CN108200118A - A kind of solution based on the request of movement simulation platform high concurrent - Google Patents

Abstract

The invention discloses a method for solving high concurrent requests based on a motion simulation platform. The method includes: configuring a server group to receive user requests; performing pre-simulation processing and enabling multi-thread processing of user requests; performing clothing heat and humidity transmission calculation and human body heat transfer calculation. Adjust calculations; feedback simulation results, and back up simulation records to the database. In the embodiment of the present invention, by configuring multiple servers to process requests, tasks are allocated reasonably, loads among servers are balanced, and a large number of request accumulations are avoided. Combined with the scheduling algorithm to balance the computing pressure of each server, thereby reducing the average response time of the server, so that each server can maintain a high computing efficiency. At the same time, the multi-threaded model can execute multiple simulation computing tasks in parallel, reducing the user's blocking waiting time, and effectively improving the processing capacity and efficiency of the motion simulation platform under high concurrent requests.

Description

A solution to high concurrent requests based on motion simulation platform

technical field

The invention relates to the technical field of computer servers, in particular to a solution to high concurrent requests based on a motion simulation platform.

Background technique

In modern society, people pay more and more attention to their physical health, and exercise is becoming an indispensable part of life. Reasonable exercise plays an important role in promoting physical health, but many people suffer from physical discomfort during exercise due to lack of understanding of their own conditions and the intensity of exercise, such as the long-distance running of students that has been frequently reported in newspapers Syncope event. In recent years, some teams have developed a human heat and humidity motion simulation algorithm, which can use the clothing heat and moisture transfer model and the human body heat regulation model to perform simulation calculations based on user-defined parameters such as body characteristics, exercise plans, and external environment, and predict The temperature, sweat rate, and comfort of the human body during exercise are recorded. This kind of motion simulation algorithm has broad application prospects. Some applications based on this algorithm have been developed to inform users of possible physical discomfort during exercise in advance and help users make better exercise plans.

At present, most of the applications based on human heat and humidity motion simulation algorithm are based on C-S (Client-Server, client-server) architecture, and the entire simulation platform consists of two parts: client and server. The user inputs parameters on the client side, packs them into user requests and sends them to the server side, and the server side uses the human body heat and humidity motion simulation algorithm to process the parameters, calculates the corresponding results and returns them to the client side for display. The server handles user requests in a linear mode, that is, it receives a user request, processes a user request, returns the result, and then receives the next request, which means that it is impossible to process multiple user requests at the same time.

Most of the existing research on sports and health simulation platforms focuses on the improvement of algorithms, and lacks consideration of the surge in the number of users after large-scale use. In the case of a small number of user requests, there is no big problem with the server processing requests in a linear mode. However, once the number of requests increases, due to the time required for the execution of the simulation algorithm and the factors of network stability, a large number of requests will accumulate on the server. At this time, the queuing time for the late arrival user request to be processed will be longer, and the average time for the user to wait for the result to be returned will be longer, resulting in a bad experience.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art. The main disadvantage of the existing motion simulation platform implementation scheme is the lack of ability to deal with high concurrent requests. When facing large-scale user requests, the work efficiency is low and the response time is long. To solve this problem, we can start from two aspects of algorithm and server performance. Improving the operating efficiency of the algorithm is certainly an effective method, but the research period is long, and the algorithm has gradually matured, so there is limited room for improvement. Therefore, starting from the platform server, improving the processing capacity of the server to deal with high concurrent user requests is of great significance for improving the performance of the sports health simulation platform in practical applications. The invention provides a solution based on the high concurrent request of the motion simulation platform, improves the working efficiency of the simulation platform server, reduces the service response delay under large-scale user requests, and enables the simulation platform users to obtain better experience.

In order to solve the above problems, the present invention proposes a solution based on high concurrent request of motion simulation platform, said method comprising:

Configure the server group and receive user requests;

Perform pre-simulation processing and enable multithreading to process user requests;

Carry out clothing heat and moisture transfer calculations and human body heat regulation calculations;

Feedback the simulation results and back up the simulation records to the database.

Preferably, the configuration server group, the specific steps of receiving user requests include:

Configure server clusters, including a scheduling server, a database server, and multiple internal servers. The IP address of the scheduling server is public, and the IP addresses of the database server and all internal servers are not public;

A TCP connection pool is used to maintain communication between the scheduling server and each internal server;

The dispatch server receives all user requests, but does not actually process them. A long connection is used between the scheduling server and the client. The scheduling server forwards user requests to the internal server for processing using the least connection strategy.

Preferably, said performing pre-simulation processing and enabling multi-threaded processing of user requests specifically includes:

The internal server listens to a specific port, and the multi-channel separation thread monitors the user request forwarded from the scheduling server, and hands it to the network IO thread to read and analyze user data. The process of parsing user data does not block the listening port;

The network IO thread obtains the parameters required for simulation calculation from the user request packet, including user ID, body characteristic parameters, clothing setting parameters, motion plan parameters and external environment parameters

After the network IO thread parses the user input parameters, it detects whether there is an idle thread in the simulation calculation thread pool, and if so, selects an idle thread and passes simulation calculation parameters to it to perform clothing heat and humidity transmission calculation and human body heat regulation calculation; otherwise Add user requests to the simulation task queue.

Preferably, the feedback simulation results specifically include:

After the simulation calculation thread is executed, the demultiplexing thread is notified. The demultiplexing thread listens to the occurrence of the event, obtains the result data and calls the network IO thread for processing. The result data includes the temperature value and sweat rate of each body part at each time node.

The network IO thread uses the TCP connection pool to send the result data back to the scheduling server.

After the scheduling server receives the result data, in order not to expose the real ip address of the internal server, the source address of the data packet is changed to the ip address of the scheduling server, and then sent to the user.

Preferably, said backing up the simulation record to the database specifically includes:

The network IO thread sends the simulation records to the database server. A simulation record includes a user ID, user input parameters and said result data.

Save data in the form of key-value pairs, with the user ID as the key. The database server judges whether an entry with the same key as the current user ID already exists in the database. If it exists, it will directly add a new simulation record to the list of the entry; if it does not exist, first create a new entry with the user ID as the key, and then Add a simulation record to the list field.

Preferably, the specific implementation method of the TCP connection pool includes:

Several TCP connections are created in advance between the scheduling server and each internal server. The source port of each TCP connection is the local port of the dispatching server and is not repeated; the destination port is the designated port of the internal server, that is, the port monitored by the simulation main program. Set the minimum connection number min of the connection pool to ensure that there are at least min TCP connections with each internal server at each moment. Set the maximum connection number max of the connection pool to ensure that the number of TCP connections with each internal server does not exceed max. Set the idle waiting time of the connection to idle. When a TCP connection is idle for more than idle and the number of current TCP connections is greater than min, the connection will be closed. When the scheduling server needs to send data to the internal server, it first checks whether there is an idle TCP connection, and if so, selects an idle connection to send data. If there is no idle connection, it is necessary to judge whether the number of existing connections between the scheduling server and the internal server has reached max, if not, create a new TCP connection to send data; if it has reached max, add the data to the waiting queue , wait until there is an idle connection before processing. Each TCP connection is not closed immediately after sending data.

Preferably, the specific implementation method of the long connection includes:

After the server receives user data, it will not disconnect from the client immediately, but will start a timer, and the connection will be kept until the timer expires. If the server receives a user request again during the timing, it restarts timing. If the client does not send a request to the server until the timer expires, the connection is closed after the timer expires.

Preferably, the specific implementation method of the least connection strategy is as follows:

The dispatch server keeps track of the number of active requests it forwards to each internal server. Initially, the number of active requests between the dispatch server and all internal servers is 0. When a request is forwarded to an internal server for processing, or the request is added to the waiting queue of the internal server's TCP connection pool, the number of active requests between the scheduling server and the internal server is increased by one; when an internal server's request When processing ends or the connection is aborted due to a timeout, the number of active requests between the dispatch server and this internal server is decremented by one. Each time a request is received, the scheduling server will give priority to the internal server with the least number of active requests to forward. When several internal servers have the same number of active requests, it is necessary to check the number of idle connections in the TCP connection pool between the scheduling server and these internal servers, and preferentially select the internal server with the largest number of idle connections for forwarding. If there are still several internal servers with the highest number of idle connections, one of these servers is randomly selected for forwarding.

Preferably, the specific implementation functions of the demultiplexed threads include:

The main responsibility of the demultiplexed thread is to monitor the events that occur in the system, and forward the events to the corresponding functions for processing. It maintains a listener list internally. Whenever a new connection is established in the TCP connection pool between the internal server and the scheduling server, the socket of the connection is added to the listening list; whenever a connection in the TCP connection pool is closed, the socket of the connection is deleted from the listening list . The objects in the listener list also include all worker threads for simulation calculations. The multi-channel separation thread calls the select, epoll and other methods of the operating system to wait for events to occur in a loop. The events of concern include two types: the arrival of user data and the completion of simulation calculations. Whenever an event occurs, the demultiplexed thread can be notified from the operating system, and obtain the socket or thread list corresponding to the event through the listening list, dispatch the event to the corresponding function for processing, and then continue to wait for the next event.

Preferably, the specific implementation functions of the network IO thread include:

The network IO thread is responsible for extracting the parameters required for the simulation calculation from the user request data packet, and sending the result data to the scheduling server after the simulation calculation is completed. The network IO thread and the demultiplexing thread can be executed concurrently, so the process of parsing user data will not block the listening port.

Preferably, the specific implementation method of the simulation computing thread pool includes:

A certain number of threads are created in advance in the internal server, dedicated to processing simulation calculations. The empirical formula for the initial number of threads is T=C/P, where T is the size of the thread pool, C is the number of server CPUs, and P is the proportion of time occupied by intensive computing tasks. Initially all threads are idle and no tasks are executed. Set the minimum number of threads min and the maximum number of threads max in the thread pool to ensure that the number of simulation computing threads (including idle and executing tasks) at any time is not less than min and not greater than max. Set the maximum idle waiting time idle. When a simulation thread is idle for more than idle and the number of threads in the current thread pool is greater than min, the thread will end. When a new simulation calculation task needs to be executed, it can be divided into the following three situations:

If there are idle threads in the thread pool, randomly select an idle thread to execute the task.

If there is no idle thread in the thread pool, but the current number of threads is less than max, create a new thread to execute the task, and add the thread to the thread pool.

If there is no idle thread in the thread pool, and the current number of threads is greater than or equal to max, add the task to the end of the task queue. The task queue adopts the first-in-first-out service mode. When a thread in the thread pool finishes executing the simulation calculation task and enters the idle state, the first task is taken out from the head of the task queue, and an idle thread is selected in the thread pool to execute it.

Preferably, after the internal server finishes executing the simulated calculation task, the thread will not be destroyed immediately, but re-marked as an idle state. Therefore, the advantage of using this technology is that threads can be reused, reducing the overhead caused by thread creation and closing, thereby reducing the time required to process user requests.

Preferably, the specific implementation method for storing data in the form of key-value pairs includes:

Each data entry contains two fields, the first field is called "key", which is used to record the user's ID, and the second field is a list, which is initially empty. When the database server receives the simulation record sent by the internal server, it judges whether there is an entry with the same key as the current user ID in the database. If it exists, insert the simulation record directly into the list of the entry; if it does not exist, first use Create a new entry with the user ID as the key, and insert a dummy record into the list field. In order to avoid the large amount of stored data, resulting in insufficient space on the database server, it is stipulated that each user ID can only save up to 5 simulation records. Every time a new simulation record is inserted into the list field, if the number of elements in the list has reached 5, delete the earliest inserted simulation record, and then insert a new simulation record.

In the embodiment of the present invention, multiple servers are configured to process user requests, user requests are allocated reasonably, loads among servers are balanced, and a large number of requests are not piled up on one server. On this basis, the scheduling algorithm is combined to balance the computing pressure of each server, thereby reducing the average response time of the server, so that each server can maintain a high computing efficiency. At the same time, the server adopts a multi-threading model, which can execute multiple simulation calculation tasks in parallel, reducing the user's blocking waiting time. In addition, in order to cope with unstable network conditions, considering the possibility of simulation results being lost during transmission, the simulation results are also backed up to the server in case of emergency. In this way, the processing capability and efficiency of the motion simulation platform under high concurrent requests can be effectively improved.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is a schematic flow diagram of a solution to high concurrent requests based on a motion simulation platform according to an embodiment of the present invention;

Fig. 2 is a schematic structural diagram of configuring a server group in an embodiment of the present invention;

Fig. 3 is a schematic diagram of a model used by an internal server to process a user request in an embodiment of the present invention.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Fig. 1 is a schematic flow diagram of a solution to high concurrent requests based on a motion simulation platform according to an embodiment of the present invention. As shown in Fig. 1, the method includes:

A1: Configure the server group and receive user requests;

A2: Perform pre-simulation processing and enable multi-threading to process user requests;

A3: Carry out clothing heat and moisture transfer calculations and human body heat regulation calculations;

A4: Feedback the simulation results and back up the simulation records to the database.

Among them, A1 further includes:

A11: Configure server clusters, including a scheduling server, a database server and multiple internal servers. The IP address of the scheduling server is public, and the IP addresses of the database server and all internal servers are not public;

A12: The TCP connection pool is used to maintain communication between the scheduling server and each internal server;

A13: The scheduling server receives all user requests, but does not actually process them. A long connection is used between the scheduling server and the client. The scheduling server forwards user requests to the internal server for processing using the least connection strategy.

A2 further includes:

A21: The internal server monitors a specific port, and the multi-channel separation thread monitors the user request forwarded from the scheduling server, and hands it to the network IO thread to read and parse user data. The process of parsing user data does not block the listening port.

A22: The network IO thread obtains the parameters required for simulation calculation from the user request packet, including user ID, body feature parameters, clothing setting parameters, motion plan parameters and external environment parameters.

A23: After the network IO thread parses the user input parameters, it detects whether there is an idle thread in the simulation calculation thread pool, and if so, selects an idle thread, passes the simulation calculation parameters obtained in A22 to it, and transfers to A3 for clothing heat and humidity transmission Calculation and human thermal regulation calculation; otherwise, add user request to simulation task queue.

Among them, A4 further includes:

A41: After the simulation computing thread finishes executing, notify the demultiplexing thread. The demultiplexing thread listens to the occurrence of the event, obtains the result data and calls the network IO thread for processing. The result data includes the temperature value and sweat rate of each body part at each time node.

A42: The network IO thread uses the TCP connection pool to send the result data back to the scheduling server.

A43: After the scheduling server receives the result data, in order not to expose the real ip address of the internal server, change the source address of the data packet to the ip address of the scheduling server, and then send it to the user.

A44: The network IO thread sends the simulation record to the database server. The simulation record includes the user ID, user input parameters, and result data obtained in A41.

A45: Save data in the form of key-value pairs, with the user ID as the key. The database server judges whether an entry with the same key as the current user ID already exists in the database. If it exists, it will directly add a new simulation record to the list of the entry; if it does not exist, first create a new entry with the user ID as the key, and then Add a simulation record to the list field.

In a specific embodiment, the specific implementation method of the TCP connection pool described in A12 is:

Several TCP connections are created in advance between the scheduling server and each internal server. The source port of each TCP connection is the local port of the dispatching server and is not repeated; the destination port is the designated port of the internal server, that is, the port monitored by the simulation main program. Set the minimum connection number min of the connection pool to ensure that there are at least min TCP connections with each internal server at each moment. Set the maximum connection number max of the connection pool to ensure that the number of TCP connections with each internal server does not exceed max. Set the idle waiting time of the connection to idle. When a TCP connection is idle for more than idle and the number of current TCP connections is greater than min, the connection will be closed. When the scheduling server needs to send data to the internal server, it first checks whether there is an idle TCP connection, and if so, selects an idle connection to send data. If there is no idle connection, it is necessary to judge whether the number of existing connections between the scheduling server and the internal server has reached max, if not, create a new TCP connection to send data; if it has reached max, add the data to the waiting queue , wait until there is an idle connection to process. Each TCP connection is not closed immediately after sending data.

Specifically, the specific implementation method of the long connection described in A13 is as follows:

After the server receives user data, it will not disconnect from the client immediately, but will start a timer, and the connection will be kept until the timer expires. If the server receives a user request again during the timing, it restarts timing. If the client does not send a request to the server until the timer expires, the connection is closed after the timer expires.

The specific implementation method of the least connection strategy described in A13 is as follows:

The dispatch server keeps track of the number of active requests it forwards to each internal server. Initially, the number of active requests between the dispatch server and all internal servers is 0. When a request is forwarded to an internal server for processing, or the request is added to the waiting queue of the internal server's TCP connection pool, the number of active requests between the scheduling server and the internal server is increased by one; when an internal server's request When processing ends or the connection is aborted due to a timeout, the number of active requests between the dispatch server and this internal server is decremented by one. Each time a request is received, the scheduling server will give priority to the internal server with the least number of active requests to forward. When several internal servers have the same number of active requests, it is necessary to check the number of idle connections in the TCP connection pool between the scheduling server and these internal servers, and preferentially select the internal server with the largest number of idle connections for forwarding. If there are still several internal servers with the highest number of idle connections, one of these servers is randomly selected for forwarding.

Specifically, the specific implementation functions of the multi-channel separation thread described in A21 are as follows:

The main responsibility of the demultiplexed thread is to monitor the events that occur in the system, and forward the events to the corresponding functions for processing. It maintains a listener list internally. Whenever a new connection is established in the TCP connection pool between the internal server and the scheduling server, the socket of the connection is added to the listening list; whenever a connection in the TCP connection pool is closed, the socket of the connection is deleted from the listening list . The objects in the listener list also include all worker threads for simulation calculations. The multi-channel separation thread calls the select, epoll and other methods of the operating system to wait for events to occur in a loop. The events of concern include two types: the arrival of user data and the completion of simulation calculations. Whenever an event occurs, the demultiplexed thread can be notified from the operating system, and obtain the socket or thread list corresponding to the event through the listening list, dispatch the event to the corresponding function for processing, and then continue to wait for the next event.

Specifically, the specific implementation functions of the network IO thread described in A21 are as follows:

The network IO thread is responsible for extracting the parameters required for the simulation calculation from the user request data packet, and sending the result data to the scheduling server after the simulation calculation is completed. The network IO thread and the demultiplexing thread can be executed concurrently, so the process of parsing user data will not block the listening port.

Further, the specific implementation method of the simulation computing thread pool described in A23 is as follows:

A certain number of threads are created in advance in the internal server, dedicated to processing simulation calculations. The empirical formula for the initial number of threads is T=C/P, where T is the size of the thread pool, C is the number of server CPUs, and P is the proportion of time occupied by intensive computing tasks. Initially all threads are idle and no tasks are executed. Set the minimum number of threads min and the maximum number of threads max in the thread pool to ensure that the number of simulation computing threads (including idle and executing tasks) at any time is not less than min and not greater than max. Set the maximum idle waiting time idle. When a simulation thread is idle for more than idle and the number of threads in the current thread pool is greater than min, the thread will end. When a new simulation calculation task needs to be executed, it can be divided into the following three situations:

1. If there are idle threads in the thread pool, randomly select an idle thread to execute the task.

2. If there is no idle thread in the thread pool, but the current number of threads is less than max, create a new thread to execute the task and add the thread to the thread pool.

3. If there is no idle thread in the thread pool, and the current number of threads is greater than or equal to max, add the task to the end of the task queue. The task queue adopts the first-in-first-out service mode. When a thread in the thread pool finishes executing the simulation calculation task and enters the idle state, the first task is taken out from the head of the task queue, and an idle thread is selected in the thread pool to execute it.

In a specific embodiment, after the internal server completes the simulation calculation task, the thread will not be destroyed immediately, but will be re-marked as an idle state. Therefore, the advantage of using this technology is that threads can be reused, reducing the overhead caused by thread creation and closing, thereby reducing the time required to process user requests.

Specifically, the specific implementation method for storing data in the form of key-value pairs described in A45 is as follows:

Each data entry contains two fields, the first field is called "key", which is used to record the user's ID, and the second field is a list, which is initially empty. When the database server receives the simulation record sent by the internal server, it judges whether there is an entry with the same key as the current user ID in the database. If it exists, insert the simulation record directly into the list of the entry; if it does not exist, first use Create a new entry with the user ID as the key, and insert a dummy record into the list field. In order to avoid the large amount of stored data, resulting in insufficient space on the database server, it is stipulated that each user ID can only save up to 5 simulation records. Every time a new simulation record is inserted into the list field, if the number of elements in the list has reached 5, delete the earliest inserted simulation record, and then insert a new simulation record.

In the embodiment of the present invention, multiple servers are configured to process user requests, user requests are allocated reasonably, loads among servers are balanced, and a large number of requests are not piled up on one server. On this basis, the scheduling algorithm is combined to balance the computing pressure of each server, thereby reducing the average response time of the server, so that each server can maintain a high computing efficiency. At the same time, the server adopts a multi-threading model, which can execute multiple simulation calculation tasks in parallel, reducing the user's blocking waiting time. In addition, in order to cope with unstable network conditions, considering the possibility of simulation results being lost during transmission, the simulation results are also backed up to the server in case of emergency. In this way, the processing capability and efficiency of the motion simulation platform under high concurrent requests can be effectively improved.

Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above-mentioned embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium can include: Read Only Memory (ROM, Read Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk, etc.

In addition, a solution to high concurrent requests based on the motion simulation platform provided by the embodiment of the present invention has been introduced in detail above. In this paper, specific examples are used to illustrate the principle and implementation of the present invention. The description of the above embodiment It is only used to help understand the method of the present invention and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation and scope of application. In summary, The contents of this description should not be construed as limiting the present invention.

Claims (9)

1. a kind of solution based on the request of movement simulation platform high concurrent, which is characterized in that the method includes：

Configuration server group receives user's request；

Emulation pre-treatment is carried out, enables multiple threads user request；

Clothing Wetness And Heat transmission is carried out to calculate and mannequin surface modeling calculating；

Simulation result is fed back, emulation record is backuped in database.

A kind of 2. solution based on the request of movement simulation platform high concurrent as described in claim 1, which is characterized in that institute Configuration server group is stated, user's request is received and includes：

Configuration server cluster, including a dispatch server, a database server and Duo Tai internal servers.Scheduling clothes The ip addresses of the ip addresses external disclosure of business device, database server and all internal servers are all underground.

It maintains to communicate using TCP connection pond between dispatch server and every internal server.

Dispatch server receives the request of all users, but without actual treatment.It is used between dispatch server and user terminal Long connection.User's request is transmitted to internal server using minimum connection strategy and handled by dispatch server.

A kind of 3. solution based on the request of movement simulation platform high concurrent as described in claim 1, which is characterized in that institute It states and carries out emulation pre-treatment, enable multiple threads user request and include：

Internal server monitors particular port, and user's request that the monitoring of multichannel separation thread is forwarded from dispatch server is given Network I/O thread reads and parses user data.The process of parsing user data does not block listening port.

Network I/O thread obtains the parameter needed for simulation calculation from user's request data package, joins including User ID, physical trait Number, clothes setup parameter, exercise program parameter and external environment parameters.

After network I/O thread has parsed user's input parameter, whether detection simulation computational threads are free idle thread in pond, if having An idle thread is selected, is passed to the simulation calculation parameter obtained, progress Clothing Wetness And Heat transmission calculates and mannequin surface modeling calculates； Otherwise user is asked to add in artificial tasks queue.

A kind of 4. solution based on the request of movement simulation platform high concurrent as claimed in claim 2, which is characterized in that institute Minimum connection strategy concrete methods of realizing is stated to include：

Dispatch server need to record it and enliven number of requests to what every internal server forwarded.When initial, dispatch server with Number of request of enlivening between all internal servers is 0.When a request is forwarded to certain internal server processing or please When seeking the waiting list in the TCP connection pond for being added into the internal server, between dispatch server and the internal server It enlivens number of request and adds one；When the request processing of an internal server terminate or connect because time-out stop when, dispatch server with Number of request of enlivening between the internal server subtracts one.After receiving request every time, dispatch server can preferentially select actively please Several minimum internal servers is asked to forward.When have several internal servers enliven number of request it is identical when, then need to check tune Degree server connect number with the free time in the TCP connection pond of these internal servers, and the preferential idle connection number of selection is most Internal server forwarding.If the idle connection number for still having several internal servers is most side by side, in these services A forwarding is randomly choosed in device.

A kind of 5. solution based on the request of movement simulation platform high concurrent as claimed in claim 2, which is characterized in that institute TCP connection pond concrete methods of realizing is stated to include：

Several TCP connections are first created between dispatch server and every internal server in advance.Wherein each TCP connection Source port is the local port of dispatch server, and does not repeat mutually；Target port is the designated port of internal server, that is, is emulated The port that main program is monitored.Set the minimum connection number min of connection pool, it is ensured that with the TCP connection between every internal server At all at least min of each moment.Set the maximum number of connections max of connection pool, it is ensured that between every internal server TCP connection is no more than max.The idle waiting time idle of connection is set, is continuously in idle condition when a TCP connection Time is more than idle, and when current TCP connection number is more than min, closes the connection.When dispatch server needs internally server During transmission data, available free TCP connection is first checked whether, if so, one idle connection transmission data of selection.If not yet Available free connection, then need to judge whether the existing connection number between dispatch server and the internal server has reached max, such as Fruit does not reach, then creates a TCP connection transmission data；If having reached max, data are added in into waiting list, Deng Daoyou It is reprocessed during free time connection.Each TCP connection will not close immediately after data have been sent.

A kind of 6. solution based on the request of movement simulation platform high concurrent as claimed in claim 3, which is characterized in that institute Multichannel separation thread specific implementation function is stated to include：

The event occurred in the main monitoring system of multichannel separation thread, and give event forwarding to the processing of corresponding function.It One monitoring list of internal maintenance.When the TCP connection pond between internal server and dispatch server has new connection to establish, The socket of the connection is added to and is monitored in list；Whenever having connection closed in TCP connection pond, deleted from monitoring in list The socket of the connection.Monitor the worker thread that the object in list further includes all simulation calculations.Multichannel separation thread dispatching The methods of select, epoll of operating system cycle waiting event occurs, and the event of concern includes two kinds：User data arrival, Simulation calculation finishes.Whenever event occurs, multichannel separation thread can be just notified from operating system, and pass through monitoring list The corresponding socket of the event or thread list are got, events dispatcher to corresponding function is handled, is then proceeded to Wait for next event.

A kind of 7. solution based on the request of movement simulation platform high concurrent as claimed in claim 3, which is characterized in that institute The network I/O thread specific implementation function of stating includes：

Network I/O thread is responsible for extracting parameter needed for simulation calculation and complete in simulation calculation from user's request data package Bi Hou, by result data degree of sending back server.Network I/O thread and multichannel separation thread can be performed concurrently, so parsing is used The process of user data will not block listening port.

8. a kind of solution based on the request of movement simulation platform high concurrent as claimed in claim 3, which is characterized in that imitative True computational threads pond concrete methods of realizing includes：

A certain number of threads are created in advance in internal server, dedicated for handling simulation calculation.Initial thread quantity is big Small empirical equation is T=C/P, and wherein T is thread pool size, and C is the quantity of server CPU, and P is shared by intensive calculations task The proportion of time.All threads are all idle when initial, without tasks carrying.Set thread pool minimum Thread Count min and Maximum thread max, it is ensured that simulation calculation Thread Count is not small (including idle and be carrying out task) at any one time In min and no more than max.Setting maximum idle waiting time idle, when an emulation thread is continuously in idle condition Time be more than idle, and current thread pond center line number of passes be more than min when, terminate the thread.When there is new simulation calculation task When needing to perform, it is divided into following three kinds of situations:

If there is free idle thread in thread pool, therefrom randomly choose an idle thread and perform the task；

If there is no idle thread in thread pool, but current thread number is less than max, then create a new thread and perform task, and The thread is added in thread pool；

If there is no idle thread in thread pool, and current thread number is more than or equal to max, then task is added in the end of task queue Tail.Task queue uses the method for service of first in first out, goes forward side by side when having performed simulation calculation task there are one thread in thread pool After entering idle state, first task is taken out, and an idle thread is selected to perform in thread pool from the head of task queue It.

9. a kind of solution based on the request of movement simulation platform high concurrent as claimed in claim 8, which is characterized in that when After internal server has performed simulation calculation task, thread will not be destroyed immediately, but re-flag as idle state.