CN101373509B - RFID intermediate part based on distributed ALE - Google Patents

Abstract

The invention provides an RFID middleware based on distributed ALE, which comprises a Global-ALE module, a Sub-ALE module and a connection pool module which are sequentially connected from top to bottom, wherein the Global-ALE module is connected with an application service logic layer in an upward direction; the connection pool module is connected with a reading tag layer in a downward direction; and the Sub-ALE module comprises a plurality of Sub-ALEs, each of which serves the ALE function, and the Sub-ALE has the mapping relationship with each other, that is, the RFID middleware adopts the same service rules. The RFID middleware adopts the distributed ALE system structure, makes full use of the distributed resources through the sub-mission distribution and the tag information distribution, and well solve the problem of the excessive load, and meanwhile, the RFID middleware keeps the consistency of a report, thereby realizing the transparent application layer.

Description

A kind of RFID middleware based on distributed ALE

Technical field

The invention belongs to RFID middleware Technology field, a kind of RFID middleware based on distributed ALE particularly is provided.

Technical background

RFID middleware (Radio Frequency Identification middleware) is identification and acquisition of information and be widely used in various fields automatically.The RFID middleware has been erected the bridge of data communication between automatic identification equipment (such as the RFID reader) and application system.Application-layer events (Application Level Event) normalized definition goes out one group of standard interface that the RFID middleware should provide the upper layer application system, and the basic functions of RFID middleware: collection/filtration (Collect/Filter).

Application-layer events (Application Level Event) standard is in 2005, by the formal externally issue of EPCglobal tissue.It defines one group of standard interface that the RFID middleware should provide the upper layer application system, and the basic functions of RFID middleware: collection/filtration (Collect/Filter).The ALE layer is between applied business logic and original tag reading layer.It receives the original tag data of sending from data source (one or more read write line), then, according to conditions such as time interval accumulative total (Accumulate) data, the EPCs that repeats or dare not interest rejected filter (Filter), simultaneously can count and make up operations such as (Count/Group), at last, these information are reported to application system.Its management equipment and subscribe to label data to equipment.

In the time of reader work, meeting is reading tag ceaselessly.Thereby, can cause same label within one minute, may be read tens times, if these data directly send to application program, will bring the very big wasting of resources.And the appearance of ALE mainly is in order to reduce the redundancy of raw data, extract effective service logic from mass data.

ALE (Application Level Event) standard has elaborated the framework of RFID middleware, as shown in Figure 1, in this framework, a plurality of physics readers is formed a logic reader.The just logic reader that ALE sees, and can be towards the physics reader.Too much when the number of tags that some logic reader reads, when a lot of logic readers were perhaps arranged, an ALE just can't handle so much label information, will cause delay or the fault of ALE.

Accompanying drawing 2 is depicted as distributed model, starts a plurality of ALE.New ALE that starts and original ALE are the reflection relations, and filtering rule, rule of classification that is to say them or the like all is the same.Then that original load is excessive ALE partial logic reader according to the loading condition of each logic reader, is distributed to the ALE of new startup.So just energy fast processing label information by filtration, packet label information, generates form and feeds back to application layer.

Though this model has solved the excessive problem of load, has caused other new problem.

The first, do not meet the principle that ALE removes data redundancy.Because reader is a reading tag information ceaselessly, a label may be read more than once by same or a plurality of readers.And go up in the surface model, the data processing between ALE1 and the ALE2 is separate, when a label all logic reader 0 and logic reader 2 read, they all can submit this label information to application layer.And application layer does not need the repetition label information.The major function of ALE is filtered these exactly and is repeated label, according to the many service logics of the rule refinement of application layer.

The second, this model is opaque to application layer.Suppose application layer to the ALE0 of first model and the ALE1 and the ALE2 of second model, import a same periods of events standard ECSpec (Event Cycle Specification) respectively, service logic definition just.Through after the regular hour, the form that they provide to application layer is respectively a form 0, and form 1, form 2. and form 1 and form 2 can not be fully integrated become form 0.That is to say that the user imports same service logic, but obtains different forms.

By last surface analysis, the reader that as seen has only the RFID middleware of an ALE to connect is too much, and perhaps the number of tags read of reader too much will cause the RFID middleware to postpone or collapse.If but with distributed form, start a plurality of ALE, and between each ALE, the logic reader is carried out simple load dispatch, can cause new problem again: do not meet ALE and remove the principle of data redundancy and opaque to application layer.Therefore the present invention proposes the method for designing of the distributed ALE in the RFID middleware: the mechanism of the ALE in the middleware is decomposed into a plurality of ALE executed in parallel, and introduces the mechanism of load balancing, coordinate the working load of each ALE.The scheme of distributed ALE in the function that has realized standard A LE regulation fully, is transparent to user and equipment control layer.

Summary of the invention

The objective of the invention is to solve the deficiency of existing method, distributed ALE architecture has been proposed, division by the subtask, the distribution of label information, make full use of distributed resource and come parallel computation, solved the excessive problem of load, also safeguarded the consistance of form simultaneously, thereby realized to application layer it being transparent.

In order to realize goal of the invention, the technical scheme of employing is as follows:

A kind of RFID middleware based on distributed ALE, comprise the Global-ALE module, Sub-ALE module and the connection pool module that connect successively from top to bottom, described Global-ALE module upwards is connected with the applied business logical layer, described connection pool module is connected with the label reading layer downwards, described Sub-ALE module comprises a plurality of Sub-ALE, each Sub-ALE has the function of ALE, and all is the reflection relation between each Sub-ALE, promptly adopts the same business rule.

Described Global-ALE module is provided with business logic modules, command interface module and form integration module.

Each Sub-ALE of described Sub-ALE module is provided with label processing module and load probe module.

Described connection pool module is provided with Sub-ALE pond and label pond, and the Sub-ALE pond administration module and the label pond administration module that are used to manage Sub-ALE pond, label pond.

Described middleware is provided with load-balancing mechanism, and this load-balancing mechanism comprises label distribution, load renewal, load dispatch.

Described label distribution comprises distribution label monocycle and cycle period;

The described distribution label monocycle is that all Sub-ALE have been carried out once complete label distribution, its all Sub-ALE to the Sub-ALE pond travel through, only available Sub-ALE is carried out distributing labels, in the time of available Sub-ALE, the label pond reading tag of arriving, if the label pond does not have label, just with the terminable recycle design of pausing slightly, inquire after the label pond, when there is label in the label pond, just once read K label, wherein K is the last threshold values that label takes place, if the number of tags of the inside, label pond is less than K certainly, then all remaining labels of reading tag pond when reading label, are just incited somebody to action all labels that read above, the network address according to current available Sub-ALE sends.

Described cycle period calls the monocycle with the terminable recycle design of pausing slightly, and each Sub-ALE is carried out distributing labels.

Described load is upgraded specific as follows:

Each available Sub-ALE is set up passage: just want to set up load channel with each Sub-ALE earlier, load information is separately in time fed back to make things convenient for each Sub-ALE;

Traversal Sub-ALE: all Sub-ALE to the Sub-ALE pond travel through, and only available Sub-ALE are carried out load and upgrade;

Set up passage: set up load channel with corresponding Sub-ALE,, then be arranged to this Sub-ALE unavailable if set up the load channel failure;

Monitor and the renewal load information: each Sub-ALE is carried out load monitor, when listening to load information, the network address and the corresponding CPU usage that comprise Sub-ALE, just according to IP network address, traversal finds corresponding Sub-ALE in the Sub-ALE pond, upgrades the CPU usage of this Sub-ALE then.

Described load dispatch is specific as follows:

The load dispatch initialization: mainly be the initialization load dispatch threshold values that will use and the parameter information of load, the identifier n=0 of initialization Sub-ALE wherein is used for all Sub-ALE of searching loop; Threshold values highLoad in the load is set, and acquiescence is 80, when the cpu busy percentage of Sub-ALE correspondence is higher than highLoad, represents that this Sub-ALE load is excessive exactly; Threshold values lowLoad under the load is set, and acquiescence is 20, when the cpu busy percentage of Sub-ALE correspondence is lower than lowLoad, represents that this Sub-ALE is in idle condition exactly; The number k=0 of available Sub-ALE is set, is convenient to add up the number of current available Sub-ALE; Setting is respectively for statistical information greater than the number hn=0 of the Sub-ALE of highLoad with less than the number ln=0 of the Sub-ALE of lowLoad;

Load analysis: on initialized basis, add up the number k of available Sub-ALE, greater than the number hn of the Sub-ALE of highLoad with less than the number ln of the Sub-ALE of lowLoad;

Load dispatch: according to load analysis, call dispatching algorithm, operation dispatching, concrete dispatching algorithm is: when the number k of available Sub-ALE equals number hn greater than the Sub-ALE of highLoad, if also have the Sub-ALE that does not start, then start a Sub-ALE; When the number k of available Sub-ALE equals number ln less than the Sub-ALE of lowLoad, if the Sub-ALE that has started in addition then stops a Sub-ALE.

Described middleware is provided with report form integration mechanism, by the report form integration module of Global-ALE the form that each Sub-ALE sends is up integrated, and is specific as follows:

Form is monitored: at first read configuration information, obtain the number k of Sub-ALE; Being algebraically maker of each Sub-ALE initialization, and algebraically is initialized as 0, is to suspend terminable recycle design slightly, to monitor the form of each Sub-ALE with that;

Report form processing:, temporarily store the form Buffer Pool into to the form that listens to; When the form in this generation is all here, just carry out report form integration; When producing total form, just carry out peacekeeping data consistency module; At last, should send to the applied business logical layer by total form;

The form storage: when listening to form, call the algebraically maker of the Sub-ALE correspondence that sends this form, obtain an algebraically identifier g, g is unique round values; This algebraically identifier is composed to this form; This form is stored in the form Buffer Pool;

Report form integration: traversal form Buffer Pool, the statistical algebra identifier is the number n of the form of g; If n equals the number k of available Sub-ALE, then with form Buffer Pool the inside, all algebraically identifiers are that the form of g extracts, and integrate, and obtain total form;

Carry out the service data consistance at last and should send to the applied business logical layer by total form.

Described middleware is provided with the conforming mechanism of service data, and is specific as follows:

Initialization: read total form, and empty a periods of events tag set;

Tag extraction: read each label of total form, and delete former record; And judge whether this label is present in the periods of events tag set,, just skip this label, if there is no, just this label is added to a last periods of events tag set if exist;

A synchronous last periods of events label: when all labels, all put in order, just, call, be synchronized among each Sub-ALE by network interface with a last periods of events tag set to behind the last periods of events tag set with total form.

A kind of RFID middleware method that the present invention is designed based on distributed ALE, its functional characteristics can be subdivided into following several aspect:

1, reader " plug and play ": in the architecture of distributed ALE, owing to introduced connection pool, making this structural system clear layer, is mutual transparent between layer and the layer.Because distributed ALE is transparent to the variation of reader, so reader has very strong dirigibility.To the reader start and stop very neatly that connect, and distributed ALE is not affected.To dynamic interpolation or remove reader, only need to set up with connection pool or close the label passage, and do not influence other runnings of distributed ALE.

2, Sub-ALE (comprising tag processor and load probe) " plug and play ": each Sub-ALE is relatively independent in distributed ALE system architecture, and is lower with the coupling of each module.Therefore can dynamically increase or delete Sub-ALE.Sub-ALE and Global-ALE related has only command channel and form passage, has only label passage and load balancing passage with connection pool related.So dynamically increase or deletion Sub-ALE the time, only need to set up or close top four passages to get final product.

3, intelligence distribution label: owing to label all is stored in the label pond according to corresponding storage algorithm, and there are the network address, cpu busy percentage and the available identifiers of each Sub-ALE in the inside, Sub-ALE pond.Label pond administration module can be selected the label distribution algorithm, gives each Sub-ALE with the label distribution in label pond.The label distribution algorithm has two kinds: one is exactly all available Sub-ALE to the Sub-ALE pond carries out poll and sends label; Two is exactly the cpu busy percentage according to all available Sub-ALE in Sub-ALE pond, sets up priority; Cpu busy percentage is high more, and priority is low more; Priority is low more, and the number of tags of primary distribution is few more.So just can be well carry out distributing labels, thereby realize good load balancing according to the loading condition of each Sub-ALE.

4, dynamic load leveling: the loading condition of this machine of load probe module load sensing of each Sub-ALE, the cpu busy percentage of this machine just, and this value sent to connection pool.The Sub-ALE pond administration module of connection pool receives the load information of each Sub-ALE, and is updated to the Sub-ALE pond.Sub-ALE pond administration module is provided with threshold values (Low Load) under threshold values in the load (High Load) and the load earlier; When the cpu busy percentage of Sub-ALE surpasses last threshold values, just represent that this Sub-ALE is in high capacity, when the cpu busy percentage of Sub-ALE is lower than threshold values down, just represent that this Sub-ALE is in idle condition.Sub-ALE pond administration module can start and stop each Sub-ALE according to corresponding loads-scheduling algorithm.

5, integrate form.

6, service data consistance.

Description of drawings

Fig. 1 is the framework synoptic diagram of existing RFID middleware;

Fig. 2 is the Distributed Architecture synoptic diagram of existing RFID middleware;

Fig. 3 is the general frame logical schematic based on the RFID middleware of distributed ALE;

Fig. 4 is the schematic flow sheet of the label distribution of distributed ALE;

Fig. 5 is the schematic flow sheet that the load of distributed ALE is upgraded;

Fig. 6 is the schematic flow sheet of the load dispatch of distributed ALE;

Fig. 7 is the schematic flow sheet of the report form integration of distributed ALE;

Fig. 8 is the conforming schematic flow sheet of the service data of distributed ALE.

Embodiment

Below in conjunction with accompanying drawing, concrete enforcement of the present invention is described in further detail.

Accompanying drawing 3 has showed that the module that this framework comprises has Global-ALE based on the general frame of the RFID middleware of distributed ALE, Sub-ALE, connection pool (Connection Pool).Specific as follows:

Global-ALE is responsible for the request of response application layer and coordinates each Sub-ALE.

Global-ALE can accept the input of the business rule of application layer, and also gives each Sub-ALE synchronously with these business rules.These business rules comprise the filtration label mode, and the packet label pattern is obtained the frequency of form and defined statement form or the like.Because handling the foundation of label information is business rule, and the business rule of each Sub-ALE has all been realized synchronously, think that the consistance of form lays a good foundation.

Application layer can be mutual by the interface and the Global-ALE that call Global-ALE.In mutual, several fundamental method: define/undefine are arranged, subscribe/unsubscribe, poll/Immediate with Global-ALE.So Global-ALE behind the interactive command of response application layer, should call by network interface, coordinates the response of each Sub-ALE, thus the consistance of maintenance command operation.

At the business rule that has received application layer and through after the tag processes of certain hour, Global-ALE should provide a form to application layer.And each Sub-ALE has handled label information and has produced subreport according to business rule.So Global-ALE only needs the integrated application layer of issuing afterwards of these subreports.

Sub-ALE is responsible for receiving label information from connection pool, behind the filter packets label information, the data after handling is submitted to Global-ALE.

Each Sub-ALE has the function of original ALE.And all the reflection relation between each Sub-ALE.The business rule that is them all is the same.So connection pool sends label information to any one Sub-ALE, the result of processing is the same.Therefore can come parallel processing by label information being given a plurality of Sub-ALE, just can solve the excessive problem of load.Each Sub-ALE has a load probe module, and it is responsible for detecting the loading problem of current processor.The index that detects load is the utilization rate of CPU.

Connection pool (Connection Pool) is responsible for receiving the label information that reader sends, and is saved in the label pond (EPC Pool), according to schedule information label is sent to each Sub-ALE then.

The main modular of connection pool comprises label pond (EPC Pool), Sub-ALE pond, label pond management (EPC Pool Manager) and Sub-ALE pond management (Sub-ALE Pool Manager).The label information that all readers read has been preserved in the label pond, and its realization mechanism is the formation of FIFO.Sub-ALE has preserved in the pond information of all Sub-ALE, and its realization mechanism is the single cycle chained list.Each node is described a Sub-ALE.Wherein identifier is represented the sequence number of this Sub-ALE; IP represents the address network address of this Sub-ALE; The value of Is Used is that True represents that current this Sub-ALE moves, otherwise is exactly idle; Cpu busy percentage represents to move the utilization rate of the processor of this Sub-ALE, and it has two static constants: go up threshold values (HighLoad) and following threshold values (Low Load).Cpu utilization value is greater than the value of High Load, and the load of expression current processor is excessive; Cpu utilization value is represented the current processor free time less than the value of Low Load; The management of label pond is responsible for receiving label information there from reader and is saved in the label pond, and the label information in reading tag pond sends to the Sub-ALE that each has been enabled.The information of the load probe of each Sub-ALE of reception is responsible in the management of Sub-ALE pond, and updates to Sub-ALE, carries out the load dispatch program according to these probe message at last.

Fig. 4 has showed the subfunction of the load balancing of distributed ALE: the schematic flow sheet of label distribution.Label distribution is being served as the role who distributes load in fact, and it is with whole loads of connection pool the inside, and just all label informations according to certain allocation algorithm, are distributed to each Sub-ALE and handled.Solved the load assignment problem like this, just laid a good foundation for load balancing.Wherein label distribution is specific as follows:

Monocycle: distribution label.Just all Sub-ALE have been carried out once complete label distribution.

Traversal Sub-ALE: all Sub-ALE to the Sub-ALE pond travel through, and only available Sub-ALE are carried out distributing labels.

Reading tag: in the time of available Sub-ALE, the label pond reading tag of arriving.If the label pond does not have label,, inquire after the label pond just with the terminable recycle design of pausing slightly.When there is label in the label pond, just once read K label (wherein K is the last threshold values that label takes place), if the number of tags certainly inside the label pond is less than K, all labels of being left of reading tag pond then.

Send label: when reading label, just all labels that will read above according to the network address of current available Sub-ALE, send.

The recursive call monocycle: with the terminable recycle design of pausing slightly, call the monocycle here, each Sub-ALE is carried out distributing labels.

Accompanying drawing 5 has been showed the subfunction of the load balancing of distributed ALE: the schematic flow sheet that load is upgraded.It is exactly load information with each Sub-ALE that load is upgraded, and the utilization factor of CPU just in time feeds back to the Sub-ALE pond administration module of connection pool.For load dispatch provides load information timely, thereby for realizing that dynamic load leveling provides the data foundation.Load is upgraded specific as follows:

Each available Sub-ALE is set up passage: just want to set up load channel with each Sub-ALE earlier, load information is separately in time fed back to make things convenient for each Sub-ALE.

Traversal Sub-ALE: all Sub-ALE to the Sub-ALE pond travel through, and only available Sub-ALE are carried out load and upgrade.

Set up passage: set up load channel with corresponding Sub-ALE here,, then be arranged to this Sub-ALE unavailable if set up the load channel failure.

Monitor and the renewal load information: each Sub-ALE is carried out load monitor, when listening to load information (network address and the corresponding CPU usage that comprise Sub-ALE), just according to IP network address, traversal finds corresponding Sub-ALE in the Sub-ALE pond, upgrades the CPU usage of this Sub-ALE then.

Accompanying drawing 6 has been showed the subfunction of the load balancing of distributed ALE: the schematic flow sheet of load dispatch.Load dispatch is exactly the loading condition according to each Sub-ALE, calls all Sub-ALE of dispatching algorithm and dispatches.Load dispatch is specific as follows:

Load dispatch initialization: mainly be the initialization load dispatch threshold values that will use and the parameter information of load.The identifier n=0 of initialization Sub-ALE wherein is used for all Sub-ALE of searching loop; Threshold values highLoad in the load (acquiescence is 80) is set, when the cpu busy percentage of Sub-ALE correspondence is higher than highLoad, represents that this Sub-ALE load is excessive exactly; Threshold values lowLoad under the load (acquiescence is 20) is set, when the cpu busy percentage of Sub-ALE correspondence is lower than lowLoad, represents that this Sub-ALE is in idle condition exactly; The number k=0 of available Sub-ALE is set, is convenient to add up the number of current available Sub-ALE; Setting is respectively for statistical information greater than the number hn=0 of the Sub-ALE of highLoad with less than the number ln=0 of the Sub-ALE of lowLoad.

Load analysis: this module is on the initialized basis of upper module, adds up the number k of available Sub-ALE, greater than the number hn of the Sub-ALE of highLoad with less than the number ln of the Sub-ALE of lowLoad.

Load dispatch: this module is called dispatching algorithm according to top load analysis, operation dispatching.Dispatching algorithm for example: when the number k of available Sub-ALE equals number hn greater than the Sub-ALE of highLoad,, then start a Sub-ALE if also have the Sub-ALE that does not start; When the number k of available Sub-ALE equals number ln less than the Sub-ALE of lowLoad, if the Sub-ALE that has started in addition then stops a Sub-ALE.

Accompanying drawing 7 has been showed the schematic flow sheet of the report form integration of distributed ALE.Report form integration is by the report form integration module of Global-ALE the form that each Sub-ALE sends up to be integrated.Specific as follows:

Form is monitored: at first read configuration information, obtain the number k of Sub-ALE; Be algebraically maker of each Sub-ALE initialization, and algebraically is initialized as 0.Be to suspend terminable recycle design slightly, to monitor the form of each Sub-ALE with that.

Report form processing:, temporarily store the form Buffer Pool into to the form that listens to; When the form in this generation is all here, just carry out report form integration; When producing total form, just carry out peacekeeping data consistency module; At last, should send to the applied business logical layer by total form.

Form storage: when listening to form, call the algebraically maker of the Sub-ALE correspondence that sends this form, obtain an algebraically identifier g (unique round values); This algebraically identifier is composed to this form; This form is stored in the form Buffer Pool.

Report form integration: traversal form Buffer Pool, the statistical algebra identifier is the number n of the form of g; If n equals the number k of available Sub-ALE, then with form Buffer Pool the inside, all algebraically identifiers are that the form of g extracts, and integrate, and obtain total form;

Service data consistance: labor below;

Send form: should send to the applied business logical layer by total form.

Accompanying drawing 8 has been showed the conforming schematic flow sheet of the service data of distributed ALE.All Sub-ALE must share all label informations that read during the periods of events, are called previousEPCSet.Be exactly total the overall just previousEPCSet of the set of all labels of the form of Global-ALE.Therefore Global-ALE adds all label informations of total form the previousEPCSet of Global-ALE to, gives each Sub-ALE synchronously with this previousEPCSet then.Specific as follows:

Initialization: read total form, and empty a periods of events tag set.

Tag extraction: read each label of total form, and delete former record; And judge whether this label is present in the periods of events tag set,, just skip this label, if there is no, just this label is added to a last periods of events tag set if exist.

A synchronous last periods of events label: when all labels, all put in order, just, call, be synchronized among each Sub-ALE by network interface with a last periods of events tag set to behind the last periods of events tag set with total form.

Claims (3)

1. RFID middleware based on distributed ALE, it is characterized in that comprising the Global-ALE module, Sub-ALE module and the connection pool module that connect successively from top to bottom, described Global-ALE module upwards is connected with the applied business logical layer, described connection pool module is connected with the label reading layer downwards

Described Sub-ALE module comprises a plurality of Sub-ALE, and each Sub-ALE has the function of ALE, and all is the reflection relation between each Sub-ALE, promptly adopts the same business rule; Described Global-ALE module is provided with business logic modules, command interface module and form integration module, each Sub-ALE of described Sub-ALE module is provided with label processing module and load probe module, described connection pool module is provided with Sub-ALE pond and label pond, and the Sub-ALE pond administration module and the label pond administration module that are used to manage Sub-ALE pond, label pond;

Described Global-ALE is responsible for the request of response application layer and coordinates each Sub-ALE, also provide a form to application layer by the form integration module, the Global-ALE module is accepted the input of the business rule of application layer, and give each Sub-ALE synchronously with these business rules, described Sub-ALE module is responsible for receiving label information from connection pool, behind the filter packets label information, handle by label processing module, and the data after will handling are submitted to the Global-ALE module, each Sub-ALE is provided with a load probe module, be responsible for detecting the loading problem of current processor, the index of the load that detects is the utilization rate of CPU, described connection pool module is responsible for receiving the label information that reader sends, and is saved in the label pond, according to schedule information label is sent to each Sub-ALE then.

2. the RFID middleware based on distributed ALE according to claim 1 is characterized in that described middleware is provided with load-balancing mechanism, and this load-balancing mechanism comprises label distribution, load renewal, load dispatch.

3. the RFID middleware based on distributed ALE according to claim 2 is characterized in that described load dispatch is specific as follows:

The load dispatch initialization: mainly be the initialization load dispatch threshold values that will use and the parameter information of load, the identifier n=0 of initialization Sub-ALE wherein is used for all Sub-ALE of searching loop; Threshold values highLoad in the load is set, and acquiescence is 80, when the cpu busy percentage of Sub-ALE correspondence is higher than highLoad, represents that this Sub-ALE load is excessive exactly; Threshold values lowLoad under the load is set, and acquiescence is 20, when the cpu busy percentage of Sub-ALE correspondence is lower than lowLoad, represents that this Sub-ALE is in idle condition exactly; The number k=0 of available Sub-ALE is set, is convenient to add up the number of current available Sub-ALE; Setting is respectively for statistical information greater than the number hn=0 of the Sub-ALE of highLoad with less than the number ln=0 of the Sub-ALE of lowLoad;

Load analysis: on initialized basis, add up the number k of available Sub-ALE, greater than the number hn of the Sub-ALE of highLoad with less than the number ln of the Sub-ALE of lowLoad;

Load dispatch: according to load analysis, call dispatching algorithm, operation dispatching, concrete dispatching algorithm is: when the number k of available Sub-ALE equals number hn greater than the Sub-ALE of highLoad, if also have the Sub-ALE that does not start, then start a Sub-ALE; When the number k of available Sub-ALE equals number ln less than the Sub-ALE of lowLoad, if the Sub-ALE that has started in addition then stops a Sub-ALE.

Images