The content of the invention
For technical problem present in prior art, it is an object of the invention to provide a kind of data transfer of regular drive
Method.The present invention by the classification of distributed tasks characteristic attribute and represent the mapping of mechanism, characteristic attribute and business rule with
Task scheduling transfer process with mechanism and rule-based driving, accurately to portray the characteristic attribute of distributed tasks, realizes machine
The decomposition of the intelligible business rule of device, identification and explain, meet distributed tasks under optimal scheduling characteristic attribute combination and
Business rule is extracted.
Below in conjunction with the accompanying drawings 2 couples of present invention are described in further detail.
The data transmission method of the present invention includes following step:
Step S2.1:Definition distribution task Characteristic Attribute Classification and expression.Fine granularity represents the feature of distributed tasks
And constraint, and expressed by resource description framework (RDF) form, it is convenient to recognize and rule-based reasoning.
Step S2.2:Characteristic attribute is recognized.The characteristic attribute defined by system, extracts the static and dynamic of distributed tasks
Characteristic attribute, and with characteristic attribute element and constraint representation.It is intrinsic or indirect, static or dynamic to an attribute, point
Do not divided and marked from 2 dimensions.
Step S2.3:Automatically learn and set business rule.It is industry by series of features attribute induction by statistical analysis
Business rule, or characteristic attribute is combined as by business rule by user interface.Business rule is made up of series of features attribute, than
Such as generated by man-machine interaction guidance, machine be appreciated that it is executable;It learns and sets the phase of business rule and characteristic attribute automatically
Mutually identification conversion is embodied in two aspects:One is characteristic attribute threshold value dynamic adaptation, and two is the dynamic combined of attribution rule, mainly
It is optimized by the assessment of considerable task scheduled transmission performance statistics.Such as, " satellite broadcasting net is used more than k receiving node "
With " using broadband networks more than kM data ", two rules are by according to the coverage rate of satellite network and broadband networks, bandwidth, utilization rate statistics
Situation and current transmission goal constraint enter Mobile state assignment to k.
Step S2.4:Business rule chain is parsed.In task scheduled transmission, according to the business rule chain that system is provided, enter
Row function decomposition into analytic function forms characteristic attribute, and matches with the constraints of distributed tasks.If matching arbitrary rule, rule are jumped out
Then chain, using the network of matching network transmission is performed.The constraints of distributed tasks be data involved by the task, node and
What the static attribute of channel three elements was determined.Such as one distributed tasks are related to certain data and have certain level of confidentiality, certain being related to
Node is only connected with local channel.It is just natural when being so scheduled optimization to the task to consider that meeting level of confidentiality requires and available
Channel in selected, two such constraint.
Step S2.5:The characteristic matching of distributed tasks.The optimal network of satisfaction is not such as matched using business rule chain, then
Matched with distributed tasks constraints according to the characteristic attribute of data, channel and node, formed and meet optimum network biography
Transmission scheme.The class method of matching way Main Basiss two:The first kind is by pre-filtering form, by the spy of data, channel and node
Levy attribute and compared filtration one by one with the constraints of distributed tasks, such as the safe class of data is higher in distributed tasks,
Then compare which channel in existing channel possesses high safety grade, then filter out.Equations of The Second Kind is, according to optimum target, to select most
Excellent transmission network.Network such as most short according to transmission time, selecting current task transmission time most short.By this two class side
Method, ultimately forms and meets optimum network transmission scheme.Node occupies multiple channels, and the coverage of channel is multiple nodes.
Because data are self-existent, only when data is transferred, just need to set up node, data, the contact of the ternary of channel;Node
Other nodes are transferred data to by channel.
Step S2.6:The amendment of business rule.Judge whether optimum using the business rule by business personnel.The such as biography
Defeated not up to predeterminated target, then start study or human users' mode automatically, corrects business rule.Business rule chain is expressed as one
The combination of serial equation or inequality.Automatically amendment mainly complete to the numerical value on the right of the equation or inequality in rule chain or
Symbol is modified.Such as transmission objectives nodal point number is more than 50 then using satellite network transmission.This business rule can be represented
For Sum (nodal point number)>50, transmission network=satellite network.This numerical value 50 can automatically learn amendment.It is i.e. big by statistical analysis
The impact of 50 this numerical value of the state of amount task and modification, finally updating this numerical value, is such as modified to 60, or 100.
Step S2.7:Flow process terminates.Optimum distributed tasks scheduled transmission is provided.Key point and corresponding technique effect.
For a transformation task, the real-time characteristic of channel between node need to be considered.Therefore present invention introduces node
It is right, as shown in Figure 3.Each node is used being made up of two nodes<The superior and the subordinate describe>Feature is defining the biography between node
Defeated orbution, i.e., described using (node is to title, superior node, downstream site, busy channel).In transformation task, make
With<Transmission direction>Feature come determine specific transmission direction, i.e. recipient and sender description.So as to ternary contact be converted
Contact for the binary between channel and data;Wherein, by a node to two connection nodes in table record data transmission procedure
Transmission the superior and the subordinate of composition, its field includes:Attribute, the data type of attribute, property value length, whether it is major key, Ke Fouwei
Empty, description explanation.Data transfer task entity relationship diagram is as shown in Figure 4.Because individual node can participate in multiple nodes pair, together
When one node to comprising Liang Ge the superior and the subordinate node.Therefore contingency table is solved between node and node pair using node-node
Many-to-many relationship.
Key point 1:The classification of distributed tasks characteristic attribute and expression mechanism.The main body of one distributed tasks include data,
Channel and node.The characteristic attribute of each main body includes static nature and behavioral characteristics again.The static nature of data includes data
Type, data name, size of data, data safety grade;The behavioral characteristics of data include data reliability requirement, data distribution
Time limit requirement, data genaration time etc..The static nature of channel includes channel capacity, channel safety grade, channel type;Channel
Behavioral characteristics include channel available width, channel reliability, channel loading.The static nature of node includes transfer capability, section
Point grade;The behavioral characteristics of node include that node state, node state are fed back.Unify in the characteristic attribute to distributed tasks
On the basis of description, the present invention carries out unified representation based on RDF for each characteristic attribute, including element definition, attribute definition etc.,
Refer to table 1.
The meteorological data transmission feature table of table 1
Classified by the characteristic attribute to distributed tasks, the distribution constraints of arbitrary distributed tasks can be converted into
Constraint to characteristic attribute.That is the constraints of a distributed tasks is represented by the condition of series of features attribute about
Beam.
Technique effect:Classification and expression mechanism by distributed tasks characteristic attribute, distributed tasks is expressed as a series of
The characteristic attribute constraint represented with RDF.
Key point 2:Mapping and matching mechanisms of the characteristic attribute with business rule.Business rule is that people is intelligible, is used for
Represent the historical experience and professional knowledge of distributed tasks.By defining business rule, and it is construed to characteristic attribute and its condition about
Beam, the machine for being capable of achieving business rule is appreciated that and performs.Such as " live data is transmitted using special line " this business rule
It is represented by:
Live data is transmitted using special line
Data inactivity feature:Data type:Live data
Channel:Channel type:Special line
By the mapping of characteristic attribute and business rule with match, business rule can be decomposed into characteristic attribute.
Business rule is often complex, can be expressed by being construed as a series of characteristic attribute of combinations.Such as " peace
The high data of congruent level use the high network transmission of safe class ", the rule can carry out table using the combination of series of features attribute
Show.
The high data of safe class use the high network transmission of safe class:
Data safety grade=1, channel safety grade>=1;&&
Data safety grade=2, channel safety grade>=2;&&
Data safety grade=3, channel safety grade>=3;&&
Data safety grade=4, channel safety grade>=4;&&
Characteristic attribute is allowed by interface definition mode with the mapping of business rule with matching mechanisms, can be by combinations of features shape
Into rule.
Technique effect:Service feature to rule mutual mapping with match, realize business rule to the knowledge of series of features
Other and decomposition;The combination of characteristic attribute forms new business rule.
Key point 3:The task scheduling flow process of rule-based driving.In task scheduling process, according to the industry of system definition
Business rule, is matched with the constraint of task.Business rule with the presence of rule chain form, i.e., in order (advise by matched rule
Then the order of chain embodies the priority between rule;Rule chain setting principle of the present invention is high priority distribution>Meet daily
Distribution of services demand>The business rule of network transmission can be optimized.By this order, on the basis of disclosure satisfy that business demand, to the greatest extent
Amount is increased network utilization.), if matching any bar, the defined operation of executing rule (is assigned to which network is performed
Deng).Such as the process is not matched to arbitrary rule, then matched with the attributive character of current data, channel and node, selects
Optimum network is transmitted.The distributed tasks of arbitrary rule, the combination die of automatic sorting characteristic attribute are not matched by statistics
Formula, learns new business rule, operates into knowledge base for business personnel.Such as counted according to historic task, extract distributed tasks
The characteristic value of adopted network, sets the characteristic attribute in business rule, that is, form new business rule.As two-level node exceedes
20, and distribute destination node more than 100, distributed tasks of the size of data less than 1M transmit this business using satellite network
Rule.Two-level node, distribution node, size of data, distribution network these be characterized in that need it is fixed.
Technique effect:Task scheduling is distributed based on business rule, and can learn to sum up business rule automatically.
Compared with prior art, the positive effect of the present invention:
1) present invention by the Characteristic Attribute Classification of distributed tasks and identify, the mapping of characteristic attribute and business rule and
With and the data transfer that drives of business rule, realize that distributed tasks constraint, business rule and characteristic attribute are triangular mutually
Mapping matching, meets distributed tasks and forms optimized scheduling according to current task constraints matching business rule.
2) present invention realizes feature by study automatically and the correction mechanism of artificial setting business rule and business rule
The statistical induction to business rule of attribute, and business rule explains that execution is characterized attribute automatically, facilitates business personnel's root
According to self-defined or adjustment transmission rule is needed, optimize overall task scheduled transmission.
Specific embodiment
Illustrate the realization of the inventive method with enforcement below.
Different distributed tasks have different distribution constraintss, the such as high data of safe class, and acquiescence needs to use
The high channel of safe class carries out data transmission;Some distributed tasks need within a specified time to be sent to all destination nodes
Deng.The constraints of these distributed tasks can be by the task of characteristic attribute, business rule parsing and conversion and regular drive
Dispatch the optimal scheduling to reach distributed tasks to transmit.
Such as according to business experience, general two-level node distributes destination node more than 100 more than 20, size of data
Distributed tasks less than 1M are transmitted using satellite network.The example is to summarize the experience out by substantial amounts of distribution, but in business
The automatic summary aspect of rule, then need to be defined characteristic attribute and classify, and returns according to substantial amounts of distributed tasks statistics
Receive and draw.
In the case, there is the characteristic attribute of data, channel and node, i.e. size of data, node level, number of nodes
(characteristic attribute table is unlisted, but it is the statistical value of characteristic attribute), the node level quantity (characteristic attribute table is unlisted, but
It is the statistical value of characteristic attribute), channel type.First by definition distribution task solid data, channel and node it is quiet
State and behavioral characteristics, these characteristic attributes are identified and are classified.As can be seen that size of data, node level, channel type
Belong to static nature and intrinsic characteristics.Intrinsic characteristics are direct features, can artificially judge or directly gather, and codomain is fixed, side
Just calculate.
After program starts, substantial amounts of history distributed tasks are carried out with constraints decomposition, and are mapped as characteristic attribute and value,
Carry out statistical analysis.The statistical analysis process is analyzed for all tasks of satellite network distribution, finds these tasks
Common trait attribute.It was found that meeting the distributed tasks of following situation, then task distribution is carried out by satellite network:
Size of data<1M,
Node level=bis- grade,
SUM (node)>100,
SUM (node level=bis- grade)>20
The attributive character drawn by statistical analysis and value, automatically study is business rule, and is set.The business rule
" must the using satellite network transmission of the task " can be then named as, and with following rule specifics:
Size of data<1M, channel type=satellite network, &&
Node level=bis- grade, channel type=satellite network, &&
SUM (node level=bis- grade)>20, channel type=satellite network, &&
SUM (node)>100, channel type=satellite network, &&
As can be seen that the business rule is completed by formulating and combining a series of characteristic attribute.
In actual dispensed task process, when distributed tasks are reached, with current business rule chain (by a series of business
Rule composition) matched in order.As current task meets a business rule therein, if network after the match is successful
State meets condition, then perform corresponding scheduling operation, and suitable network is otherwise matched again.Such as it is currently needed for the distribution dispatched
After the constraints of task is by decomposing mapping, its characteristic attribute has:Size of data is 512K, and the number of nodes of distribution is 200
Individual, the quantity of two-level node is 50.After matching by business rule chain, it is found that business rule " must be passed using satellite network
The match is successful for defeated task ", then obtains the state of satellite network, and distributed tasks are dispatched to into satellite network if meeting,
Carry out data transmission, suitable network is otherwise matched again.
As current distributed tasks do not match arbitrary business rule, the constraints of such as current distributed tasks is reflected by decomposition
After penetrating, it is 512K to be expressed as size of data, and the number of nodes of distribution is 200, and the quantity of first nodes is 10, two-level node
Quantity be 15.Obviously it can not be matched completely with business rule " must being transmitted using satellite network for task ".Then start special
Matching mechanisms are levied, data, node and channel characteristics are matched with distributed tasks constraint, select suitable network to be passed
It is defeated.It is most short for optimized scheduling target that the matching process can select transmission time, selects suitable networks to be transmitted.
The difficulty action accomplishment of the distributed tasks is marked, to correct business rule;If transmitted using the business rule, failure time
Number is continuously more than set point number, the codomain for relearning each characteristic attribute is needed, to correct business rule.By substantial amounts of biography
Defeated discovery, size of data is 512K, and the number of nodes of distribution is 200, and the quantity of first nodes is 10, the number of two-level node
Measure and still select satellite network as best transmission network for 15 such transformation tasks.By given threshold or manually sentence
Disconnected mode, can correct " must the using satellite network transmission of the task " this business rule, be expressed as:
Size of data<1M, channel type=satellite network, &&
Node level=bis- grade, channel type=satellite network, &&
SUM (node level=bis- grade)>10, channel type=satellite network, &&
Node level=one-level, channel type=satellite network, &&
SUM (node level=one-level)>10, channel type=satellite network, &&
SUM (node)>100, channel type=satellite network, &&
Such as above-mentioned example, the optimization task scheduling for being possible to realize historical experience and the combination of real-time status feature of the invention
Distribution transmission.
The above, only presently preferred embodiments of the present invention is not intended to limit protection scope of the present invention.