CN1831776A - Consistency maintemance method of marking backtrack - Google Patents

Abstract

A method for holding consistence of multicopy data includes program control algorithm, backtracking algorithm and operation - execution algorithm, execution of local user operation and execution of allopatric user operation. The execution of local user operation is as follows: executing operation on data copy by program, revising local state vector, using local state vector as operation time seal being broadcasted to the other users; the execution of allopatric user operation is as flows: receiving operation sent by allopatric user, backtracking address space of local data copy to its generation time, executing operation on this time directly.

Description

The consistency maintaining method of mark retrace

Technical field

The invention belongs to computer supported cooperative work (Computer Support Cooperative Work) technical field, be specifically related to a kind of consistency maintaining method of mark retrace, this method can be so that a plurality of user edits same data object in real time by existing application software, by a kind of real-time collaborative technology of not having lock, can not influence each other between a plurality of users, the present invention is the operating collision between the process user automatically, realizes the consistency maintenance of data between each user.

Background technology

" computer supported cooperative work " (CSCW, Computer Support Cooperative Work) can be defined as: in the environment that computer technology is supported (being Computer Support), a common task (being Cooperative Work) is finished in a groupware collaborative work.Its target is to design the application system of various collaborative works.

The product of CSCW can become groupware (GroupWare).CSCW and groupware have nuance, and CSCW is a subject, and groupware is a concrete technology or an entity, and the CSCW system of specific implementation is considered to the example of groupware one class, but two terms are also used with sometimes.

The formation and development of CSCW has certain certainty.At first, in the information society in modern times, people's life style and labor style have group, interactivity, characteristics such as distributivity and collaborative.Secondly, the develop rapidly of computer technology (comprising parallel and distributed treatment technology, multimedia technology, database technology, cognitive science or the like), communication and computer networking technology has constituted the technical foundation that CSCW realizes.In addition, the proposition of this notion of concurrent engineering (Concurrent Engineering) also plays an important role.Concurrent engineering is the systems approach of integrated, concurrent designing product and correlated process, and it emphasizes Team Work (group work), and is closely-related with the research of CSCW to the technical support of Team Work.Therefore we can say that CSCW is in modern society, is background in people's collaborative work mode, with computing machine and development of Communication Technique be fused to the basis, has application fields and be a precondition and formation naturally.It relates to numerous ambits, as computing machine, and management, communication, compartment system, artificial intelligence, sociology, psychology, theory of organization or the like.

The purpose of CSCW is exactly the support that provides under computer environment people's teamwork, therefore says, communication, cooperation, coordination are the three elements of CSCW.

The basis of CSCW is communication, it is between the user who distributes (local communication can be thought the special case of compartment system) geographically that the group communication of nature takes place, therefore network service is vital, and handles the multimedia file transmission and Data Control is very complicated in co-operative environment.And computer based or be the communication of medium with the computing machine, not fully and other communication form combine.Asynchronous text based Email and bulletin board are different with synchronous phone and face-to-face talk: people can not transmit file between two telephone numbers arbitrarily.Computer Processing technology and the communication technology combined to help to address this problem.

The form of CSCW is cooperation, and communicates by letter similarly, and cooperation is the important content of group work.In group activity, any activity all must be that multi agent cooperation is finished.Effectively the collaboration requirements people must share information.But current infosystem especially Database Systems is kept apart people mutually under many circumstances.Such as, when using same CAD database, two designers advance line operate Time, and they can not revise the different piece of same design object simultaneously and know his modification that the co-worker made; They must just can know the work that the other side does by mutual inspection.Many tasks all need good shared environment, in due course the action message of friendly notice group and each user's activity.

The key of CSCW is to coordinate.If the activity of a group is coordinated, its communication and cooperation will be strengthened greatly so.The work group that can not well coordinate will certainly often clash and the duplication of labour between its member.When task of the common composition of several sections, coordination itself is counted as a requisite activity.Current database application provides the visit to shared object, yet most of Software tool only provides the support to single user, to this critical function of coordination of supporting group done but seldom.

Reinforcement along with enterprise and mechanism's globalization consciousness (providing its product or service) towards global range, therefore the effect that aspect performance such as can support the system of this cooperation will be at standardized administration, to raise the efficiency, reduce cost can not be substituted also will become requisite information facility., the trend of the extensive cooperation of this trend is bright and clear gradually along with the appearance of some important application patterns.These application models comprise: e commerce transactions, Virtual Organization, collaborative building, long-distance education, tele-medicine or the like, and we believe that these application models will be to generation far-reaching influences such as future social people's work, study and lives,

The famous market analysis IDC of mechanism every two years can issue the market analysis and the prediction of a collaboration software, data are pointed out, as far back as calendar year 2001 U.S.'s collaboration software market share reached 3% to 5%, the last report points out that global TCA market scale will reach 7,500,000,000 dollars by 2009.

The report of Computer World Information (CCW Research) issue shows that Chinese collaboration software market scale reached 5.89 hundred million yuan as far back as 2004, wherein works in coordination with the tool software market scale about 0.6 hundred million yuan, about 5.29 hundred million yuan of collaborative platform and applied software markets; Collaboration software market mainly concentrates on three zones in East China, south China and North China, and wherein the zone, East China accounts for 26.76% of whole markets.2004, the collaborative instrument of China, platform and applied software market total sales were above 5.89 hundred million yuans.Since 2000, along with popularizing of informationalized propelling, internet and IT application, Chinese collaboration software market formed gradually.Especially Email becomes after the tool of communications important in enterprise's day-to-day operations management day by day, and collaboration software market becomes an important component part of applied software market gradually.Expected that collaboration software market will be with 34.45% annual compound growth rate development 2004～2008 years, by 2008, the market total value will reach 19.26 hundred million yuan.

Groupware is a vital classification in CSCW field, promptly supports people to carry out the software systems of collaborative work.Group editor is that the groupware application field is a kind of.It allows a plurality of users to participate in the editor and the modification of a sharing data objects simultaneously by network.Group editor is not limited only to text, can also be applied to the collaborative modification (literal, figure, medium etc.) of various types of data object, is an important technology safeguarding many copy datas object consistency.The research of group editing technique is the top international conference ACM CSCW in CSCW field and the important content on the ACM Group always, and special symposial has also been held five.Group editor's correlation technique has also obtained using widely, is implemented as two systems that support the multi-user to organize editor based on MS Word/MS PowerPoint.

Target the most basic of group editor is to hide the impression of user to network delay.By safeguarding a data copy, realize making an immediate response to local operation at each user side.So consistency maintenance is one of topmost challenge among the group editor.In the past twenty years, many relevant technology have also obtained tremendous development, but the solution that many difficult problems of consistency maintenance aspect still do not improve.In the world on the consistency maintenance problem, mostly be conceived to operate the method for conversion at present.But the operation conversion method still exists very big difficulty in the some difficult problems that solve consistency maintenance and aspect the support of Undo operation.

Summary of the invention

The objective of the invention is to propose a kind of consistency maintaining method based on mark retrace.

What the present invention proposed returns the consistency maintaining method of moulding based on mark, is the consistency maintenance engine that will make up an application program, comprises control algolithm, back-track algorithm and the operation execution algorithm of program in this engine.The state vector maintaining method and the broadcasting method that also comprise a this locality in the engine.Consistency maintaining method is divided into two parts: to the execution of local user manipulation, to the execution of strange land user operation.The former execution flow process is: program is to the data trnascription executable operations of this locality, revises local state vector, with the state vector of this locality timestamp as operation, in operation, is broadcast to all other users then; The latter's execution flow process is: control algolithm is received the operation that the strange land user transmits, when executive condition is satisfied in operation, the timestamp that adheres to according to operation, the address space of recalling the local data copy produces constantly to it, on this address space, operation can directly be carried out, if recover address space to treat next operation.The present invention can be so that a plurality of user edits same data object in real time, by a kind of real-time collaborative technology of not having lock, can not influence each other between a plurality of users, the present invention is the operating collision between the process user automatically, realizes the consistency maintenance of data between each user.

Introduce some relevant key concepts below earlier.

Group editor is the vital classification in groupware application field.It allows a plurality of users to participate in the editor and the modification of a sharing data objects simultaneously by network, and group editor's a important goal is exactly making an immediate response of local operation, and promptly the nothing of local operation postpones to carry out.The nothing that realizes operation postpones the full copy type structure of an important means of execution with regard to data, be that each user keeps a data trnascription, the user operates in this locality and can carry out immediately, sends to other user then, but full copy type structure can cause the consistency maintenance problem.

Real-time CSCW system must satisfy following 3 condition for consistence simultaneously and be only consistent: 1. data consistency (convergence): after executing one group of co-operating, the content of copying data should keep in full accord on each node: 2. cause and effect consistance (causality preservation): any two operation C1, C2, if the cause-effect relationship on their life periods: C1 → C2, then on all nodes, C1 carried out before C2; 3. wish consistance (intentionpreservation): any two operation C1, C2 is if they are separate: C1 ‖ C2, then their results of carrying out with any order on any node expected results unanimity of carrying out in this locality of Ying Yuqi.

Based on the timestamp (State Vector based timestamp) of state vector, be widely used for differentiating the causal sequence (Causal Relationship) of operation room.A timestamp is a vector, website of each dimension expression of vector, and the value representation on this one dimension has been carried out the quantity of the operation in this website in the vector.Because the operation from same website all is that order is carried out, so which operation that timestamp can be illustrated under the state of this timestamp executed, if all executed operations are included in the executed operation of another timestamp in the timestamp, then two timestamps satisfy cause-effect relationship, otherwise are concurrency relations.In the timestamp pattern, each website all can have its state vector, all can together broadcast together with the timestamp of current website after each operation produces.

One piece of linear structure that document is made up of character nodes.Be without loss of generality, suppose that each character all is inseparable, and each operation is all only at a character.In fact the insertion of any one complexity, deletion action can resolve into a set at the operation of monocase apace, and this hypothesis does not influence the ability to express of operation.We are the expression of operation again.Because operation is only towards unique character, so a deletion action can only be represented with a position P, the deletion action that timestamp is SVOdel is represented: state vector of the website when this operation produces is SVOdel, the target of operation be deletion from a left side several P visual characters.And an insertion operation can be represented with inserting content C and insertion position P, the insertion operation expression that timestamp is SVOins: the state vector of the website when this operation produces is SVOins, the target of operation is that content C is inserted into from (needing to consider two special circumstances here: the left side of the first from left visual character, perhaps the right of the rightest visual character) between the several P-1 in a left side and P the visual character.

Because the existence of concurrent operations, when site-local received that a strange land is operated, the position of operation may be offset, and operation can't be carried out immediately.Address space conversion (Address Space Transformation) is by the address space of conversion document, state when the address space of at first recalling (Retrace) document produces to each operation, in the address space after conversion, position when the position in the document is represented with the operation generation represents it is consistent, operation can be carried out in new address space immediately, after carrying out, operation can recover the address space of document, to prepare the execution of next operation.

Under the timestamp pattern, guarantee that cause and effect consistance (Causality Preservation) only needs website of assurance to carry out the executive condition that operation is satisfied in a strange land operation: promptly each is operated all to have under the situation that guarantees causal sequence and carries out.Even but satisfy the causal sequence of operating, the order difference that operation arrives between different websites, the order of execution still can be different.We only need guarantee that it satisfies convergence (Convergence) and wish is safeguarded (Intention Preservation) when the executive condition of operation is satisfied in an operation.

One piece of linear structure that document is made up of character nodes, each character may have a plurality of operations at it, and each operation only can be at a character, so we are stored in the timestamp of operation together with it on the linear node at its pairing character place.On each node of the linear structure of document, except the character information and pairing a plurality of operation information of storing it, also has effective/invalid (Effective/Ineffective) flag information, represent whether current character nodes is visible on User, as shown in Figure 1.

The consistency maintenance engine of changing based on address space of the present invention is by control algolithm, and trace-back process and operation execution algorithm three parts are formed.Because the influence of concurrent operations when program is received an operated from a distance, is compared when the state of document produces with operation variation has been taken place, operation can't be carried out immediately.For correctness and the consistance that guarantees operating structure, we need at first recall the address space state of document.The process of recalling is according to the subsidiary timestamp of operation, set gradually the effective/disarmed state of each node in the document, make when to some extent effectively node produces with operation and be consistent, under this address space state, operation can be positioned between two effective nodes exactly, deletion action is carried out between can be directly, can obtain unique executing location altogether and insert operation according to the Range-Scan algorithm, revise local state vector then, and recover address space to treat the execution of next operation.Specifically describe as follows:

1, control algolithm

Control algolithm is an execution flow process to strange land user operation; After receiving a strange land user operation, if the executive condition of operation is satisfied in operation, then engine sends operation to the control algolithm execution, the timestamp that control algolithm is at first adhered to according to operation, call back-track algorithm and recall the address space of local data copy, carry out this operation according to action type then, operate the complete state vector of revising this locality afterwards according to the state vector maintaining method of this locality, recover the address space of local data copy at last, to treat the execution of next operation.Concrete operating process is as follows:

Suppose that current website is website S, its state vector is SVS, and corresponding document is DocS.The timestamp of operation O is SVO from the strange land of website R operation O when one, and is the executive condition that satisfies operation, and then the implementation at website S can be expressed as flow process:

The definition of position of operation, all be when producing by calculating operation in the linear structure on the left of the number of significant character node obtain.When the order of any two character nodes is unique, Retracing (DocS, state when SVO) having guaranteed that space, current address (the significant character node in the document linear structure) produces with operation is the same, so control algolithm can calculate correct significant character node in the time of can guaranteeing to operate execution.And after the state vector SVS adjustment, (DocS SVS) has guaranteed to recover the document status of address space after comprising new operation implementation effect to Retracing.

2, back-track algorithm

Back-track algorithm is the timestamp according to appointment, recall the address space of local data copy, back-track algorithm is investigated each Object node in the data trnascription, for certain Object node wherein, at the operation of this Object node, a counter is set, for the operation that continues before the cause and effect, if the wish of operation is to keep this Object node, then counter is added one; If the wish of operation is this node of deletion, then counter subtracts one; If last insert to be operating as continue before the cause and effect and counter is 0, this node then is set for effectively, be invalid otherwise this node is set; Comprehensive all Object nodes effectively/disarmed state, be the address space behind the recalling of local data copy.Concrete operating process is as follows:

If the document linear structure of website S is expressed as DocS, wherein the state vector of S is expressed as SVS, and establishing any one timestamp that satisfies causal sequence is SV, and then trace-back process can be expressed as following flow process:

The target of trace-back process is to recall the address space of document to given timestamp SV, promptly investigates each character nodes respectively, determines the effective/invalid flag of each character nodes under timestamp SV.Implementation is exactly to hide the implementation effect of all operations that produce after timestamp SV, only is retained in the implementation effect of the operation that produces before the SV.For each character nodes, suppose insertion operation and deletion action that a node adheres to, their timestamp is respectively: SVOins and SVOdel, in fact have only four kinds of situations to need to consider: (1) SVOins 〉=SV, then this node be set to invalid; (2) SVOins≤SV, and not at the deletion action of this character nodes, then this node is effective; (3) SVOins≤SV and SVOdel 〉=SV, this node is effective; (4) SVOins≤SV and SVOdel≤SV, this node are invalid.Utilize counter Counter and insert the timestamp state of operation, we can judge the situation of each character nodes exactly, thereby the effective/disarmed state of node is set.

3, operation execution algorithm

The operation execution algorithm finds a unique position for a strange land user's insertion operation on all data trnascriptions; An insertion operation is defined on the address space of operating the timestamp appointment of adhering to, is expressed as two positions between effective Object node; But still may exist a plurality of invalid operation between two effective Object nodes, the operation execution algorithm can be by the unique position relation of determining between new operation and a plurality of invalid operation of TOrder function.

If operation O is a deletion action, then only in address space, find the relevant position (counting the significant character node of appointment in the linear structure of DocS from left to right) of operation.

If being one, operation O inserts operation, in address space, find the relevant position of operation (to count the linear structure of DocS from left to right, find two adjacent significant character nodes of appointment), set up a new character nodes, the mark of initialization node also is inserted into definite position in the insertion scope to it.

Because Retracing (DocS, SVO) generation that address space is returned to operation constantly, the execution of deletion action is simple, only find corresponding character nodes, to operate attached to this character nodes, the trace-back process of Zhi Hanging gives correct effective/invalid flag can for this node thereafter.And insert operation is to seek corresponding insertion position, an interval between two significant character nodes in the corresponding document in address space.In this interval, may have other invalid character nodes in the document, we need a Range-Scan process determine new insertion operation insertion position and these idle character nodes about concern, to obtain a definite position.In order to realize this Range-Scan process, we have introduced a TOrder function, are used for obtaining the ordering relation of any two insertion operation rooms.Because deletion action is only revised label information, do not influence the order of character nodes, so the discussion below us is only around inserting operation.

Insert operation and can compare for two, we have used the skill [15] of keeping the operation conversion sequence among the OT to come definition of T Order relation, be used for representing two insert characters about relation, TOrder () less in the left side.Consider two character nodes CNa and CNb, the operation of its correspondence produces respectively from website a and b, and timestamp is respectively SVa and SVb, TOrder (CNa)＜TOrder (CNb) then, and if only if (1) sum (SVa)＜sum (SVb); Perhaps (2) a＜b when sum (SVa)=sum (SVb), wherein $\mathrm{sum}\left(\mathrm{SV}\right)=\underset{i=0}{\overset{N-1}{\mathrm{\Σ}}}\mathrm{SV}\left[i\right].$ TOrder () relation is an ordering relation (Totalordering), be (Transitivity) that transmits, and any two character nodes is comparable.Then Cao Zuo execution algorithm can be expressed as following flow process:

Among the present invention, local state vector maintaining method is in order to safeguard local state vector; Judge which operation executed by the state vector of this locality; The quantity that the state vector record is carried out from the operation of different websites, after a new operation was performed, just the operation number of executions with this website added one, and upgraded local state vector.

Among the present invention, broadcasting method is to give all other user with user's operational notification an of this locality, which operation executed is local state vector can judge, represented that equally also these operations and the new local operation of carrying out satisfy cause-effect relationship, so broadcasting method together is notified to other user with this state vector as the timestamp of operating, allow other user know the state of the data trnascription that this operation execution is.

Description of drawings

Fig. 1 is the storage organization diagram.

Fig. 2 is the sequential diagram of operation.

Fig. 3 is address space transfer process diagram.

Fig. 4 is the implementation diagram of operation.

Specific implementation method

Introduce whole consistency maintaining method in detail below in conjunction with an example.

Following example has been described the ﹠amp based on Mark; The processing procedure of the consistency maintenance engine of Retrace.

Suppose that a system has three user U1, U2 and U3, their one of together modify is initially empty shared document.The generation of operation and execution sequence are as shown in Figure 2.Here we only need to pay close attention to operation O4 and O5, and example shown in Figure 2 is used in still can be in the back.Operation O5 attempts character " c " is inserted into position 3 between " b " and " d ".We have a look user U1 and how to come executable operations O5---and " c " is inserted into position 3 with character.Because character " b " has been operated O4 and has deleted before this, position 3 representative now be the position on character " d " right side, rather than the left side of " d " originally.

The address space of address space conversion (Address Space Transformation) conversion document.Corresponding back-track algorithm, we see the front that user U1 is executable operations O5 directly, are because change has taken place the address space of document, the position of position 3 representatives state during with generation when operation O5 carries out is different.For can executable operations O5, the state of the address space that we need recall (Retracing) document when operation O5 produces, as shown in Figure 3.By this method, in the address space after the conversion, position 3 expressions be exactly position between character " a " and " d ", operate O5 and can in new address space, carry out immediately.For the conversion of supporting documentation address space, we have introduced mark (Mark) technology.Mark represents that a character nodes is effectively or invalid a given moment.

As shown in Figure 4, corresponding control algolithm is when user U1 attempts executable operations Insert[" c ", 3] time, the state vector of operation is＜1,1,2 〉, the state of document is shown in the picture left above among Fig. 4 of figure below.At first, control algolithm is by hiding operation Delete[" b ", 2] influence, document is dated back to constantly＜1,1,2 〉, the view of this moment is " abd ", illustrates as the right side.Then " c " is inserted between " b " and " d ".After carrying out the state vector of site-local be adjusted to＜2,1,2.Then document is returned to new state vector, i.e. recovery operation Delete[" b ", 2] influence and add operation Insert[" c ", 3] effect, document is reverted to view " acd ", shown in the lower-left figure among Fig. 4.

For a new operation O, in conjunction with the flow process in the Range-Scan algorithm, the process of determining an executing location is one, and from left to right each inserts the process of operation in this insertion scope of sequential scanning.In this scope, there are two classes to insert operation, a part of operation compare with operation O be causal sequence formerly, we stipulate that new insertion operation O necessarily is inserted in this part character that inserts operation left side, and when promptly scanning the character of this part operation, scanning process just can directly be through with.It is concurrent with operation O that another part inserts operation, operates us for this part and compare by TOrder () function.Algorithm finishes the insertion position, back and remains among the variable P, is about to the left side that new operation is inserted in position P.

Claims (7)

1, based on the consistency maintaining method of mark retrace, it is characterized in that making up the consistency maintenance engine of an application program, the control algolithm, back-track algorithm and the operation execution algorithm that comprise program in this engine, the state vector maintaining method and the broadcasting method that also comprise a this locality, make a plurality of users edit same data object in real time, by a kind of real-time collaborative technology of not having lock, do not influence each other between a plurality of users, thereby the operating collision between the automatic process user, the consistency maintenance of realization data between each user.

2, consistency maintaining method according to claim 1 is characterized in that described control algolithm is an execution flow process to strange land user operation; After receiving a strange land user operation, if the executive condition of operation is satisfied in operation, then engine sends operation to the control algolithm execution, the timestamp that control algolithm is at first adhered to according to operation, call back-track algorithm and recall the address space of local data copy, carry out this operation according to action type then, operate the complete state vector of revising this locality afterwards according to the state vector maintaining method of this locality, recover the address space of local data copy at last, to treat the execution of next operation.

3, consistency maintaining method according to claim 1, it is characterized in that back-track algorithm in the described engine is the timestamp according to appointment, recall the address space of local data copy, back-track algorithm is investigated each Object node in the data trnascription, for certain Object node wherein, at the operation of this Object node, a counter is set, for the operation that continues before the cause and effect,, then counter is added one if the wish of operation is to keep this Object node; If the wish of operation is this node of deletion, then counter subtracts one; If last insert to be operating as continue before the cause and effect and counter is 0, this node then is set for effectively, be invalid otherwise this node is set; Comprehensive all Object nodes effectively/disarmed state, be the address space behind the recalling of local data copy.

4, consistency maintaining method according to claim 1 is characterized in that the insertion operation of described operation execution algorithm for a strange land user, finds a unique position on all data trnascriptions; An insertion operation is defined on the address space of operating the timestamp appointment of adhering to, is expressed as two positions between effective Object node; But still may exist a plurality of invalid operation between two effective Object nodes, the operation execution algorithm can be by the unique position relation of determining between new operation and a plurality of invalid operation of TOrder function.

5, consistency maintaining method according to claim 1 is characterized in that the state vector maintaining method of described this locality, in order to safeguard local state vector; Judge which operation executed by the state vector of this locality; The quantity that the state vector record is carried out from the operation of different websites, after a new operation was performed, just the operation number of executions with this website added one, and upgraded local state vector.

6, consistency maintaining method according to claim 1, it is characterized in that broadcasting method in the described engine is to give all other user with user's operational notification an of this locality, which operation executed is local state vector can judge, represented that equally also these operations and the new local operation of carrying out satisfy cause-effect relationship, so broadcasting method together is notified to other user with this state vector as the timestamp of operating, allow other user know the state of the data trnascription that this operation execution is.

7, consistency maintaining method according to claim 4, it is characterized in that operating in the execution algorithm, the TOrder functions specify ordering relation of an operation room, for two operations, the TOrder function provides the magnitude relationship of two operations, the TOrder function has transitivity, and any two operations are comparable.

Images