes, thereby realize that computational resource utilizes to greatest extent.Dynamic cluster refers to one under running status, and based on common calculating purpose, the computational resource of formation is shared set by various on the internet the computing equipments of possibly connecting.This set has following characteristic.
(1) life cycle.Dynamic cluster has life cycle.Common dynamic cluster all to pass through generation, development, maturation, steadily, desolate and decay a plurality of stages.But not all dynamic cluster all will pass through these complete processes.Or some cluster directly decay crossing over some stage after the generation can be in generation for a long time and decay outside certain stage.The generation of dynamic cluster is to initiate a calculating through a computing node, causes the response of other nodes, thereby forms cluster; When responsive node increased, cluster just was in developing stage; Developing stage can arrive a limit, and Here it is the stage of ripeness; After the maturity period, cluster scale can tend to balance gradually, and scale no longer enlarges, and does not also obviously dwindle, i.e. plateau; Along with most nodes are accomplished calculating and left this cluster, the node in the cluster reduces gradually, thereby causes the formation in desolate stage; After all computing nodes all left in the cluster, cluster was withered away, and got into and decayed the stage.
(2) common purpose.The basis that dynamic cluster forms and keeps is common calculation task.Calculation task refers to the transmission that takes place in data or the service process, the mutual and computing activity of providing.A dynamic cluster comprises a plurality of nodes; These nodes all carry out correlation computations round one or more correlation computations tasks.They are the beneficiary of related data and service, also are the indirect suppliers of related data and service.
(3) independent assortment.Node in the dynamic cluster has the characteristic of independent assortment.No matter how the cluster scale size changes, and the node in the cluster all adds under to the interested condition of correlation computations task.After node has been enjoyed related data or service, perhaps lose the data or service that correlation computations provides after, then can leave system at any time and not receive any restriction.
(4) highly dynamic.The principal character of dynamic cluster is exactly that this cluster always is in the middle of the continuous variation.This variation both had been embodied in cluster itself and had had life cycle, was also embodied in node and constantly added or leave, even be embodied in the variation on the topological structure that cluster has.These behavioral characteristics reflect cluster state at once; Assurance to these states can be played important support effect for the system that builds on such cluster.
(5) mesh topology.As a kind of application form of distributed computing technology, any cluster all can show with a kind of form of topological structure.To the understanding of topological structure, also can hold appropriately to the characteristic of cluster.Dynamic cluster among the present invention shows as network structure.But this description is too rough.Understanding to this structure can be with reference to related article [15].Why forming network structure but not other topological structures, is because such structure can be so that can share computational resource separately between each node to greatest extent, thereby under the dynamic situation of height, realizes the maximization of calculated performance.Fig. 1 has represented the topological structure at once of dynamic cluster.
(6) highly heterogeneous.Though each node is based on common task and is organized into cluster, still has many differences in this cluster.What is called is highly heterogeneous to be showed by following characteristics.The first, the heterogeneity of node resource.Usually show between each node have the difference of computational resource; And the node that has a various computing resource is different to the potential contribution of system.Should take appropriate algorithm [14,16] both to guarantee the normal operation of whole cluster, also can make each node can access the service on himself resource base of maximum using.The second, the heterogeneity of data or service.Data are to maintain the dynamic cluster life cycle to be in the basic of stable state with service.In the middle of High-Performance Computing Cluster, such data or service can show complicated form, like the data of various forms.The difference of data format cause between node mutual when cooperating in a different manner [14,16] handle data of different nature, reach the reasonable use of computational resource.The 3rd, the heterogeneity of node physical location.Because the formation of dynamic cluster is decided by data and service quality fully, and non-artificial prior definition systems stabilisation, this causes possibly comprising in the cluster from the big node of physical location difference.Suitable routing mode [17,18] is the high efficiency of whole cluster both, also can guarantee that node obtains fair service on the different physical locations simultaneously.The 4th, the heterogeneity of nodes ' behavior.The behavior of node in dynamic cluster also had any different.Some node can occur with computational resource contributor's form, and some then is to ask for and contribution is had both, and also some just asks for [19] merely.The dynamic cluster algorithm must be considered these characteristics, to guarantee the high efficiency of cluster.
In second step, set up dynamic cluster.Had the needed basis of dynamic cluster and to large-scale internet in dynamic cluster had understanding after, just can interdependent node organize with formation dynamic cluster.Based on the characteristic of above-mentioned dynamic cluster, it is different fully with the cluster under stable environment to set up the method for dynamic cluster.The drift net agreement is an important solutions wherein.The drift net agreement can be summarized as follows simply.
(1) foundation is based on the routing algorithm of community network [15].The front simplicity of explanation to the basic demand of route technology.This routing plan is the resource finding algorithm that is based upon on community network or complex network [15] basis.Because community network is the topological structure that meets Internet resources distribution rule that computing node is set up through interconnected user on the network's social property in essence, utilizes this network just can accomplish the effective searching to computational resource rightly.When utilizing community network to carry out route search, relate generally to the combination problem of structuring and destructuring route technology.The outstanding feature of structuring route technology is in confirming finite time, to obtain route results to any resource (rare or popular), but such result is based upon under the strict constraints.And this condition current internet environment is difficult to be guaranteed.The destructuring route technology is based upon on the topological structure (being community network) that the Internet forms naturally, the internet environment that exists is not had any mandatory requirement.On such basis, will guarantee to obtain Search Results fast, necessarily requiring has deep understanding just can reach to concrete application and corresponding topological structure.But even so, unstructured searching is still not enough in the assurance of route fairness, and is also difficult to sharing of route pressure simultaneously.In view of the foregoing, can appropriately combine said two devices.Because have group in the community network, group is representing metastable computing environment, approaches the desired condition of structuring route.Like this, can take in whole the Internet, to carry out the destructuring route, and in group, carry out the structuring route.Through said method, routing algorithm has just been accomplished the popular resource of effective searching and has been guaranteed the route justice to scarce resource.
(2) set up the lightweight data transmission scheme.The lightweight data definition can send the data that finish for only needing the finite data bag.Usually the lightweight data total data amount size that can allow in the computing node bandwidth unit time.The concrete state of node decided when concrete size can be according to actual bandwidth and operation.In a word, such data are very little to the pressure that computing node, bandwidth and interdependent node cause.However, to, a plurality of nodes, also can cause source node pressure excessive when sending this data when a node if potential receiving node is too much.At this time can utilize message based reliable multicast algorithm [2] to solve this problem.Because this data are sent fast, and are not high to the requirement of reasonableness of topological structure, as long as reach the purpose of sharing the source node burden.When the transmission of heavyweight and lightweight data mixing is arranged in the system, it is also conceivable that and utilize the topological structure of heavyweight data multicast formation to realize the lightweight transfer of data; This method can be with reference to the lightweight data multicast mechanism based on BT agreement [14].This machine-processed principle is that the stability and high efficiency topological structure that utilizes the heavyweight data multicast to form reduces network pressure and delay, thereby reaches the extensive multicast of lightweight data.
(3) set up and the irrelevant heavyweight data transmission scheme of sequential.Opposite with the lightweight data, the heavyweight data need be sent a plurality of packets could accomplish the data transmission, and it sends number will be far longer than the lightweight data.Therefore, sending node and recipient node and network of relation are all caused much higher pressure.This pressure is mainly reflected in that sending duration is long, to take resource many.According to the characteristic of dynamic cluster, realize long-time extensive heavyweight transfer of data, its agreement must guarantee to make computing node to operate under the metastable high dynamic environment.Reach above-mentioned target; A computing node is when participating in such transmission; At first to form with a plurality of potential nodes asynchronous concurrent mutual; Remedying the low problem of bandwidth availability ratio between any two nodes, the acquisition of potential node can be through being provided with central server in system, and promptly all nodes of participating in the heavyweight transfer of data will arrive all that central server is registered and state variation announcement server periodically; The maximum characteristics of this central server be can be initiatively with system in computing node get in touch, rather than always be in the status that passive wait computing node is reported oneself state.Simultaneously come the transmission weight grade data alternately, also can reduce of the influence of network dynamic this node with a plurality of nodes; This is the method that improves systematic function and fault-tolerant ability through reproduction technology [20].Secondly, in this process, also require any node constantly to go to seek other possibility both candidate nodes; When its potential ability of both candidate nodes is higher than current certain mutual node; Should initiatively perhaps contribute poor performance few node initiatively to replace; Because any one both candidate nodes all will be carried out enough for a long time alternately with present node, present node just can the evaluation node relative ability through volume of transmitted data is compared in this reciprocal process.Once more,, take different strategies, only make the node of asking for and between node is contributed and asked for, set up proportional relation etc. as rewarding contributor, punishment for different behavior nodes in the cluster; Reach above-mentioned target, each node also need be monitored the node of getting in touch in reciprocal process, with the behavior of accurate assessment node in cluster.At last, safeguard in the system when contributing the node interests more, also will consider fairness; This is because not all dynamic cluster can both reach the transmission requirement of heavyweight data.Owing to arranged the abundant central server of computational resource in the system, central server guarantees that with mode initiatively weak cluster obtains the help on the computational resource.What need supplementary notes is, mentioned above principle for the irrelevant heavyweight data of sequential be effective; Reason is that such data can split arbitrarily and make up, and can relatively easily realize asynchronous and concurrent.
(4) set up and sequential associated weight grade data transmission plan.For with sequential associated weight grade data, aforesaid way just need be done modify.With sequential associated weight grade data except the characteristic that possesses general heavyweight data, also have a temporal aspect, promptly data must with in order and the mode that is not less than a certain speed transmit.Be sequential between each packet of data like this, and cross when low,, also be not worth even be sent to the destination when data packet transmission rates.Ask for something transfer of data and data use the data of carrying out simultaneously can have above-mentioned characteristic, like the Voice & Video data.This data all have requirements at the higher level to the computational resource that each node in the stability of cluster and the cluster has.Compare with the irrelevant heavyweight transfer of data of sequential, because can't data be split arbitrarily and combination, the main distinction is that asynchronous or concurrent difficulty increases.But if do not have asynchronously or concurrent, it is just nonsensical that cluster exists, and the shared computational resource that just can not cooperate with each other between the node causes overall performance to improve.General way is to be divided into two parts to this data, and a part is the approaching data (the urgent data of sequential) of sequential, and a part is that sequential requires urgent not enough data (the urgent data of non-sequential).For the data that generated, the ratio that the former accounts in the whole data is relatively little, and the ratio that the latter accounts in the whole data is much bigger.And, also can be divided into similar two parts to it through the way of buffering for the data of instant generation.Can take pure synchronization transfer method [16] for the urgent data of sequential, also can take asynchronous system [16], the two can be aided with the help of contribution type resource.For the urgent data of non-sequential, can be fully according to treating with the irrelevant heavyweight data transfer mode of sequential.Because the urgent data of non-sequential finally can be evolved into the urgent data of sequential, the latter's transmission can alleviate the former pressure greatly.Certainly, under the condition of finite bandwidth resource, the former must have higher priority to obtain computational resource.
(5) realization constitutes the transmission of entity to multiple heterogeneous data.In practical application, multiple heterogeneous data can be formed the significant entity to the user.Data mutual group in this entity lumps together just meaningful, perhaps departs from original meaning and just lost after the partition.More to be a plurality of entities according to fixing order be disposed in complicated situation gives expression to certain sense together.Will form following general layout like this, promptly transmit with characteristic separately, and between entity, also have sequential property at the inner data demand of different nature of an entity.Transfer of data in this case just needs the appropriate above-mentioned three kinds of transmission methods to different pieces of information of using, and under the prerequisite that guarantees the resource maximum using, guarantees intrinsic sequential property between the data.The problem that drift net agreement that Here it is mainly solves.Its elementary tactics at first is the lightweight high priority data in each entity; Reason is mostly these lightweight data are text data, even the lightweight data volume sum in whole entity is compared also minimum with the heavyweight data; Simultaneously these lightweight data itself also comprise the descriptor to entity, and this data of prioritised transmission help receiving node has overall understanding to entity and the relation between them, thereby scheduling resource is more suitably accomplished the transmission of all solid datas.Secondly, behind all lightweight Data Transfer Dones, again heavyweight data relevant with sequential in the entity that makes number one are transmitted; Meanwhile, with the irrelevant heavyweight data of sequential suitable download probability is provided for arranging in first the entity; If bandwidth resources still have residue, the heavyweight data allocations that can consider to be followed successively by in all the other entities is downloaded probability.At last; After heavyweight transfer of data relevant with sequential in first entity finishes; If itself and the irrelevant heavyweight data of sequential also do not have end of transmission; Then directly transmit heavyweight data relevant in second entity with sequential, and only for the irrelevant heavyweight data of sequential suitable download probability being provided in first entity.The rest may be inferred, accomplishes the transmission of all entities.This strategy can be summarized as lightweight data override, the heavyweight data relevant with sequential are taken second place and be placed on last consideration with the irrelevant heavyweight data of sequential.This universal law with the real world applications scene matches: one has and generally includes text (lightweight data), picture (lightweight data), audio frequency (with sequential relevant lightweight data) and video (with sequential associated weight grade data) etc. in the implication entity, seldom can occur and heavyweight data that sequential is irrelevant.
In the 3rd step, set up the event-driven mechanism in the dynamic cluster.One of basis that dynamic cluster forms is exactly to be prerequisite each other alternately with the event-driven.Be that a node request connects, initiates transfer of data and closes connection or the like basic act and all accomplish through event-driven mechanism.This and traditional the Internet are that master's mode is completely different with the ballot.And after the basis of setting up dynamic cluster, this event-driven mechanism has just formed basic framework.Combine lightweight multicast mechanism efficiently again, just can form a message based event-driven mechanism.This mechanism can adapt to the characteristic of the Internet dynamic change, guarantees event-driven mechanism work under the reliable extensive environment of efficient stable.This machine-processed realization has improved the interactive efficiency between the Internet computing node, has avoided voting mechanism to the pressure of system's increase and to the system change torpid reaction.Event-driven has guaranteed that concern person can need not pay extra cost and vote before incident takes place; Simultaneously, after incident takes place, can react at once.The remote events driving mechanism becomes the basis that the reflection the Internet changes under the internet environment.The Internet is a system that constantly changes in essence, but this variation can't be reflected to the correlation computations node immediately under the support that does not have the remote events driving mechanism.In order to reduce computation burden, during current internet is mainly used in addition the refusal these variations are followed the tracks of.The present invention has solved this problem effectively.
In the 4th step, news data is sorted according to user behavior and dynamic cluster scale.Value assessment is one of difficult problem of current internet press service to news data, and this problem was set forth in front.In the news service method of the present invention, news data will be sent to potential computing node with event driven mode after being published, rather than passive wait on Internet Server.The user can select to read, ignores, reject behaviors such as perhaps deletion according to the hobby of self to the data that obtain on the computing node of self controlling.The accumulative total effect of these behaviors has just formed the assessment that news data is worth.When comprising the heavyweight data in the news data, at this moment the issuing process of news data meeting last much longer can weigh news value through the size of corresponding dynamic cluster scale.The measurement of this value can be to accomplish near real-time speed because system possesses instant event-driven mechanism.The non-instantaneity of current internet makes consistency maintenance difficulty, further cause through frequency of reading sort slowly and also error big.Way of the present invention has been avoided the problems referred to above.
The 5th step is through monitoring the ageing of dynamic cluster LCA news data.It is incontrovertible that news data has ageing by force.But ageing assessment is to be based upon on the instantaneity basis.News service method of the present invention can obtain immediately latest data generation, change, duplicate and the data of dynamic change such as deletion, this is for holding the ageing basis of having set up of news.The assurance of effect of time for news is actual to be that the combination of news life cycle and news ordering is used.The present invention can utilize the state of the strong server monitoring news of computing capability; This status monitoring is actually the statistics that news data is caused dynamic cluster in a period of time.When a news continuous cluster that causes in special time, perhaps cluster is never decayed, and can prolong monitoring time; Weigh the value of news simultaneously according to cluster scale; When a news data can't cause dynamic cluster in the sufficiently long time, can think that this news life cycle finishes.Based on reliable monitoring mechanism, the ageing of news can both effectively be guaranteed in long-term period, special period and instant period.Supplementary notes be to be one of main task during dynamic cluster is safeguarded for the monitoring of dynamic cluster; Only in this way assisting on the computational resource just can be given to the cluster a little less than the computing capability by system, and utilizes computational resource more than needed in the strong cluster of computing capability.On this basis, carry out the monitoring to the news data dynamic cluster, its job specification is the same, need not consume extra resource.
In the 6th step, guarantee the strong retractility of news data delivery system through making full use of computational resource.So-called strong retractility is the synonym that forms the large scale system ability.On the basic platform of dynamic cluster,, computing node will move after freely adding with the mode of dynamic cluster, and each node all can be contributed self computational resource according to the drift net agreement when obtaining computational resource in such cluster.Certainly, obtain total amount and respective organization mode that strong retractility not only is decided by computational resource, also be decided by the concrete demand of calculating resource.For press service, the computational resource that news data is consumed when transmission is maximum, and its displaying can be dealt with for current computing node; Because it is little that transmission news data resource requirement total amount and the total computational resource of potential computing node are compared proportion, is aided with the such efficient resource organizational form of dynamic cluster again, realize that the strong retractility press service of unlimited scale has just been accomplished easily.
In the 7th step, set up exchange method based on dynamic cluster.Dynamic cluster can also be in order to carry out the mutual of strong retractility except a good platform as information system.When carrying out,, in dynamic cluster, accomplish such task and can reach near infinitely great scale because message itself belongs to the lightweight data based on message mechanism mutual.In the time need carrying out between the news user, must do suitable scheduling alternately to such based on the heavyweight data interaction.This scheduling needs just when having served as that the user need carry out based on the heavyweight data interaction with the concurrent mode of height at high proportion.This time is even computational resource also deficiency can occur under the support of drift net agreement.Fortunately, this situation is uncommon in the middle of practical application.This is because dynamic cluster is to run in the community network environment; The social property of network causes having only the minority node in its leading role of system.In should using, mean that the minority node is initiatively issued and alternately, most of node is in passive accepting state.In a word, the information system based on the dynamic cluster technology can provide than the strong much bigger interactive mode of current information system.
List of references
【1】Ian?Foster，Carl?Kesselman；The?Grid-Blueprint?for?a?New?Computing?Infrastructure；Second?Edition，Elsevier?Inc.，ISBN?7-111-16054-1，2004
【2】Obraczka，K..；Multicast?Transport?Protocols：a?Survey?and?Taxonomy；Communication?Magazine，IEEE，Volume?36，Issue?1，Jan.1998Page(s)：94-102
【3】Hadzilacos?V.，Toueg?S.；A?Modular?Approach?to?Fault-tolerant?Broadcasts?andRelated?Problems；Technical?report，Dept?of?Computer?Science，University?of?Toronto，1994
【4】Jean?Bacon，John?Bates，Richard?Hayton，Ken?Moody；Using?Events?to?BuildDistributed?Applications，Services?in?Distributed?and?Networked?Environments；1995，SecondInternational?Workshop?on，5-6June?1995Page(s)：148-155
【5】Wiesmann?M.，et?al；Understanding?Replication?in?Databases?and?Distributed?Systems；in?Proceedings?20
th?International?Conference?on?Distributed?Computing?Systems(ICDCS?2000)，Taipei，Republic?of?China，IEEE
【6】Mah?B.A.；An?Empirical?Model?of?HTTP?Network?Traffic；INFOCOM’97，SixteenthAnnual?Joint?Conference?of?the?IEEE?Computer?and?Communications?Societies，ProceedingsIEEE，Volume?2，7-11?April?1997Page(s)：592-600
【7】Callaghan?B.；WebNFS?Client?Specification；Internet?RFC?2054，October?1996
【8】Callaghan?B.；WebNFS?Server?Specification；Internet?RFC?2055，October?1996
【9】Satyanarayanan?M.；Distributed?File?Systems；in?Mullender?S.，Distributed?Systems，anAdvanced?Course，2
nd?Edition，Wokingham：ACM?Press/Addison-Wesley，1989，Page(s)：353-383
【10】Brin?S，Page?L.；The?Anatomy?of?a?Large-Scale?HyPertextual?Web?Search?Engine；WWW7，Computer?Networks?30(1-7)，1998，Page(s)：107-117
【11】Vinton?Cerf；Specification?of?Internet?Transmission?Control?Program；RFC?675，December?1974
【12】Rosenberg?J.，et?al；STUN-Simple?Traversal?of?User?Datagram?Protocol(UDP)Through?Network?Address?Translators(NATs)；RFC?3489，March?2003
【13】Ananth?Grama，et?al；Introduction?to?Parallel?Computing；Second?Edition，Benjamin/Cummings?Publishing?Company?Inc，1994，ISBN：0-201-64865-2
【14】Cohen?B.；Incentives?Build?Robustness?in?BitTorrent；in?Workshop?on?Economics?ofPeer-to-Peer?Systems，Berkeley?USA，May?2003
【15】Newman?M?E?J.；The?Structure?and?Function?of?Complex?Networks；SIAM?Review，2003，45，Page(s)：167-256
【16】Vlavianos?Aggelos，et?al；BiToS：Enhancing?BitTorrent?for?Supporting?StreamingApplications；INFOCOM 2006，25
th?IEEE?International?Conference?on?ComputerCommunications?Proceedings，April?2006，Page(s)：1-6
【17】Rowstron?A.，Druschel?P.；Pastry：Scalable，Decentralized?Object?Location?andRouting?for?Large-Scale?Peer-to-Peer?Systems；in?Proceedings?of?the?2001?ACM?SIGCOMMConference?on?Applications，Technologies，Architectures，and Protocols?for?ComputerCommunication(SIGCOMM’01)，San?Diego，CA，August?2001，page(s)：247-254
【18】Gribble?S.D.；Measuring?and?Analyzing?the?Characteristic?of?Napster?and?GnutellaHosts；Multimedia?Systems?Journal，9(2)，August?2003，page(s)：170-184
【19】Adar?E.，Huberman?B.A.；Free?Riding?on?Gnutella；First?Monday，Internet?Journal，Oct.2000，Available?at?
http://www.firstmonday.dk/issues/issue5?10/adar/index.html
【20】Gifford?D.K.；Weighted?Voting?for?Replicated?Data；in?Proceeding?7
th?Symposium?onOperating?Systems?Principles，ACM，page(s)：150-162
Claims (7)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200810117824 CN101645911B (en) | 2008-08-05 | 2008-08-05 | Dynamic cluster-based news service method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200810117824 CN101645911B (en) | 2008-08-05 | 2008-08-05 | Dynamic cluster-based news service method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101645911A CN101645911A (en) | 2010-02-10 |
| CN101645911B true CN101645911B (en) | 2012-12-12 |
Family
ID=41657631
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200810117824 Expired - Fee Related CN101645911B (en) | 2008-08-05 | 2008-08-05 | Dynamic cluster-based news service method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101645911B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102436601A (en) * | 2011-11-09 | 2012-05-02 | 江苏联著实业有限公司 | Mobile Internet news value evaluation system |
| US9436917B2 (en) * | 2013-02-05 | 2016-09-06 | Cisco Technology, Inc. | Accelerating learning by sharing information between multiple learning machines |
| JP7394346B2 (en) * | 2020-06-05 | 2023-12-08 | 株式会社ジェイ・キャスト | Article processing system and article processing method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1645829A (en) * | 2005-01-27 | 2005-07-27 | 中国科学院计算技术研究所 | Topological matching method for structured P2P system |
| CN101083656A (en) * | 2007-07-05 | 2007-12-05 | 上海交通大学 | Data stream technique based multi-source heterogeneous data integrated system |
-
2008
- 2008-08-05 CN CN 200810117824 patent/CN101645911B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1645829A (en) * | 2005-01-27 | 2005-07-27 | 中国科学院计算技术研究所 | Topological matching method for structured P2P system |
| CN101083656A (en) * | 2007-07-05 | 2007-12-05 | 上海交通大学 | Data stream technique based multi-source heterogeneous data integrated system |
Non-Patent Citations (2)
| Title |
|---|
| Aggelos Vlavianos.BiToS: Enhancing BitTorrent for Supporting Streaming Applications.《INFOCOM 2006. 25th IEEE International Conference on Computer Communications Proceedings》.2006, * |
| Jean Bacon等.Using Events to Build Distributed Applications.《Services in Distributed and Networked Environments, 1995., Second International Workshop on》.1995, * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101645911A (en) | 2010-02-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jelasity et al. | T-man: Gossip-based fast overlay topology construction | |
| Xia et al. | A survey of bittorrent performance | |
| Yang et al. | Performance of peer-to-peer networks: Service capacity and role of resource sharing policies | |
| CN102571839A (en) | P2P content delivery method based on social attribute of users and system adopting same | |
| Ioannidis et al. | On the design of hybrid peer-to-peer systems | |
| CN101808104A (en) | Method for constructing internet operating in streaming manner | |
| Chandler et al. | Toward p2p-based multimedia sharing in user generated contents | |
| CN101645911B (en) | Dynamic cluster-based news service method | |
| CN1937553B (en) | Reciprocal network data dispatching method based on flow media data frame | |
| Matos et al. | Scaling up publish/subscribe overlays using interest correlation for link sharing | |
| CN101645872B (en) | Business service system of Internet large-scale application environment and working method thereof | |
| He et al. | P2P Traffic Optimization based on Congestion Distance and DHT. | |
| Zhong et al. | Topological model and analysis of the P2P BitTorrent protocol | |
| Guo et al. | Design and implementation of real-time management system architecture based on GraphQL | |
| Ghosh et al. | Utilizing Continuous Time Markov Chain for analyzing video-on-demand streaming in multimedia systems | |
| CN101800655A (en) | Peer-to-peer service system establishing method for contributing resources to application of large-scale internet | |
| Wang et al. | SmartPeerCast: a Smart QoS driven P2P live streaming framework | |
| Ha et al. | A novel Hybrid CDN-P2P mechanism For effective real-time media streaming | |
| Cherbal et al. | Peer-to-Peer lookup process based on data popularity | |
| Gu et al. | ASAP: An advertisement-based search algorithm for unstructured peer-to-peer systems | |
| CN101645826B (en) | Cluster-based structuralized routing method | |
| Hai et al. | A P2P E-Commerce Model Based on Interest Community | |
| Jia et al. | Modelling of P2P‐Based Video Sharing Performance for Content‐Oriented Community‐Based VoD Systems in Wireless Mobile Networks | |
| Sarma et al. | Uniform load sharing on a hierarchical content delivery network interconnection model | |
| CN101800689A (en) | Role-based routing method in large-scale noncentral Internet |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121212 Termination date: 20150805 |
|
| EXPY | Termination of patent right or utility model |