CN102122252A - Method for designing thread pool capable of ensuring temporal succession - Google Patents

Abstract

The invention relates to a method for designing a thread pool capable of ensuring temporal succession. The method comprises the following steps of: maintaining a task for the thread pool and marking the task into to a Hash mapping table of a thread ID, for managing the task distribution; adding an attribute parameter to thread parameters: a load (nLoad) used for identifying the total quantity of the distributed tasks of each thread; and instead of sharing the same task queue by all the threads, configuring a corresponding task queue for each thread in the thread pool, wherein each thread can search for a new task from the corresponding task queue, and a single monitoring thread is arranged. The method provided by the invention has the advantages that: 1) by utilizing the new added function of the mapping table, the temporal successions for inputting and outputting data of each task can be guaranteed to be completely consistent, and 2) the loads on all threads can be guaranteed to be always consistent by the monitoring thread started at regular time, thereby achieving the best concurrent property.

Description

A kind of thread pool method for designing that guarantees time continuity

Technical field

The present invention relates to a kind of method for designing of thread pool, refer more particularly to the thread pool method for designing that guarantees time continuity.

Background technology

Thread pool is a kind of processing form of multithreading.Thread pool can be pre-created the task of some threads with executive utility, and these threads are normally with the form tissue of formation.Generally, the task quantity that need carry out of thread pool is greater than the quantity of thread in the thread pool.After certain thread had been finished a task, can please looking for novelty in task queue of task was carried out, and the task in task queue is all finished.Afterwards the thread in the thread pool will be suspended or dormancy up to there being new task to arrive.

Traditional thread pool all can have the difference of priority for carrying out of task, is complete fair play for the task of same priority.But equality and be not suitable for all application scenarioss completely.For example in real-time data base, for all free successional requirement of all data, but because all tasks executed in parallel in thread pool, may back arriving of task finish, thereby make the data of thread pool output and the data of input on time continuity, put upside down prior to task before.

Summary of the invention

Technical matters to be solved by this invention provide a kind of in real-time data base the method for designing of the used thread pool of deal with data.This method can also guarantee the time continuity of data that each thread is handled in the thread pool and the load balancing of each thread when data-handling capacity is provided efficiently.

For solving the problems of the technologies described above, thread pool of the present invention need be safeguarded the Hash mapping table of a task identification to Thread Id, is used for the distribution of task is managed; The thread parameter increases a property parameters: load (nLoad), in order to identify the task total amount that each thread distributes, this parameter quantizes with the frequency (nFreq) that this task in a period of time arrives thread pool, wherein the length of time period (nPeriod) can be provided with (unit for minute) according to application scenarios, and quantitative formula is as follows:

$\mathrm{nLoad}=\frac{\mathrm{nPeriod}*60}{\mathrm{nFreq}}$

If the nLoad result who calculates is 0, then be designated as 1; All threads are no longer shared same task queue, but are the corresponding with it task queue of each thread configuration in the thread pool, and each thread is searched new task from corresponding task queue; One monitor thread is arranged separately, be in dormant state at ordinary times, can regularly start the loading condition of each thread of monitoring, the standard of judgement is the load parameter of thread.

In the such scheme, described step is as follows:

Step 1, initialization, thread pool is created N thread by being provided with, and initialization task is identified to the Hash mapping table of Thread Id, the task queue of each thread of initialization, the load attribute LThread in the thread is initialized as 0;

Step 2, the interpolation of task, the load value LTask that at first calculates new interpolation task according to the load calculation formula; Use the sequential search method to find out the lightest thread of load in the thread pool; Give this thread with this Task Distribution, updating task is identified to the Hash mapping table of Thread Id simultaneously, and the load value LTask of this task is added on the former load value LThread of corresponding thread, promptly

LThread _Newly=LThread _Old+ Ltask;

Step 3, the deletion of task is at first deleted the load value of this task from the load value of corresponding thread, and updating task is identified to the Hash mapping table of Thread Id then, and the mapping relations of this task are deleted from mapping table;

Step 4, the renewal of task when the frequency of task changes, need be upgraded its corresponding threads load, and the mode that adopts this task of deletion earlier to add this task is again finished;

Step 5, the load balancing of thread pool, the interpolation of task is to carry out according to the standard that each threads load of assurance reaches unanimity in second step.But when constantly going out the deletion of current task, the loading condition of each thread will occur inconsistent, and wherein the worst situation is that only remaining a certain thread has task, and the whole dormancy of other threads directly cause concurrent performance to descend.Therefore the loading condition that needs each thread in the monitor thread timing scan thread pool, the form according to the employing variance of judgement:

$\mathrm{vLoad}=\frac{{(\mathrm{Load}1-\mathrm{mLoad})}^2+...+{(\mathrm{Loadn}-\mathrm{mLoad})}^2}{N}$

Wherein vLoad represents the variance yields of load, and N represents the number of thread, Load _i(i=1,2,3 ... N) represent the load value of each thread, and mLoad represents the mean value of all load values, as follows:

$\mathrm{mLoad}=\frac{\mathrm{Load}1+\mathrm{Load}2+...+\mathrm{Loadn}}{N}$

The load variance of calculating gained when monitor thread surpasses LoadMax (be defaulted as 100, can be made as littler), then will carry out the operation of load balancing, and the task that load is departed from the bigger thread of draw value is re-assigned on the less thread of load;

Step 6, task executions when a certain task arrives, is at first obtained the Thread Id of this task of processing from mapping table according to its task identification, then this task is delivered in the task queue of this thread correspondence to get final product.Each thread obtains new task and carries out from task queue separately.

In the such scheme, described method is for the task of arriving thread pool, at first in mapping table, search the Thread Id that these task needs are delivered according to task identification by the task distribution module, then task is sent in the task queue of corresponding thread, the task that thread in the thread pool constantly please be looked for novelty from task queue is separately carried out, and exports the result then.

Preferably, described step 2 may further comprise the steps:

Step 201, working load value computing formula according to the frequency of new interpolation task, calculates the load numerical value of this task;

The mode that step 202, employing travel through in proper order finds out one of thread of load value minimum in all threads;

Step 203 is added this load value on the minimum load thread that finds in 202 to;

Step 204 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool;

Step 205 judges whether the variance yields that calculates in 204 surpasses default LoadMax (default value is 100);

Step 206 if surpassed LoadMax, need be carried out load balancing;

Step 207 is calculated the new hash value that adds task identification, and is added in the mapping table.If add failure, the load value of this task need be deleted from the load value of corresponding thread.

Preferably, described step 3 may further comprise the steps:

Step 301 is calculated the hash value that needs deleted task identification;

Step 302 finds the Thread Id of this task correspondence by hash value in mapping table, and notes the load value of this task;

Step 303 deducts the load value of this task from the total load value of corresponding thread;

Step 304 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool;

Step 305 judges whether the variance yields that calculates in 304 surpasses default LoadMax (default value is 100);

Step 306 if surpassed LoadMax, need be carried out load balancing;

Step 307 is deleted this task from mapping table.If the deletion failure need recover the load value of this task from the load value of corresponding thread.

Preferably, described step 5 may further comprise the steps:

Step 401, monitor thread generally are in dormant state, regularly activate to start (default time is spaced apart 60 minutes);

Step 402 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool;

Step 403 judges that whether the variance yields that calculates in 402 surpasses default LoadMax (default value is 100), if do not exceed, reenters dormant state;

Step 404﹠amp; 405, if exceed LoadMax, travel through the load value of all threads, one of the thread of finding out the load value maximum and one of thread of load value minimum;

Step 406, arbitrary task transfers is to the thread of load minimum in the thread with the load maximum;

Step 407, updating task are identified to the mapping table of Thread Id, the mapping Thread Id of the task of being moved are changed into the ID of less thread.

The present invention is owing to adopted technique scheme to have following advantage:

1, the present invention utilizes newly-increased mapping table function, and the time continuity in the time of can guaranteeing the data input and output of each task is in full accord.

2, can guarantee that by the monitor thread that regularly starts the load on each thread can reach unanimity, to reach best concurrent performance.

Description of drawings

Below in conjunction with the drawings and specific embodiments technical scheme of the present invention is further described in detail.

Fig. 1 is a thread pool structural drawing of the present invention;

Fig. 2 is the schematic flow sheet of interpolation new task of the present invention;

Fig. 3 is the schematic flow sheet of deletion task of the present invention;

Fig. 4 is a load balancing schematic flow sheet of the present invention.

Embodiment

As shown in Figure 1, for the task of arriving thread pool, at first in mapping table, search the Thread Id that these task needs are delivered according to task identification by the task distribution module, then task is sent in the task queue of corresponding thread, the task that thread in the thread pool constantly please be looked for novelty from task queue is separately carried out, and exports the result then.

Fig. 2 is the flow process of adding new task in thread pool, and concrete steps are as follows:

Step 201, working load value computing formula according to the frequency of new interpolation task, calculates the load numerical value of this task.

The mode that step 202, employing travel through in proper order finds out one of thread of load value minimum in all threads.

Step 203 is added this load value on the minimum load thread that finds in 202 to.

Step 204 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool.

Step 205 judges whether the variance yields that calculates in 204 surpasses default LoadMax (default value is 100).

Step 206 if surpassed LoadMax, need be carried out load balancing, and detailed process sees that step 401 is to 407.

Step 207 is calculated the new hash value that adds task identification, and is added in the mapping table.If add failure, the load value of this task need be deleted from the load value of corresponding thread.

Fig. 3 is the flow process of deletion task from thread pool, the following steps that this flow process specifically comprises:

Step 301 is calculated the hash value that needs deleted task identification.

Step 302 finds the Thread Id of this task correspondence by hash value in mapping table, and notes the load value of this task.

Step 303 deducts the load value of this task from the total load value of corresponding thread.

Step 304 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool.

Step 305 judges whether the variance yields that calculates in 304 surpasses default LoadMax (default value is 100).

Step 306 if surpassed LoadMax, need be carried out load balancing, and detailed process sees that step 401 is to 407.

Step 307 is deleted this task from mapping table.If the deletion failure need recover the load value of this task from the load value of corresponding thread.

Fig. 4 is the flow process of thread pool load balancing, the following steps that this flow process specifically comprises:

Step 401, monitor thread generally are in dormant state, regularly activate to start (default time is spaced apart 60 minutes).

Step 402 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool.

Step 403 judges that whether the variance yields that calculates in 402 surpasses default LoadMax (default value is 100), if do not exceed, reenters dormant state.

Step 404﹠amp; 405, if exceed LoadMax, travel through the load value of all threads, one of the thread of finding out the load value maximum and one of thread of load value minimum.

Step 406, arbitrary task transfers is to the thread of load minimum in the thread with the load maximum.

Step 407, updating task are identified to the mapping table of Thread Id, the mapping Thread Id of the task of being moved are changed into the ID of less thread.

It should be noted last that, above embodiment is only unrestricted in order to technical scheme of the present invention to be described, although the present invention is had been described in detail with reference to preferred embodiment, those of ordinary skill in the art is to be understood that, can make amendment or be equal to replacement technical scheme of the present invention, and not breaking away from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of the claim scope of the present invention.

Claims (8)

1. a thread pool method for designing that guarantees time continuity is characterized in that, described thread pool need be safeguarded the Hash mapping table of a task identification to Thread Id, is used for the distribution of task is managed; The thread parameter increases a property parameters: load (nLoad), in order to identify the task total amount that each thread distributes, this parameter quantizes with the frequency (nFreq) that this task in a period of time arrives thread pool, wherein the length of time period (nPeriod) can be provided with (unit for minute) according to application scenarios, and quantitative formula is as follows:

$\mathrm{nLoad}=\frac{\mathrm{nPeriod}*60}{\mathrm{nFreq}}$

If the nLoad result who calculates is 0, then be designated as 1; All threads are no longer shared same task queue, but are the corresponding with it task queue of each thread configuration in the thread pool, and each thread is searched new task from corresponding task queue; One monitor thread is arranged separately, be in dormant state at ordinary times, can regularly start the loading condition of each thread of monitoring, the standard of judgement is the load parameter of thread.

2. a kind of thread pool method for designing that guarantees time continuity according to claim 1 is characterized in that described step is as follows:

Step 1, initialization, thread pool is created N thread by being provided with, and initialization task is identified to the Hash mapping table of Thread Id, the task queue of each thread of initialization, the load attribute LThread in the thread is initialized as 0;

Step 2, the interpolation of task, the load value LTask that at first calculates new interpolation task according to the load calculation formula; Use the sequential search method to find out the lightest thread of load in the thread pool; Give this thread with this Task Distribution, updating task is identified to the Hash mapping table of Thread Id simultaneously, and the load value LTask of this task is added on the former load value LThread of corresponding thread, promptly

LThread _Newly=LThread _Old+ Ltask;

Step 3, the deletion of task is at first deleted the load value of this task from the load value of corresponding thread, and updating task is identified to the Hash mapping table of Thread Id then, and the mapping relations of this task are deleted from mapping table;

Step 4, the renewal of task when the frequency of task changes, need be upgraded its corresponding threads load, and the mode that adopts this task of deletion earlier to add this task is again finished;

Step 5, the load balancing of thread pool, the interpolation of task is to carry out according to the standard that each threads load of assurance reaches unanimity in second step.But when constantly going out the deletion of current task, the loading condition of each thread will occur inconsistent, and wherein the worst situation is that only remaining a certain thread has task, and the whole dormancy of other threads directly cause concurrent performance to descend.Therefore the loading condition that needs each thread in the monitor thread timing scan thread pool, the form according to the employing variance of judgement:

$\mathrm{vLoad}=\frac{{(\mathrm{Load}1-\mathrm{mLoad})}^2+...+{(\mathrm{Loadn}-\mathrm{mLoad})}^2}{N}$

Wherein vLoad represents the variance yields of load, and N represents the number of thread, Load _i(i=1,2,3 ... N) represent the load value of each thread, and mLoad represents the mean value of all load values, as follows:

$\mathrm{mLoad}=\frac{\mathrm{Load}1+\mathrm{Load}2+...+\mathrm{Loadn}}{N}$

The load variance of calculating gained when monitor thread surpasses LoadMax (be defaulted as 100, can be made as littler), then will carry out the operation of load balancing, and the task that load is departed from the bigger thread of draw value is re-assigned on the less thread of load;

Step 6, task executions when a certain task arrives, is at first obtained the Thread Id of this task of processing from mapping table according to its task identification, then this task is delivered in the task queue of this thread correspondence to get final product.Each thread obtains new task and carries out from task queue separately.

3. a kind of thread pool method for designing that guarantees time continuity according to claim 2, it is characterized in that, described method is for the task of arriving thread pool, at first in mapping table, search the Thread Id that these task needs are delivered according to task identification by the task distribution module, then task is sent in the task queue of corresponding thread, the task that thread in the thread pool constantly please be looked for novelty from task queue is separately carried out, and exports the result then.

4. according to claim 2 or 3 described a kind of thread pool methods for designing that guarantee time continuity, it is characterized in that described step 2 may further comprise the steps:

Step 201, working load value computing formula according to the frequency of new interpolation task, calculates the load numerical value of this task;

The mode that step 202, employing travel through in proper order finds out one of thread of load value minimum in all threads;

Step 203 is added this load value on the minimum load thread that finds in 202 to;

Step 204 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool;

Step 205 judges whether the variance yields that calculates in 204 surpasses default LoadMax (default value is 100);

Step 206 if surpassed LoadMax, need be carried out load balancing;

Step 207 is calculated the new hash value that adds task identification, and is added in the mapping table.If add failure, the load value of this task need be deleted from the load value of corresponding thread.

5. according to claim 2 or 3 described a kind of thread pool methods for designing that guarantee time continuity, it is characterized in that described step 3 may further comprise the steps:

Step 301 is calculated the hash value that needs deleted task identification;

Step 302 finds the Thread Id of this task correspondence by hash value in mapping table, and notes the load value of this task;

Step 303 deducts the load value of this task from the total load value of corresponding thread;

Step 304 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool;

Step 305 judges whether the variance yields that calculates in 304 surpasses default LoadMax (default value is 100);

Step 306 if surpassed LoadMax, need be carried out load balancing;

Step 307 is deleted this task from mapping table.If the deletion failure need recover the load value of this task from the load value of corresponding thread.

6. a kind of thread pool method for designing that guarantees time continuity according to claim 4 is characterized in that described step 3 may further comprise the steps:

Step 301 is calculated the hash value that needs deleted task identification;

Step 302 finds the Thread Id of this task correspondence by hash value in mapping table, and notes the load value of this task;

Step 303 deducts the load value of this task from the total load value of corresponding thread;

Step 304 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool;

Step 305 judges whether the variance yields that calculates in 304 surpasses default LoadMax (default value is 100);

Step 306 if surpassed LoadMax, need be carried out load balancing;

Step 307 is deleted this task from mapping table.If the deletion failure need recover the load value of this task from the load value of corresponding thread.

7. a kind of thread pool method for designing that guarantees time continuity according to claim 5 is characterized in that described step 5 may further comprise the steps:

Step 401, monitor thread generally are in dormant state, regularly activate to start (default time is spaced apart 60 minutes);

Step 402 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool;

Step 403 judges that whether the variance yields that calculates in 402 surpasses default LoadMax (default value is 100), if do not exceed, reenters dormant state;

Step 404﹠amp; 405, if exceed LoadMax, travel through the load value of all threads, one of the thread of finding out the load value maximum and one of thread of load value minimum;

Step 406, arbitrary task transfers is to the thread of load minimum in the thread with the load maximum;

Step 407, updating task are identified to the mapping table of Thread Id, the mapping Thread Id of the task of being moved are changed into the ID of less thread.

8. a kind of thread pool method for designing that guarantees time continuity according to claim 6 is characterized in that described step 5 may further comprise the steps:

Step 401, monitor thread generally are in dormant state, regularly activate to start (default time is spaced apart 60 minutes);

Step 402 is utilized the variance computing formula, calculates the load variance of all threads in the thread pool;

Step 403 judges that whether the variance yields that calculates in 402 surpasses default LoadMax (default value is 100), if do not exceed, reenters dormant state;

Step 404﹠amp; 405, if exceed LoadMax, travel through the load value of all threads, one of the thread of finding out the load value maximum and one of thread of load value minimum;

Step 406, arbitrary task transfers is to the thread of load minimum in the thread with the load maximum;

Step 407, updating task are identified to the mapping table of Thread Id, the mapping Thread Id of the task of being moved are changed into the ID of less thread.

Images