CN106569896B - A data distribution and parallel processing method and system - Google Patents

Abstract

The invention discloses a kind of data distribution and method for parallel processing and systems, which comprises parses to initial query instruction, obtains true inquiry instruction；True inquiry instruction is distributed in different back end by work distribution node；It executes the true inquiry instruction received respectively on each back end, and query result is returned into work distribution node, while keeping the update of data information table on each server node；On each server node, the corresponding inquiry table of the true inquiry instruction is backed up, so that preserving inquiry table copy on each server node；On each server node, when detecting the presence of concurrency conflict, by calling the inquiry table copy on current server node concurrently to execute true inquiry instruction.The data distribution and method for parallel processing and system can not only guarantee the consistency of data and can be improved the speed and efficiency of data query.

Description

A data distribution and parallel processing method and system

technical field

The present invention relates to the technical field of data processing, in particular to a data distribution and parallel processing method and system.

Background technique

The parallel processing method is a processing method that adopts parallel means in information processing to achieve efficient development and calculation. In other words, it refers to the completion of multiple tasks within the same time frame, which may or may not be of the same nature. As long as there is overlap in time, it can be called parallel. It is mainly used in high-performance processors, large-scale database management, complex mathematical modeling and other fields, and its application scope is still expanding. It mainly contains three elements: simultaneity, concurrency, and flow. In order to achieve high performance and meet the needs of a large amount of calculation, a parallel computing model can be used for data processing.

Big data analytics are increasingly common in many business areas, including financial companies, government agencies, and insurance agencies. Its applications range from generating simple reports to executing complex analytical workloads. The difficulty is: as the amount of data increases, how to perform analytical queries and storage in these areas. Generally speaking, distributed system design needs to consider data consistency (Consistency), availability (Availability) and partition tolerance (Partition tolerance). Data consistency means that if data is replicated to multiple servers, all servers also need to be updated when the data is updated; availability means timely response to user needs; partition fault tolerance is the scalability of nodes. Massively Parallel Processing (MPP) architecture can address these challenges by distributing storage and queries across multiple nodes and processes. MPP can be connected through Internet communication, and its terminals can be jointly processed by multiple low-cost servers. In the MPP architecture, each module has an independent CPU, storage, etc. to manage the data of each module, and allows to dynamically add or delete server nodes. Therefore, the data and computing power between these modules are not shared, and the data is distributed in local storage.

Specifically, MPP is connected by a plurality of SMP (Symmetric Multi-Processor) servers through a certain node interconnection network, and works cooperatively to complete the same task. From the user's point of view, it can be regarded as a server system. Its basic feature is that it consists of multiple SMP servers. Each SMP server is called a node or a server node. It is connected through the node interconnection network. Each node only accesses its own local resources such as memory and storage. Shared (share nothing) structure; MPP has better expansion capability, and theoretically its expansion is unlimited. Referring to FIG. 4 , it is a schematic structural diagram of an MPP architecture.

In an MPP system, the CPU within each node cannot access the memory of another node. The information exchange between nodes is realized through the node interconnection network, and this process is generally called data redistribution. The MPP server requires a complex mechanism to schedule and balance the load and parallel processing of various nodes. At present, some servers based on MPP technology often shield this complexity through system-level software (such as database). The MPP database distributes tasks to multiple servers and nodes in parallel, and after each node completes the calculation, the respective results are aggregated together to obtain the final result. Although MPP can realize the connection of large-scale CPUs, it is precisely because of its non-shared characteristics that it is difficult to maintain data consistency. Since the MPP system does not share resources, it needs to occupy more resources than the SMP. When the communication time is long, the MPP system cannot give full play to the resource advantage to achieve high efficiency.

Therefore, the inventor found in the process of implementing the present invention that the prior art has at least the following problems: the consistency of the data in the database is not good, and the query speed and efficiency of the data are low.

SUMMARY OF THE INVENTION

In view of this, the purpose of the present invention is to provide a data distribution and parallel processing method and system, which can not only ensure the consistency of data but also improve the speed and efficiency of data query.

Based on the above-mentioned purpose, the data distribution and parallel processing method provided by the present invention include:

Receive the initial query instruction input by the client;

Analyzing the initial query instruction to obtain a real query statement that meets the standard, and use it as a real query instruction;

Send the real query instruction to the work distribution node, and distribute the real query instruction to different data nodes through the work distribution node; wherein, the work distribution node and the data node refer to all server nodes in the database Nodes divided according to preset functions; and the work assignment nodes are used to manage task assignments and query results of data nodes;

Execute the received real query instruction on each data node, and return the query result to the work distribution node, while keeping the data information table updated on each server node;

On each server node, back up the query table corresponding to the real query instruction, so that each server node saves a copy of the query table;

On each server node, when a concurrency conflict is detected, the real query instruction is executed concurrently by calling the copy of the query table on the current server node.

Optionally, when it is detected that there is a concurrency conflict, the step of concurrently executing the real query instruction by calling the copy of the query table on the current server node includes:

Detecting whether there is a concurrency conflict in the real query instruction in each server node;

If there is no concurrency conflict, submit a query task; wherein, the query task is a query task corresponding to a real query instruction;

If there is a concurrency conflict, call the adaptive backup on the current server node to compare with the current query task, and execute the query task on the adaptive backup;

If the query tasks in all server nodes are successfully executed, the decision to submit the query task is returned, otherwise, the decision to withdraw the query task is returned.

Optionally, the method for dividing all server nodes in the database into work assignment nodes and data nodes according to preset functions includes:

Detect the local storage capacity and computing processing capacity of each server node;

Determine whether the computing processing capability of the current server node is greater than the preset performance threshold, if so, the current server node is set as a work distribution node, otherwise the current server node is set as a data node;

or,

It is judged whether the local storage amount of the current server node is greater than the preset storage threshold, and if so, the current server node is set as a data node; otherwise, the current server node is set as a work distribution node.

Optionally, the step of sending the real query instruction to a work distribution node and distributing the real query instruction to different data nodes through the work distribution node further includes:

Detect the specific usage of each server node;

According to the detection results, a preset balancing strategy is adopted, and data nodes and corresponding spaces are allocated on each server node.

Optionally, the balancing strategy is a strategy with the lowest space usage rate.

Optionally, the step of maintaining the update of the data information table on each server node further includes:

The location information and related operation information of the data information table are copied and stored.

Optionally, the step of maintaining the update of the data information table on each server node further includes:

Back up all data in each data node to the specified data node;

Keep all data in each data node and the synchronous update of the corresponding data backup;

The location information of each data node is saved on each data node.

Optionally, the step of parsing the initial query instruction further includes:

Perform lexical analysis on the characters input by the user to obtain words that meet the standard;

Perform grammatical analysis on multiple consecutive words to obtain sentences that conform to grammatical logic, and construct an abstract syntax tree at the same time;

Convert the logical SQL statement to the real SQL statement according to the abstract syntax tree, and obtain the real query statement that meets the standard, which is used as the real query instruction.

The present invention also provides a data distribution and parallel processing system, including:

an input unit for receiving an initial query instruction input by the user terminal;

a parsing unit, configured to parse the initial query instruction to obtain a real query statement that meets the standard, and use it as a real query instruction;

A distribution unit, configured to send the real query instruction to a work distribution node, and distribute the real query instruction to different data nodes through the work distribution node; wherein, the work distribution node and the data node refer to All server nodes in the database are divided into nodes according to preset functions; and the work assignment nodes are used to manage task assignment and query results of data nodes;

The processing unit is used to execute the received real query instruction on each data node respectively, and return the query result to the work distribution node, and at the same time keep the update of the data information table on each server node;

a backup unit, configured to back up the lookup table corresponding to the real inquiry instruction on each server node, so that each server node saves a copy of the lookup table;

The concurrency unit is used on each server node to concurrently execute the real query instruction by calling the copy of the query table on the current server node when a concurrency conflict is detected.

It can be seen from the above that the method and system for data distribution and parallel processing provided by the present invention divides all server nodes in the database into work distribution nodes and data nodes, so that the work distribution nodes can manage the task distribution of the data nodes. and the return of query results; in this way, data nodes can achieve load balancing according to different usage states. By keeping the data information table updated, the consistent synchronization of the data in the server node can be ensured. By backing up a copy of the query table on each server node, each server node can execute query instructions concurrently, thereby shortening corresponding time and improving query efficiency. The consistency of distributed data can be guaranteed by invoking an adaptive backup to compare with the current query task when there is a concurrency conflict. Therefore, the data distribution and parallel processing method and system of the present invention can not only ensure the consistency of data but also improve the speed and efficiency of data query.

Description of drawings

1 is a flowchart of an embodiment of a data distribution and parallel processing method provided by the present invention;

Fig. 2 is the flow chart of solving the concurrent conflict in the data distribution and parallel processing method provided by the present invention;

3 is a flowchart of an embodiment of a data distribution and parallel processing system provided by the present invention;

FIG. 4 is a schematic structural diagram of an MPP provided by the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are for the purpose of distinguishing two entities with the same name but not the same or non-identical parameters. It can be seen that "first" and "second" It is only for the convenience of expression and should not be construed as a limitation to the embodiments of the present invention, and subsequent embodiments will not describe them one by one.

Referring to FIG. 1 , it is a flowchart of an embodiment of a data distribution and parallel processing method provided by the present invention. The data distribution and parallel processing method includes:

Step 101: Receive an initial query instruction input by the user terminal; wherein, the user usually inputs some continuous character strings at the user terminal, and the database usually needs to be parsed to obtain identifiable sentences or instruction information.

Step 102: Parse the initial query instruction to obtain a real query statement that meets the standard, and use it as a real query instruction; wherein, the common query parsing part mainly includes steps such as syntax analysis, lexical analysis, semantic analysis, and statement analysis. The function of syntax analysis is to replace an input string with a structure that describes the string, so that the computer can more easily understand what the user-input string means. This stage consists of three processes, namely lexical analysis, syntax analysis, and outputting abstract syntax trees. The lexical analyzer is a DeterministicFinite Automata, which can convert the input character set into words according to our defined lexical grammar. After lexical analysis comes grammatical analysis. The result of lexical analysis is used as the input of grammatical analysis. On the basis of lexical analysis, grammatical analysis determines whether the words input by the user conform to grammatical logic. Taking SQL as an example, semantic analysis involves related theories and concepts of SQL standards and SQL optimization. Semantic analysis includes logical analysis and physical analysis. Logical analysis is a general mathematical analysis process and has nothing to do with the underlying distributed environment. Physical analysis, on the other hand, transforms the results of logical analysis, which is closely related to the underlying execution environment.

Abstract Syntax Tree (AST) is a representation of the tree structure of user input sentences, each node on the tree is a word, and the structure of the tree reflects the grammar. The abstract syntax tree is constructed along with the process of syntax analysis. After the syntax analysis is completed, the syntax analyzer will output an abstract syntax tree, and the user's input corresponds to the structure content of the abstract syntax tree.

Step 103: Send the real query instruction to the work distribution node, and distribute the real query instruction to different data nodes through the work distribution node; wherein, the work distribution node and the data node refer to the database All server nodes are divided into nodes according to preset functions; and the work distribution node is used to manage the task distribution and query results of the data nodes; wherein, in the MPP architecture, the entire database includes a large number of relatively independent servers, each of which serves as a An independent node, so referring to a single server as a server node, node, server, etc., means the same in what it actually means. Generally speaking, in a database, one work distribution node can be set to manage the remaining server nodes, that is, data nodes; it is also possible to set multiple work distribution nodes to manage their respective data nodes. In this way, by setting work distribution nodes to manage data nodes, the storage capacity, access pressure and other factors of each server node can be roughly balanced. For example, data in data nodes with high load can be migrated to data nodes with low load , to achieve load balancing.

Step 104: Execute the received real query instruction on each data node respectively, and return the query result to the work distribution node, while maintaining the update of the data information table on each server node; wherein, the data information table uses In order to record the update of data and maintain the consistency of data on each server node through communication between server nodes; that is, any modification made to the data will be recorded in the data information table. The update refers to the synchronous modification on its backup or copy, so that the data information table can also be guaranteed to be consistent.

Step 105, on each server node, back up the query table corresponding to the real query instruction, so that each server node saves a copy of the query table; wherein, the query table refers to the real query instruction (for example, : select[id] command, etc.) after command query, a data table that stores the query results obtained; the query table also includes all query commands. A query task in a distributed database refers to a sequence of operations. Any operation on a distributed database eventually becomes a sequence of storage operations on the database, and its execution is also distributed. Therefore, the query table is backed up at each server node, so that each server node can execute query instructions through the query table or a copy of the query table, which not only enables concurrent execution of query instructions, but also improves query efficiency by maintaining a copy of the query table.

Step 106: On each server node, when a concurrency conflict is detected, the real query instruction is executed concurrently by calling the copy of the query table on the current server node. Among them, in the prior art, when a concurrency conflict is encountered, it is necessary to wait for a global response before the query can be performed, resulting in low data query speed and efficiency, and poor data consistency at the same time. The invention realizes the concurrent execution of the query statement by storing the query table copy in advance, and then executing the query task by calling the query table copy when encountering a concurrency conflict, which can greatly shorten the response time and improve the speed and efficiency of data query.

It can be seen from the above embodiments that the method for data distribution and parallel processing provided by the present invention divides all server nodes in the database into work distribution nodes and data nodes, so that the work distribution nodes can manage the task distribution of the data nodes and the query results. Return; in this way, the data nodes can achieve load balancing according to different usage states. By keeping the data information table updated, the consistent synchronization of the data in the server node can be ensured. By backing up a copy of the query table on each server node, each server node can execute query instructions concurrently, thereby shortening corresponding time and improving query efficiency. The consistency of distributed data can be guaranteed by invoking an adaptive backup to compare with the current query task when there is a concurrency conflict. Therefore, the data distribution and parallel processing method provided by the present invention can not only ensure the consistency of data, but also improve the speed and efficiency of data query. At the same time, based on the above steps, the present invention also overcomes the problem of concurrent conflicts in the query, so that the entire query process is more stable and reliable.

In some optional embodiments of the present invention, it is necessary to first divide the database into several subsets according to preset rules. For example, all the data in the database are evenly divided into disjoint n equal parts by column. It consists of several columns; alternatively, divide all the data into disjoint n equal parts by rows, and each part consists of several rows; or divide the data into n disjoint subsets according to the similarity of the data, such as greater than The data with the average value is classified into the same subset, and the other data is classified into another subset; or, the data is divided into n disjoint subsets according to the difference of the data, such as the data with the variance in the column greater than a certain constant k It is classified into the same subset, and other data is classified into another subset; in this way, the entire database is divided into several independent individuals and subsets, each subset is processed on the same server, a total of n servers. Enables independent decision making on each server, given whether the current task should be aborted or submitted. The principle of judgment is that the query task can be submitted when there is no concurrency conflict. If there is a concurrency conflict, for example, when a file is being written by one program, it cannot be written by another program at the same time. Call the adaptive replication backup of the subset on the current server and compare it with the current task to make a decision.

The adaptive backup refers to backing up important data on multiple servers for disaster and fault tolerance in a distributed system. There are many adaptive replication backup methods, including data block backup, dynamic backup, etc. If you want to maintain the consistency of data in multiple servers deployed in a distributed manner, you must ensure that data write operations on all servers are performed, or none of them are performed. However, when a server executes a local transaction, it cannot know the result of the execution of local transactions in other servers. So I don't know whether this transaction should be aborted or committed. If there is a concurrency conflict on a certain subset, that is, it is occupied by the write operation of a certain program, and other programs cannot perform the write operation on the current subset, the present invention proposes to replace the local execution of the current subset by accessing the backup of the current subset. In this way, the local transaction execution results of all servers can be obtained, and a decision can be given. The specific method is: if the transaction operations of all participating subsets of the task are actually executed successfully, the decision to submit the query task is returned; If the actual execution of the set's transaction operation fails, it returns a decision to withdraw the query task. Therefore, the MPP-based shared-nothing resource structure requires a global deadlock detection mechanism, a two-phase commit protocol, etc. to ensure transaction integrity and recovery. However, the solution adopted in the present invention does not require these time-consuming methods, thereby improving the speed of the query task.

In some optional embodiments of the present invention, referring to FIG. 2 , when a concurrent conflict is detected, the step 106 of concurrently executing the real query instruction by calling the copy of the query table on the current server node further includes:

Step 201, the real query instruction is detected in each server node to determine whether there is a concurrency conflict, if there is no concurrency conflict, go to step 202, if there is a concurrency conflict, go to step 203;

Step 202, according to Step 201, if there is no concurrency conflict, submit a query task; wherein, the query task is the query task corresponding to the real query instruction;

Step 203, according to Step 201, if there is a concurrency conflict, call the adaptive backup on the current server node to compare with the current query task, and execute the query task on the adaptive backup;

Step 204, determine whether the query tasks in all the server nodes are successfully executed, if so, go to step 205, otherwise, go to step 206;

Step 205, according to step 204, if the query tasks in all the server nodes are successfully executed, the decision to submit the query task is returned;

Step 206, according to step 204, if there is an unsuccessful query task, the decision to withdraw the query task is returned.

In this way, in view of the concurrency conflict problem in the server node, the query task can also be executed by invoking the adaptive backup. Specifically, each server has record information of an adaptive backup. When there is a concurrent conflict of query commands and an adaptive backup needs to be called, the current server can determine the storage of the record by accessing the adaptive backup record on the current server. location, that is, in which server node the adaptive backup is stored, and then compare the adaptive backup with the current query task and make a decision, which implements the query task on the adaptive backup data on the current server, thereby maintaining data consistency At the same time, the overall query task can be effectively executed concurrently, thereby greatly improving the speed and efficiency of data query.

In some optional embodiments of the present invention, the method for dividing all server nodes in the database into work assignment nodes and data nodes according to preset functions includes:

Detect the local storage capacity and computing processing capacity of each server node;

Determine whether the computing processing capability of the current server node is greater than the preset performance threshold, if so, the current server node is set as a work distribution node, otherwise the current server node is set as a data node;

or,

It is judged whether the local storage amount of the current server node is greater than the preset storage threshold, and if so, the current server node is set as a data node; otherwise, the current server node is set as a work distribution node.

In this way, different server nodes can correspondingly implement different functions according to their own characteristics, and can maximize the use of the resources of the server itself. At the same time, this distribution method also makes the management of the database more orderly and efficient.

In some optional embodiments of the present invention, the step of sending the real query instruction to a work distribution node, and distributing the real query instruction to different data nodes through the work distribution node further includes:

Detect the specific usage of each server node;

According to the detection results, a preset balancing strategy is adopted, and data nodes and corresponding spaces are allocated on each server node.

Wherein, the specific usage includes: the communication situation of the server, the storage space usage rate of each data node, and other situations or data related to the performance of the server node. In this way, different query tasks can be implemented according to different situations of the server node. distribute. Ability to optimize task processing speed and efficiency.

Further, in some optional embodiments, the balancing strategy is a strategy with the lowest space usage rate.

In some optional embodiments of the present invention, the step of maintaining the update of the data information table on each server node further includes: copying and storing the location information and related operation information of the data information table. In this way, by saving the location information and related operation information of the data information table, information recording of all data processing can be realized, so that the user can realize the withdrawal operation and other processing according to the information.

In some optional embodiments of the present invention, the step of maintaining the update of the data information table on each server node further includes:

All data in each data node is backed up to the designated data node; that is, data security is ensured by backing up the data, and adaptive backup data is obtained.

Keep the synchronous update of all data in each data node and the corresponding data backup; by saving the synchronous update of the data, all the data in each data node and the backed up data are kept consistent.

The location information of each data node is stored on each data node, which not only enables the data node to be found through the location information, but also facilitates the implementation of a recall operation.

The location information of each data node can be saved in the database by updating the primary database copy specifying the database address. All the information of the original data can be stored in the data node, and according to the information, the retrieval of the backup data is performed in the data of the corresponding data node or the backup data node.

In some optional embodiments of the present invention, the step of parsing the initial query instruction further includes:

Perform lexical analysis on the characters input by the user to obtain words that meet the standard;

Perform grammatical analysis on multiple consecutive words to obtain sentences that conform to grammatical logic, and construct an abstract syntax tree at the same time;

Convert the logical SQL statement to the real SQL statement according to the abstract syntax tree, and obtain the real query statement that meets the standard, which is used as the real query instruction.

In this way, through the step-by-step analysis of the initial query command, the real query command can be obtained, and then the subsequent data query task can be executed.

Optionally, the data distribution and parallel processing method of the present invention pre-stores a copy of the query table on each server node, so when there is a concurrent conflict between the query commands in the current server node, the current server node cannot execute the query command. , after returning the result that the processing node cannot query, the processing node finds the backup server where the adaptive backup corresponding to the current server data is located through the data information table, and then instructs the backup server to query the backup data according to the query instruction in the copy of the query table, Complete the data query.

Referring to FIG. 3, it is a flowchart of an embodiment of the data distribution and parallel processing system provided by the present invention. The data distribution and parallel processing system includes:

An input unit 301, configured to receive an initial query instruction input by a user terminal;

A parsing unit 302, configured to parse the initial query instruction to obtain a real query statement that meets the standard, and use it as a real query instruction;

A distribution unit 303, configured to send the real query instruction to a work distribution node, and distribute the real query instruction to different data nodes through the work distribution node; wherein, the work distribution node and the data node are Refers to the nodes in which all server nodes in the database are divided according to preset functions; and the work assignment nodes are used to manage the task assignment and query results of data nodes;

The processing unit 304 is used to execute the received real query instruction on each data node respectively, and return the query result to the work distribution node, while maintaining the update of the data information table on each server node;

A backup unit 305, configured to back up the look-up table corresponding to the real query instruction on each server node, so that each server node saves a copy of the look-up table;

The concurrency unit 306 is configured to, on each server node, concurrently execute the real query instruction by calling the copy of the query table on the current server node when a concurrency conflict is detected.

It can be seen from the above embodiments that the data distribution and parallel processing system of the present invention backs up the query table through the backup unit 305, and then implements the concurrent processing of query tasks through the concurrent unit 306, so that the data distribution and parallel processing are performed. The processing system can not only ensure the consistency of data but also improve the speed and efficiency of data query.

Those of ordinary skill in the art should understand that the discussion of any of the above embodiments is only exemplary, and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples; under the spirit of the present invention, the above embodiments or There may also be combinations between technical features in different embodiments, steps may be carried out in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in detail for the sake of brevity.

Additionally, well known power/ground connections to integrated circuit (IC) chips and other components may or may not be shown in the figures provided in order to simplify illustration and discussion, and in order not to obscure the present invention. . Furthermore, devices may be shown in block diagram form in order to avoid obscuring the present invention, and this also takes into account the fact that the details regarding the implementation of these block diagram devices are highly dependent on the platform on which the invention will be implemented (i.e. , these details should be fully within the understanding of those skilled in the art). Where specific details (eg, circuits) are set forth to describe exemplary embodiments of the invention, it will be apparent to those skilled in the art that these specific details may be used without or with changes The present invention is carried out below. Accordingly, these descriptions are to be considered illustrative rather than restrictive.

Although the present invention has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations to these embodiments will be apparent to those of ordinary skill in the art from the foregoing description. For example, other memory architectures (eg, dynamic RAM (DRAM)) may use the discussed embodiments.

Embodiments of the present invention are intended to cover all such alternatives, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omission, modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (9)

1. a data distribution and parallel processing method, is characterized in that, comprises: Receive the initial query command input by the client; Analyzing the initial query instruction to obtain a real query statement that meets the standard, and use it as a real query instruction; Send the real query instruction to the work distribution node, and distribute the real query instruction to different data nodes through the work distribution node; wherein, the work distribution node and the data node refer to all server nodes in the database Nodes divided according to preset functions; and the work assignment nodes are used to manage task assignments and query results of data nodes; Execute the received real query instruction on each data node, and return the query result to the work distribution node, and keep the update of the data information table on each server node; wherein, the data information table is used to record data update; On each server node, the query table corresponding to the real query instruction is backed up, so that a copy of the query table is saved on each server node; A data table that stores the obtained query results; On each server node, when a concurrency conflict is detected, the real query instruction is executed concurrently by calling the copy of the query table on the current server node. 2. The method according to claim 1, wherein the step of concurrently executing the real query instruction by calling the copy of the query table on the current server node when a concurrency conflict is detected comprises: Detecting whether there is a concurrency conflict in the real query instruction in each server node; If there is no concurrency conflict, submit a query task; wherein, the query task is a query task corresponding to a real query instruction; If there is a concurrency conflict, call the query table backup on the current server node to compare with the current query task, and execute the query task on the query table backup; If the query tasks in all server nodes are successfully executed, the decision to submit the query task is returned, otherwise, the decision to withdraw the query task is returned. 3. The method according to claim 1, wherein the method for dividing all server nodes in the database into work assignment nodes and data nodes according to preset functions comprises: Detect the local storage capacity and computing processing capacity of each server node; Determine whether the computing processing capability of the current server node is greater than the preset performance threshold, if so, the current server node is set as a work distribution node, otherwise the current server node is set as a data node; or, It is judged whether the local storage amount of the current server node is greater than the preset storage threshold, and if so, the current server node is set as a data node; otherwise, the current server node is set as a work distribution node. 4. The method according to claim 1, wherein the real query instruction is sent to a work distribution node, and the real query instruction is distributed to different data nodes through the work distribution node. Steps also include: Detect the specific usage of each server node; According to the detection results, a preset balancing strategy is adopted, and data nodes and corresponding spaces are allocated on each server node. 5 . The method according to claim 4 , wherein the balancing strategy is a strategy with the lowest space utilization rate. 6 . 6. The method according to claim 1, wherein the step of maintaining the update of the data information table on each server node further comprises: The location information and related operation information of the data information table are copied and stored. 7. The method according to claim 1, wherein the step of maintaining the update of the data information table on each server node further comprises: Back up all data in each data node to the specified data node; Keep all data in each data node and the synchronous update of the corresponding data backup; The location information of each data node is saved on each data node. 8. The method according to claim 1, wherein the step of parsing the initial query instruction further comprises: Perform lexical analysis on the characters input by the user to obtain words that meet the standard; Perform grammatical analysis on multiple consecutive words to obtain sentences that conform to grammatical logic, and construct an abstract syntax tree at the same time; Convert the logical SQL statement to the real SQL statement according to the abstract syntax tree, and obtain the real query statement that meets the standard, which is used as the real query instruction. 9. A data distribution and parallel processing system, comprising: an input unit for receiving an initial query instruction input by the user terminal; a parsing unit, configured to parse the initial query instruction to obtain a real query statement that meets the standard, and use it as a real query instruction; A distribution unit, configured to send the real query instruction to a work distribution node, and distribute the real query instruction to different data nodes through the work distribution node; wherein, the work distribution node and the data node refer to All server nodes in the database are divided into nodes according to preset functions; and the work distribution nodes are used to manage task distribution and query results of data nodes; The processing unit is used to execute the received real query instruction on each data node respectively, and return the query result to the work distribution node, and at the same time keep the update of the data information table on each server node; wherein, the data information Tables are used to record data updates; A backup unit, configured to back up the look-up table corresponding to the real query instruction on each server node, so that each server node saves a copy of the look-up table; wherein, the look-up table refers to the real query instruction After the instruction query is performed, a data table that stores the obtained query result; The concurrency unit is used on each server node to concurrently execute the real query instruction by calling the copy of the query table on the current server node when a concurrency conflict is detected.