Batch task processing method and user behavior integral query method

Abstract

The embodiment of the invention provides a batch task processing method and a user behavior integral query method, and the method comprises the steps: distributing tasks to corresponding calculation clusters and disaster recovery clusters according to an identification interval of identification information of each task, carrying out the partition scheduling of the batch tasks through a plurality of calculation clusters, and carrying out the user behavior integral query. And tasks are synchronously scheduled and processed by using the computing cluster and the disaster recovery cluster, so that the task parallel capability and the processing efficiency are improved. Moreover, by performing odd-bit and even-bit task fragmentation processing on each identification interval, the computing cluster and the disaster recovery cluster can respectively schedule and process different task fragmentation, resource waste and data conflict caused by scheduling and processing the same task by the computing cluster and the disaster recovery cluster are avoided, meanwhile, load balance of the computing cluster and the disaster recovery cluster is ensured, and the load balancing efficiency of the computing cluster and the disaster recovery cluster is improved. In addition, under the condition that identification interval configuration changes, task fragments of the computing cluster and the disaster recovery cluster do not need to be reconfigured, and the stability and efficiency of batch task processing are improved.

Description

Technical Field

[0001] One or more embodiments of the present specification relate to the field of big data processing technology, and in particular, to a batch task processing method and a method for querying user behavior points. Background Art

[0002] Batch task processing refers to the method of batch processing of a certain amount of data or tasks during the data processing process. The characteristics of batch tasks are that the data volume is usually large, and the tasks can be automatically executed in a certain order without human intervention until the batch task processing is completed.

[0003] The processing efficiency of batch tasks depends on the task scheduling capability of the computing cluster. Since batch tasks generally do not have high requirements for timeliness, related technical solutions adopt a centralized scheduling approach, using one or more computing clusters to execute tasks sequentially. The data processing cycle is long, which makes it difficult to meet the needs of some scenarios with high timeliness requirements. Summary of the invention

[0004] To improve the efficiency of batch task processing, the embodiments of this specification provide a batch task processing method and device, a user behavior score query method and device, an electronic device, a storage medium, and a computer program product.

[0005] In a first aspect, an embodiment of the present specification provides a batch task processing method, which is applied to a data processing system, wherein the data processing system includes multiple computing clusters and a disaster recovery cluster corresponding to each computing cluster, and the method includes:

[0006] Distribute the batch tasks to the computing clusters and their disaster recovery clusters corresponding to the identification intervals according to the identification intervals to which the identification information of each task in the batch tasks belongs, wherein each computing cluster corresponds to one identification interval, and the identification information of the multiple computing clusters corresponding to the identification intervals is a continuous value;

[0007] For each computing cluster, the computing cluster reads and executes a task with a first identifier from the distributed batch tasks, and the disaster recovery cluster corresponding to the computing cluster reads and executes a task with a second identifier from the distributed batch tasks, wherein one of the first identifier and the second identifier is identifier information of an odd number of bits in the identifier interval, and the other is identifier information of an even number of bits in the identifier interval;

[0008] The task processing results of each task are stored in the computing cluster and its disaster recovery cluster.

[0009] In some embodiments, the task includes a first task and a second task different from the first task;

[0010] The process of the computing cluster reading and executing the task of the first identification includes: in a continuous time period, the computing cluster sequentially and alternately reads and executes the first task of the first identification and the second task of the first identification;

[0011] The process of the disaster recovery cluster reading and executing the task with the second mark includes: in a continuous time period, the disaster recovery cluster sequentially and alternately reads and executes the first task with the second mark and the second task with the second mark.

[0012] In some embodiments, in continuous time periods, the computing cluster alternately reads and executes the first task of the first identifier and the second task of the first identifier in sequence, including: in odd time periods, the computing cluster reads and executes the first task of the first identifier; in even time periods, the computing cluster reads and executes the second task of the first identifier.

[0013] In some embodiments, in continuous time periods, the disaster recovery cluster reads and executes the first task with the second identifier and the second task with the second identifier alternately in sequence, including: in odd time periods, the disaster recovery cluster reads and executes the first task with the second identifier; in even time periods, the disaster recovery cluster reads and executes the second task with the second identifier.

[0014] In some implementations, the task includes a user behavior score calculation task and / or a user behavior parameter calculation task, wherein the user behavior score calculation task depends on the user behavior parameters.

[0015] In some implementations, when the task includes a user behavior score calculation task, the process of executing the user behavior score calculation task includes:

[0016] Acquire user behavior data, wherein the user behavior data includes historical behavior data and current behavior data, and the current behavior data includes at least part of user behavior parameters obtained in advance;

[0017] A plurality of behavior factor parameters are generated according to the historical behavior data and the current behavior data, and the user behavior score is generated according to the plurality of behavior factor parameters, and the task processing result includes the user behavior score.

[0018] In some implementations, the acquiring of the user behavior data includes: reading the user behavior data from a database corresponding to the user representation according to a user identifier.

[0019] In some implementations, storing the task processing result of each task in the computing cluster and its disaster recovery cluster includes: storing each user behavior score in a database corresponding to the user identifier.

[0020] In a second aspect, the embodiments of this specification provide a method for querying user behavior points, including:

[0021] Acquire a user query instruction, wherein the user query instruction includes a user identifier;

[0022] According to the user identifier, query the user behavior score corresponding to the current computing cycle from the database, the user behavior score of each computing cycle in the database is calculated by the batch task processing method described in any of the above embodiments, and the task processing result includes the user behavior score;

[0023] In response to the user behavior score corresponding to the user identification in the current calculation cycle not being calculated, executing the user behavior score calculation task to obtain the user behavior score in the current calculation cycle; in response to the user behavior score corresponding to the user indication in the current calculation cycle being calculated, reading the user behavior score;

[0024] Display the user behavior score for the current calculation cycle.

[0025] In a third aspect, an embodiment of the present specification provides a batch task processing device, which is applied to a data processing system, wherein the data processing system includes multiple computing clusters and a disaster recovery cluster corresponding to each computing cluster, and the device includes:

[0026] The task scheduling module is configured to distribute the batch tasks to the computing clusters and their disaster recovery clusters corresponding to the identification intervals according to the identification intervals to which the identification information of each task in the batch tasks belongs, wherein each computing cluster corresponds to one identification interval, and the identification information of the multiple computing clusters corresponding to the identification intervals is a continuous value;

[0027] The task execution module is configured to, for each computing cluster, read and execute a task with a first identifier from the distributed batch tasks, and a disaster recovery cluster corresponding to the computing cluster reads and executes a task with a second identifier from the distributed batch tasks, wherein one of the first identifier and the second identifier is identifier information of an odd number of bits in an identifier interval, and the other is identifier information of an even number of bits in an identifier interval;

[0028] The result storage module is configured to store the task processing result of each task in the computing cluster and its disaster recovery cluster.

[0029] In some implementations, the task includes a first task and a second task different from the first task; and the task execution module is configured to:

[0030] In a continuous time period, the computing cluster sequentially and alternately reads and executes a first task with a first identifier and a second task with a first identifier;

[0031] In a continuous time period, the disaster recovery cluster reads and executes the first task with the second mark and the second task with the second mark alternately in sequence.

[0032] In some implementations, the task execution module is configured to:

[0033] In odd time periods, the computing cluster reads and executes a first task with a first identifier; in even time periods, the computing cluster reads and executes a second task with a first identifier;

[0034] In an odd-numbered time period, the disaster recovery cluster reads and executes the first task with the second identifier; in an even-numbered time period, the disaster recovery cluster reads and executes the second task with the second identifier.

[0035] In some implementations, the task includes a user behavior score calculation task and / or a user behavior parameter calculation task, wherein the user behavior score calculation task depends on the user behavior parameters.

[0036] In some implementations, the task execution module is configured to:

[0037] Acquire user behavior data, wherein the user behavior data includes historical behavior data and current behavior data, and the current behavior data includes at least part of user behavior parameters obtained in advance;

[0038] A plurality of behavior factor parameters are generated according to the historical behavior data and the current behavior data, and the user behavior score is generated according to the plurality of behavior factor parameters, and the task processing result includes the user behavior score.

[0039] In some implementations, the task execution module is configured to:

[0040] The user behavior data is obtained by reading from a database corresponding to the user representation according to the user identifier;

[0041] Each user behavior score is stored in a database corresponding to the user identification.

[0042] In a fourth aspect, the embodiments of this specification provide a user behavior score query device, including:

[0043] An instruction acquisition module is configured to acquire a user query instruction, wherein the user query instruction includes a user identifier;

[0044] a points query module configured to query the user behavior points corresponding to the current calculation cycle from the database according to the user identifier, wherein the user behavior points of each calculation cycle in the database are calculated by the batch task processing method described in any of the above embodiments, and the task processing result includes the user behavior points; in response to the user behavior points corresponding to the user identifier in the current calculation cycle not being calculated, executing the user behavior points calculation task to obtain the user behavior points of the current calculation cycle; in response to the user behavior points corresponding to the user identifier in the current calculation cycle being calculated, reading the user behavior points;

[0045] The points display module is configured to display the user behavior points in the current calculation cycle.

[0046] In a fifth aspect, an embodiment of the present specification provides an electronic device, including:

[0047] processor;

[0048] A memory stores computer instructions, wherein the computer instructions, when executed by a processor, implement the method described in any of the above embodiments.

[0049] In a sixth aspect, an embodiment of this specification provides a computer program product, which, when executed, implements the method described in any of the above embodiments.

[0050] The method of the implementation mode of this specification includes distributing tasks to the corresponding computing cluster and its disaster recovery cluster according to the identification interval described in the identification information of each task, partitioning and scheduling batch tasks through multiple computing clusters, and using the computing cluster and the disaster recovery cluster to synchronously schedule and process tasks, thereby improving the task parallel capability and processing efficiency. In addition, by performing task slicing processing of odd and even digits for each identification interval, the computing cluster and the disaster recovery cluster can respectively schedule and process different task slicings, thereby avoiding resource waste and data conflicts when both schedule and process the same task, and ensuring the load balancing of the computing cluster and the disaster recovery cluster. In addition, when the configuration of the identification interval changes, there is no need to reconfigure the task slicing of the computing cluster and the disaster recovery cluster, thereby improving the stability and efficiency of batch task processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is an architectural diagram of a data processing system in an exemplary embodiment of this specification.

[0052] Figure 2 FIG. 1 is a logical diagram of task scheduling of a data processing system in an exemplary embodiment of the present specification.

[0053] Figure 3is a flowchart of a batch task processing method in an exemplary embodiment of this specification.

[0054] Figure 4 It is a schematic diagram of a batch task processing method in an exemplary embodiment of this specification.

[0055] Figure 5 is a flowchart of a batch task processing method in an exemplary embodiment of this specification.

[0056] Figure 6 It is a system architecture diagram that the financial service platform relies on in an exemplary implementation of this specification.

[0057] Figure 7 It is a flowchart of a method for querying user behavior points in an exemplary embodiment of this specification.

[0058] Figure 8 1 is a structural block diagram of a batch task processing device in an exemplary embodiment of this specification.

[0059] Fig. 9 It is a structural block diagram of a user behavior score query device in an exemplary implementation of this specification.

[0060] Fig.10 is a structural block diagram of an electronic device in an exemplary embodiment of this specification. DETAILED DESCRIPTION

[0061] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this manual are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0062] User behavior points refer to quantitative scores generated based on user historical behavior or data to evaluate a user's ability to behave. For example, in the case of a financial service platform, the financial service platform often generates a corresponding credit score for each user, which can reflect the user's credit ability. For another example, the financial service platform can also generate a corresponding financial management score for the user, which can reflect the user's financial management ability.

[0063] In the financial service platform, user behavior points need to be dynamically updated with the behavior data continuously generated by users, so that the user behavior points can accurately reflect the user's current behavior ability. For a large number of users, the task of calculating the user behavior points of each user is a batch task, also called a batch task, which means that the financial service platform needs to calculate the user behavior points of each user based on the behavior data of each user and store them. When the user calls for query, the latest user behavior points are read from the database and displayed to the user.

[0064] In the related art, the batch task of calculating user behavior points does not have high requirements for timeliness. For example, user behavior points are generally updated weekly, that is, every 7 days is a cycle, so the financial service platform can calculate the user behavior points of all users in a round within 7 days. At present, the batch task processing for user behavior points adopts a centralized scheduling solution, using one or more computing clusters to sequentially schedule and process tasks until the batch task processing is completed.

[0065] As the requirements for data timeliness continue to increase, the batch task processing methods in related technologies are difficult to meet the business needs of modern finance for agile response. For example, in the scenario of daily update of user behavior points, it is necessary to complete a round of calculation and storage of user behavior points for all users every day. The batch task processing of tens of millions is difficult to meet the daily update timeliness requirements.

[0066] Based on this, the embodiments of this specification provide a batch task processing method and device, a user behavior points query method and device, a data processing system, a storage medium and a computer program product, which aim to utilize the system disaster recovery mechanism and combine it with a distributed scheduling method to improve the scheduling and processing speed of batch tasks and meet the needs of task scenarios with high timeliness requirements.

[0067] System Disaster Recovery refers to the measures and calculations taken to ensure that the data system can quickly recover or continue to provide services when an accident occurs and the service is interrupted. Among them, system backup is a key component of system disaster recovery. For example, for a data system, in order to ensure the normal operation of the service when encountering natural disasters, technical failures, human errors or other unexpected events, a disaster recovery system can be set up for the data system. The configuration of the disaster recovery system is exactly the same as that of the data system, and the capabilities it can provide are also the same, so that the disaster recovery system can ensure the normal operation of the service when a disaster occurs.

[0068] In the related art, when processing batch tasks, the data system is used as the main system, and the batch task processing is concentrated in the main system. The main system schedules and executes the tasks sequentially, and the disaster recovery system is only used as a data backup, resulting in a waste of computing resources. In order to improve the efficiency of batch task processing, the data system and the disaster recovery system can be used to jointly schedule and execute batch tasks, but this approach faces two major problems: one is the data conflict problem. In order to avoid data conflicts between the two systems, an efficient and stable task scheduling solution is required to reasonably schedule tasks; the other is the load balancing problem, which requires a reasonable distribution of the load of the data system and the disaster recovery system.

[0069] Based on this, the embodiments of this specification provide a batch task processing method and device, which can be applied to a data processing system. Figure 1 The following is a diagram showing the architecture of a data processing system according to an exemplary embodiment of the present specification. Figure 1 Provide explanation.

[0070] In the implementation mode of this specification, the data processing system includes multiple computing clusters, and each computing cluster is provided with a corresponding disaster recovery cluster. Figure 1 In the example, the data processing system includes n computing clusters Rz41A, Rz42A, ..., Rz4nA, and n disaster recovery clusters Rz41B, Rz42B, ..., Rz4nB. Taking Rz41A and Rz41B as examples, the computing cluster Rz41A and the disaster recovery cluster Rz41B have exactly the same configuration and data, and the capabilities they can provide are also exactly the same.

[0071] Each computing cluster or disaster recovery cluster can be understood as a physical computer room, which may include one or more servers for providing computing power, and may also include one or more databases for providing storage power. A server or database may refer to a physical device or a logical device configured through virtualization technology. In the implementation manner of this specification, the number of servers and databases included in different computing clusters may be the same or different, and there is no restriction on the specific number of servers or databases included in the computing cluster.

[0072] In the implementation of this specification, it is necessary to partition and manage multiple computing clusters and disaster recovery clusters according to the task identification information of the batch task. The task identification information refers to the task identification used to classify the batch task. For example, in some implementations, the batch task takes the user behavior score calculation task as an example, and the task identification information can be the user's UID (User Identification) or a part of the digital number in the UID.

[0073] See for example Figure 1As shown, in an example, assuming that the system includes 4 computing clusters Rz41A~Rz44A and 4 disaster recovery clusters Rz41B~Rz44B, the task identification information can use two digits of the user's UID, so that the identification information interval corresponding to all tasks is [00,99], totaling 100 identification information.

[0074] In this example, the identification information interval [00,99] can be evenly divided into 4 identification intervals according to the number of computing clusters, and allocated to each computing cluster and its disaster recovery cluster. Figure 1 In the example, the identifier interval corresponding to computing cluster Rz41A is [00,01,02,…,24]. Similarly, the identifier interval corresponding to the disaster recovery cluster Rz41B of computing cluster Rz41A is also [00,01,02,…,24]. The identifier interval corresponding to computing cluster Rz42A is [25,26,27,…,49]. Similarly, the identifier interval corresponding to the disaster recovery cluster Rz42B of computing cluster Rz42A is also [25,26,27,…,49]. The identifier interval corresponding to computing cluster Rz43A is [50,51,52,…,74]. Similarly, the identifier interval corresponding to the disaster recovery cluster Rz43B of computing cluster Rz43A is also [50,51,52,…,74]. The identifier interval corresponding to the computing cluster Rz44A is [75,76,77,…,99]. Similarly, the identifier interval corresponding to the disaster recovery cluster Rz44B of the computing cluster Rz44A is also [75,76,77,…,99], as shown in the following Table 1:

[0075] Table 1 Identification interval

[0076] Computing Cluster Disaster Recovery Cluster Identification interval Rz41A Rz41B [00,01,02,…,24] Rz42 Rz42B [25,26,27,…,49] Rz43 Rz43 [50,51,52,…,74] R Rz4 [75,76,77,…,99]

[0077] By dividing the task identification information into intervals, tasks can be distributed according to the identification intervals during batch task processing, thereby realizing distributed scheduling and processing of batch tasks. Figure 2 Shown in Figure 1 The task scheduling and distribution logic based on the system architecture is combined with Figure 2 Provide explanation.

[0078] like Figure 2As shown, the data processing system of the implementation mode of this specification adopts a three-layer task distribution mechanism. In this example, at the first level (level 1), the system task scheduler distributes batch tasks to Rz41A~Rz44A and Rz41B~Rz44B according to the aforementioned identification interval, laying the foundation for subsequent refined task processing. At the second level (level 2), taking Rz41A and Rz41B as an example, the identification intervals corresponding to Rz41A and Rz41B are split, and a task processing process is created for each identification information (UID), so that each system process can focus on processing a task of an identification information. At the third level (level 3), the system can grab specific tasks from each identification information (UID) for processing. For example, taking UID:00 as an example, tasks "2088...00x" and "2088...00y" are tasks under the identification information "00".

[0079] In an example, each system process can capture multiple tasks for processing at a time. For example, each system process can capture 10 tasks for processing. Therefore, for a computing cluster of Rz41A, 250 tasks can be processed in parallel at the same time. For computing clusters of Rz41A to Rz44A, 1,000 tasks can be processed in parallel at the same time.

[0080] In the above task scheduling and processing process, for the same group of computing clusters and disaster recovery clusters, since the corresponding identification intervals are the same, data conflicts may occur. For example, taking computing cluster Rz41A and disaster recovery cluster Rz41B as an example, Rz41A schedules and processes tasks with identification information "00". At the same time, if Rz41B also schedules and processes tasks with identification information "00", it will cause repeated calculations, resulting in a waste of computing resources, and when writing task processing results, repeated writing is likely to cause database errors.

[0081] Therefore, in the implementation of this specification, based on the above system, the interval sharding scheduling method can be further combined to ensure the task parallel stability and load balancing of the computing cluster and the disaster recovery cluster. Figure 3 Provide explanation.

[0082] like Figure 3 As shown, in some implementations, the batch task processing method exemplified in this specification includes:

[0083] S310 . Distribute the batch tasks to the computing cluster and its disaster recovery cluster corresponding to the identification interval according to the identification interval to which the identification information of each task in the batch tasks belongs.

[0084] In the implementation manner of this specification, when scheduling and processing batch tasks, the system may first determine the identification interval to which each task belongs according to the identification information of each task.

[0085] For example, in the previous example, the batch task takes the user behavior score calculation task as an example. Each task represents the calculation of the user behavior score for a user. The identification information of each task can be represented by the two-digit number in the UID. The system task scheduler can obtain the identification information of each task, and then distribute the task to the computing cluster and its disaster recovery cluster corresponding to the identification interval based on the aforementioned division of the identification interval.

[0086] For example, in one example, the identification information of a task is "03". Therefore, through the identification interval shown in the above Table 1, it can be determined that the identification information of the task belongs to the identification interval [00, 01, 02, ..., 24], so that the system task scheduler can distribute the task to the computing cluster Rz41A and the disaster recovery cluster Rz41B.

[0087] Through the above method and process, the scheduling and distribution of batch tasks at the first level (level 1) can be realized, that is, the batch tasks are distributed to each computing cluster and its disaster recovery cluster according to different identification intervals.

[0088] S320: For each computing cluster, the computing cluster reads and executes a task with a first identifier from the distributed batch tasks, and a disaster recovery cluster of the computing cluster reads and executes a task with a second identifier from the distributed batch tasks.

[0089] In combination with the foregoing, it can be understood that each computing cluster has the same identification interval corresponding to its disaster recovery cluster, and therefore the distributed batch tasks are also the same. For example, taking computing cluster Rz41A and disaster recovery cluster Rz41B as an example, in order to avoid data conflicts between the two during task scheduling and processing, in the implementation manner of this specification, the tasks in the identification interval can be sharded so that the computing cluster Rz41A and the disaster recovery cluster Rz41B respectively process different task shards.

[0090] For example, in some implementations, the identification intervals corresponding to the computing cluster and the disaster recovery cluster may be divided into a first identification and a second identification, and the first identification may be identification information of odd-numbered bits in the identification interval, so that the second identification may be identification information of even-numbered bits in the identification interval. Alternatively, the first identification may be identification information of even-numbered bits in the identification interval, so that the second identification may be identification information of odd-numbered bits in the identification interval.

[0091] For example, in an example, still taking the computing cluster Rz41A and the disaster recovery cluster Rz41B as an example, the corresponding identification interval is [00,01,02,…,24], and the first identification can be divided into [00,02,04,06,08,10,12,14,16,18,20,22,24], and the second identification can be divided into [01,03,05,07,09,11,13,15,17,19,21,23].

[0092] Therefore, when scheduling tasks at the second level (level 2), computing cluster Rz41A can only call tasks with task identification information of the first identifier from all distributed batch tasks, and disaster recovery cluster Rz41B can only call tasks with task identification information of the second identifier from all distributed batch tasks. Then at the third level (level 3), computing clusters Rz41A and Rz41B can grab specific task execution from the scheduled task slices and obtain corresponding task processing results.

[0093] It can be understood that by dividing the identification interval into the first identification and the second identification, the computing cluster and the disaster recovery cluster can schedule and process different task slices respectively, avoiding scheduling and processing the same task by both, resulting in resource waste and data conflicts.

[0094] In addition, it is worth mentioning that the scheduling method in the implementation of this specification is different from simply processing batch tasks of the computing cluster and the disaster recovery cluster in batches. Instead, it divides the batch tasks into odd and even groups through identification information, which can ensure load balancing between the computing cluster and the disaster recovery cluster.

[0095] For example, in the above example, we can see that the identification intervals of computing cluster Rz41A and disaster recovery cluster Rz41B are continuous values. Through the task slicing processing of odd and even digits, it can be ensured that the number of task identification information allocated to computing cluster Rz41A and disaster recovery cluster Rz41B is basically consistent, so that the number of specific tasks scheduled and processed by the two is basically consistent, and the resources of computing cluster and disaster recovery cluster can be fully scheduled and utilized to ensure load balancing.

[0096] Moreover, the implementation of this specification performs task slicing for odd and even digits of the identification interval, which is different from the solution of fixedly configuring identification intervals for computing clusters and disaster recovery clusters. It can avoid the redundant operation of repeatedly configuring task slicing when the identification interval configuration changes.

[0097] For example, in a related technical example, if the computing cluster Rz41A is pre-configured to schedule tasks in the interval [00,01,…,11], and the disaster recovery cluster Rz41B is pre-configured to schedule tasks in the interval [12,13,14,…,24], once the identification intervals corresponding to the computing cluster Rz41A and the disaster recovery cluster Rz41B change, it is necessary to reconfigure the task shards that the computing cluster and the disaster recovery cluster need to schedule.

[0098] For example, the system consists of Figure 1 The architecture shown is expanded to include 10 computing clusters and 10 disaster recovery clusters. At this time, the identification intervals of computing cluster Rz41A and disaster recovery cluster Rz41B change to [00,01,02,…,09]. In the above solution, it is necessary to reconfigure the task slices of computing cluster Rz41A to be fixedly scheduled to [00,01,02,03,04], and configure the task slices of disaster recovery cluster Rz41B to be fixedly scheduled to [05,06,07,08,09]. The configuration operation is complicated.

[0099] In the embodiment of the present disclosure, no matter how the identification intervals of the computing cluster Rz41A and the disaster recovery cluster Rz41B change, the computing cluster Rz41A and the disaster recovery cluster Rz41B will always be scheduled in shards according to the identification information of the odd and even bits, thereby eliminating the need to reconfigure the task scheduling logic of the computing cluster Rz41A and the disaster recovery cluster Rz41B, greatly simplifying the complexity of system configuration operations.

[0100] S330: storing the task processing result of each task in the computing cluster and its disaster recovery cluster.

[0101] After the above task scheduling and processing process, the task processing result of each task can be calculated in turn, and then the task processing result can be stored.

[0102] For example, taking the user behavior score calculation task as an example, combined with Figure 2 As shown, after the computing cluster Rz41A calls and executes the task "2088...00x", the user behavior points corresponding to the task can be obtained, and then the calculated user behavior points are stored in the computing cluster Rz41A and the disaster recovery cluster Rz41B. It can be understood that in order to maintain data synchronization between Rz41A and Rz41B, both the task processing results calculated by Rz41A and the task processing results calculated by Rz41B must be synchronously written into Rz41A and Rz41B for storage. In addition, it can be understood that due to the aforementioned task slicing scheduling of the computing cluster Rz41A and the disaster recovery cluster Rz41B, there will be no data conflict when writing the task processing results, thereby accelerating data storage efficiency.

[0103] In some embodiments, when storing task processing results, a corresponding database can be pre-configured for each identification information, so that the task processing results of the task of the identification information are stored in the corresponding database. That is, a database is configured and maintained for each user of the identification information. The database is only used to store user data corresponding to the identification information, optimize the data access path, and improve data reading and writing efficiency.

[0104] The above only takes a group of computing clusters and disaster recovery clusters (i.e., Rz41A and Rz41B) as an example to illustrate the process of the batch task processing method. For other computing clusters and disaster recovery clusters, the above method and process can be executed in sequence, and no further description is given. Through the above batch task processing method, batch tasks are processed in sequence until all batch tasks are processed.

[0105] From the above, it can be seen that in the implementation of this specification, batch tasks are partitioned and scheduled through multiple computing clusters, and computing clusters and disaster recovery clusters are used to synchronously schedule and process tasks, thereby improving task parallelism and processing efficiency. In addition, by processing odd-numbered and even-numbered task slices for each identification interval, the computing cluster and the disaster recovery cluster can schedule and process different task slices respectively, avoiding resource waste and data conflicts when both schedule and process the same task, and ensuring load balancing between the computing cluster and the disaster recovery cluster, and avoiding the need to repeatedly configure redundant operations of task slices when the identification interval configuration changes.

[0106] It is worth noting that in some scenarios, for complex batch task processing jobs, the job process may need to include multiple types of tasks. For example, taking the scenario of calculating user behavior points as an example, in order to comprehensively evaluate user behavior points, more user behavior parameters may be used, and the calculation task of these user behavior parameters is also a batch task processing. For example, taking financial management points as an example of user behavior points, in order to calculate the financial management points of a certain user, it may be necessary to obtain the user's historical period of holding ratios, yields and other parameters. These parameters are user behavior parameters, and the process of calculating user behavior parameters for a large number of users is also a batch task processing process. From this, it can be understood that for complex batch task processing scenarios, multiple types of tasks may be included.

[0107] In some implementations of this specification, in order to improve the batch task processing efficiency of this complex multi-task scenario, a time-alternating peak-shifting processing method is proposed. The computing cluster or disaster recovery cluster can process two types of tasks alternately on the time axis, thereby making full use of cluster resources and greatly improving the batch task processing efficiency. Figure 4 Provide explanation.

[0108] In some implementations, the task type of a batch task includes a first task and a second task. For example, the first task is a user behavior score calculation task. Specifically, the user behavior score calculation task may be a financial management score calculation task, a credit score calculation task, etc. The second task is a user behavior parameter calculation task. Specifically, the user behavior parameter calculation task may be an AUM (Assets Under Management) calculation task, a position ratio calculation task, etc.

[0109] In the implementation of this specification, when the cluster schedules and executes tasks, the time axis can be divided into multiple continuous time periods according to the duration of task processing. For example, in one example, the duration of a computing cluster to capture and execute a task is about 200ms to 400ms. Considering factors such as data fluctuations, a time period can be set to 1s, that is, every 1s on the time axis is a time period.

[0110] In this example, combined with Figure 2 and Figure 4 As shown, taking computing cluster Rz41A and disaster recovery cluster Rz41B as an example, the scheduling of the first task or the second task by Rz41A and Rz41B can be referred to in the above Figure 3 In summary, the computing cluster Rz41A can schedule the first task or the second task with the first identifier, and the disaster recovery cluster Rz41B can schedule the first task or the second task with the second identifier.

[0111] For the computing cluster Rz41A, in the implementation manner of this specification, the computing cluster Rz41A can alternately read and execute the first task and the second task of the first identifier in sequence in continuous time periods. For example, in time period T, Rz41A can grab and execute the first task of the first identifier, and in time period T+1, Rz41A grabs and executes the second task of the first identifier, and in time period T+2, Rz41A continues to grab and execute the first task of the first identifier... and so on.

[0112] Similarly, for the disaster recovery cluster Rz41B, in the implementation manner of this specification, the disaster recovery cluster Rz41B can alternately read and execute the first task and the second task of the second identifier in sequence in continuous time periods. For example, in time period T, Rz41B can capture and execute the first task of the second identifier, and in time period T+1, Rz41B captures and executes the second task of the second identifier. In time period T+2, Rz41B continues to capture and execute the first task of the second identifier... and so on.

[0113] In the above example, when the cluster executes the task and obtains the task processing result, the task processing result can be stored in the corresponding database. Those skilled in the art may refer to the above and will not elaborate on it here.

[0114] Combination Figure 4 As shown in the figure, for each computing cluster or disaster recovery cluster, by alternating and staggering the processing tasks on the timeline, different types of batch task processing operations can be finely scheduled in time to ensure that the computing cluster can be in a computing state in each time period. This not only makes full use of the cluster computing resources and improves the processing efficiency of complex multi-tasking scenarios, but also disperses the database write pressure and improves the overall database processing efficiency and response speed.

[0115] In the following implementation modes of this specification, the user behavior points calculation task will be taken as an example to illustrate the specific task processing process. In addition, in the following implementation modes, the user behavior points will be illustrated using "financial management points" as an example. However, those skilled in the art will understand that user behavior points are not limited to "financial management points" but may also be, for example, "credit points", etc., which will not be elaborated in this specification.

[0116] The financial management score is an indicator launched by the financial service platform to objectively quantify the user's financial management ability. For users, understanding their own financial management score can provide more accurate financial management strategy guidance, further optimize asset allocation, and promote the steady growth of assets. The unique value of the financial management score system lies in its ability to transform users' complex financial management behavior patterns into intuitive quantitative scores, greatly enhancing the transparency and operability of personal financial management, and providing users with a new perspective on scientific and personalized asset management.

[0117] In the related art, the financial score update mechanism of the financial service platform is generally weekly, that is, the financial score of the user is updated once every 7 days. In the implementation of this specification, through the above-mentioned batch task processing method process, the batch task processing cycle for calculating the financial score of the user can be shortened by more than ten or even several hours, so that the financial score can be updated daily, that is, the financial score of the user is updated once a day, which greatly meets the high timeliness requirements of the financial service platform for data processing.

[0118] In the above implementation mode of this specification, the scheduling and processing process of the batch task of calculating the financial points has been explained in combination with the data processing system. Figure 5 The specific process of calculating the user's financial management score is explained.

[0119] like Figure 5 As shown, in some embodiments, the batch task processing method exemplified in this specification, the process of executing the user behavior score calculation task includes:

[0120] S510: Obtain user behavior data.

[0121] S520: Generate multiple behavior factor parameters according to the historical behavior data and the current behavior data, and generate user behavior points according to the multiple behavior factor parameters.

[0122] It is worth noting that in the schemes for calculating user financial management scores in related technologies, a quantitative evaluation is generally performed based on the data of the user's historical period, thereby generating a financial management score.

[0123] In the implementation of this specification, the behavioral factors that affect the financial management score are decomposed into two parts: "offline factors" and "online factors". Offline factors refer to the influencing factors abstracted from the user's offline behavior data, while online factors refer to the influencing factors abstracted from the user's current behavior data. Therefore, in the implementation of this specification, the user behavior data obtained also includes two parts: one is the historical behavior data used to represent the user's historical behavior, and the other is the current behavior data used to represent the user's current behavior.

[0124] For example, historical behavior data may include: average holding time of a fund, number of fixed investments in a fund, number of transactions in a historical period, etc. These behavior data are calculated based on historical data and are used to reflect the user's historical financial management habits. Current behavior data may include: user holdings, holding ratio of good products, AUM, etc. These behavior data are all related to the user's current behavior and can reflect the user's current financial situation.

[0125] In some implementations, the specific parameters of the historical behavior data and the current behavior data can also be calculated by the aforementioned batch task processing method, which is not described in detail in this specification. Taking a certain user as an example, when obtaining the user behavior data of the user, the user behavior data of the user can be obtained by reading the user ID (such as UID) of the user from the aforementioned corresponding database. Those skilled in the art can understand this, and this specification is not described in detail.

[0126] In some embodiments, after obtaining the user behavior data, multiple behavior factor parameters can be generated based on the historical behavior data and current behavior data included in the user behavior data. Each behavior factor parameter can be understood as a parameter obtained after quantifying a certain user behavior data, which is used to represent a sub-item score of the financial management score. After obtaining these behavior factor parameters, the final user behavior score can be calculated based on the behavior factor parameters, for example, by weighted fusion. The user behavior score is the financial management score. For the calculation process of quantifying and weighted fusion of the behavior factor parameters, those skilled in the art can refer to the relevant technology, and will not be described in detail.

[0127] After calculating the user's financial management score, the financial management score can be stored in the database corresponding to the user, waiting for the user to query and call. Through the process of the above implementation, the financial management score corresponding to each user can be calculated. In some examples of this implementation, combined with the aforementioned batch task processing method process, the above batch task processing method process can be executed once a day, and the latest financial management score is stored in each database, waiting for the user to query and call.

[0128] From the above, it can be seen that in the implementation of this specification, by combining historical behavior data and current behavior data to calculate user behavior points, the user's behavior ability can be evaluated more comprehensively, thereby improving the accuracy of the user's behavior points.

[0129] In some embodiments, this specification provides a method for querying user behavior points, which can be applied to a financial service platform. Figure 6 The system architecture diagram of the financial service platform in some embodiments of this specification is shown below. Figure 6 Provide explanation.

[0130] like Figure 6 As shown, the system architecture of the financial service platform may include a server 11, a network 12, and several electronic devices, such as a PC (Personal Computer) 13, a mobile phone 14, and the like.

[0131] The server 11 may be a physical server including an independent host, or the server 11 may be a virtual server hosted by a host cluster. During operation, the server 11 may run the server-side program of the financial service platform to realize the service end of the financial service platform. Figure 1 The data processing system shown.

[0132] PC 13 and mobile phone 14 are only some types of electronic devices that can be used by users. In fact, users can obviously also use electronic devices such as tablet devices, laptops, PDAs (Personal Digital Assistants), wearable devices (such as smart glasses, smart watches, etc.), etc., and one or more embodiments of this specification do not limit this. During operation, the electronic device can run the client-side program of the financial service platform to implement the client of the financial service platform.

[0133] Among them, the client application of the above-mentioned financial service platform can be started and run on the electronic device. The client-side program can be a native application installed on the electronic device, or the client-side program can be a small program, a quick application or other similar forms. Of course, when using web technologies such as HTML5 or similar, the relevant functions can be implemented through the page displayed by the browser. The browser here can be an independent browser application or a browser module embedded in certain applications.

[0134] As for the network 12 for interaction between electronic devices such as PC 13 and mobile phone 14 and server 11, it is possible to select a wired or wireless network to achieve communication based on the communication mode supported by the corresponding electronic device, and this specification does not limit this. For example, PC 13 can support both wired and wireless communication, so it can use a wired or wireless network to achieve communication as needed, while mobile phone 14 usually only supports wireless communication, so it can use a wireless network to achieve communication.

[0135] In one example, a user can query his / her user behavior score through a client of a financial service platform running on a mobile phone 14, and then the server 11 reads and returns the user behavior score to the user by executing the implementation method described below in this specification. Figure 7 Provide explanation.

[0136] like Figure 7 As shown, in some embodiments, the query method of user behavior points exemplified in this specification includes:

[0137] S710: Obtain user query instructions.

[0138] Combination Figure 6 As shown, the user can click the control for querying the user behavior score through the client interface of the financial service platform, so that the client can detect the user query instruction and send the user query instruction to the server. In the implementation mode of this specification, the user query instruction includes a user identifier, which can be, for example, a UID.

[0139] S720. Query the user behavior points corresponding to the current calculation period from the database according to the user identifier. In response to the user behavior points not being calculated, execute the user behavior points calculation task to obtain the user behavior points of the current calculation period. In response to the user behavior points being calculated, read the user behavior points.

[0140] In the implementation mode of this specification, after obtaining the user query instruction, the server can query the user behavior score of the current calculation cycle in the corresponding database. The calculation cycle refers to the time period for periodic calculation of user behavior points. For example, in an example, the financial service platform calculates the user behavior points once a day, and the calculation cycle is 1 day.

[0141] When a user queries the user behavior score for the day, the user behavior score for the day may have been calculated through the aforementioned batch task processing method and stored in the database. In this case, the user behavior score in the current calculation cycle can be directly read from the database, and the user behavior score is the latest user behavior score for the day.

[0142] However, at the time of the user's query, the user's user behavior score for the day may not have been calculated yet. This is because batch tasks need to be processed in sequence. If the batch task has not processed the user's task at the time of the user's query, then the database does not have the user behavior score for the day. In this case, in order to be able to respond to the user's query operation immediately, there is no need to wait for the batch task to be processed, but to call and execute the user's calculation task separately to generate the latest user behavior score. At the same time, the user's calculation task can be deleted in the batch task.

[0143] S730: Display the user behavior score of the current calculation cycle.

[0144] In either of the above cases, after obtaining the user behavior points of the user in the current calculation period, the server can return the user behavior points to the client and render them on the client interface, so that the user can view the latest user behavior points of the day.

[0145] In some embodiments, the financial service platform may further display one or more of the aforementioned user behavior parameters, that is, in the aforementioned embodiments, the user behavior parameters corresponding to each user will also be stored in the database. When the user triggers a query instruction to query a certain user behavior parameter on the client, the financial service platform can display the queried user behavior parameter to the user according to the aforementioned method process, which will not be repeated here.

[0146] From the above, it can be seen that in the implementation of this specification, the accuracy of the user behavior score is improved by combining historical behavior data and current behavior data to calculate the user behavior score. In addition, by combining batch task running with real-time calculation, when the user queries, the task processing can be prioritized for the querying user, thereby satisfying the user's real-time query. Moreover, through the aforementioned batch task scheduling method, the batch task processing efficiency is improved, the system response speed is improved, and the task requirements with high timeliness requirements are met.

[0147] In some implementations, this specification provides a batch task processing device, which can be applied to a data processing system, wherein the data processing system includes multiple computing clusters and a disaster recovery cluster corresponding to each computing cluster. Figure 8 As shown, the device comprises:

[0148] The task scheduling module 10 is configured to distribute the batch tasks to the computing clusters and their disaster recovery clusters corresponding to the identification intervals according to the identification intervals to which the identification information of each task in the batch tasks belongs, wherein each computing cluster corresponds to one identification interval, and the identification information of the multiple computing clusters corresponding to the identification intervals is a continuous value;

[0149] The task execution module 20 is configured to, for each computing cluster, read and execute a task with a first identifier from the distributed batch tasks, and a disaster recovery cluster corresponding to the computing cluster reads and executes a task with a second identifier from the distributed batch tasks, wherein one of the first identifier and the second identifier is identifier information of an odd number of bits in the identifier interval, and the other is identifier information of an even number of bits in the identifier interval;

[0150] The result storage module 30 is configured to store the task processing result of each task in the computing cluster and its disaster recovery cluster.

[0151] In some implementations, the task includes a first task and a second task different from the first task; the task execution module 20 is configured to:

[0152] In a continuous time period, the computing cluster sequentially and alternately reads and executes a first task with a first identifier and a second task with a first identifier;

[0153] In a continuous time period, the disaster recovery cluster reads and executes the first task with the second mark and the second task with the second mark alternately in sequence.

[0154] In some implementations, the task execution module 20 is configured to:

[0155] In odd time periods, the computing cluster reads and executes a first task with a first identifier; in even time periods, the computing cluster reads and executes a second task with a first identifier;

[0156] In an odd-numbered time period, the disaster recovery cluster reads and executes the first task with the second identifier; in an even-numbered time period, the disaster recovery cluster reads and executes the second task with the second identifier.

[0157] In some implementations, the task includes a user behavior score calculation task and / or a user behavior parameter calculation task, wherein the user behavior score calculation task depends on the user behavior parameters.

[0158] In some implementations, the task execution module 20 is configured to:

[0159] Acquire user behavior data, wherein the user behavior data includes historical behavior data and current behavior data, and the current behavior data includes at least part of user behavior parameters obtained in advance;

[0160] A plurality of behavior factor parameters are generated according to the historical behavior data and the current behavior data, and the user behavior score is generated according to the plurality of behavior factor parameters, and the task processing result includes the user behavior score.

[0161] In some implementations, the task execution module 20 is configured to:

[0162] The user behavior data is obtained by reading from a database corresponding to the user representation according to the user identifier;

[0163] Each user behavior score is stored in a database corresponding to the user identification.

[0164] In some embodiments, this specification provides a user behavior score query device, such as Fig. 9 As shown, the device comprises:

[0165] The instruction acquisition module 40 is configured to acquire a user query instruction, wherein the user query instruction includes a user identifier;

[0166] The points query module 50 is configured to query the user behavior points corresponding to the current calculation cycle from the database according to the user identifier, wherein the user behavior points of each calculation cycle in the database are calculated by the batch task processing method described in any of the aforementioned embodiments, and the task processing result includes the user behavior points; in response to the user behavior points corresponding to the user identifier in the current calculation cycle not being calculated, executing the user behavior points calculation task to obtain the user behavior points of the current calculation cycle; in response to the user behavior points corresponding to the user identifier in the current calculation cycle being calculated, reading the user behavior points;

[0167] The points display module 60 is configured to display the user behavior points in the current calculation cycle.

[0168] In some implementations, this specification provides an electronic device, including:

[0169] processor;

[0170] A memory stores computer instructions, wherein the computer instructions, when executed by a processor, implement the method described in any of the above embodiments.

[0171] In some implementations, the embodiments of this specification provide a computer program product, which, when executed, implements the method described in any of the aforementioned implementations.

[0172] Fig.10 It is a schematic structural diagram of an electronic device provided by an exemplary embodiment, through which the above-mentioned processor or storage medium function can be implemented.

[0173] Please refer to Fig.10 At the hardware level, the electronic device includes a processor 702, an internal bus 704, a network interface 706, a memory 708, and a non-volatile memory 710, and may also include hardware required for other functions. One or more embodiments of this specification may be implemented based on software, such as the processor 702 reading the corresponding computer program from the non-volatile memory 710 into the memory 708 and then running it. Of course, in addition to the software implementation, one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

Claims

1. A batch task processing method, characterized in that: Applied to a data processing system, the data processing system includes multiple computing clusters and a disaster recovery cluster corresponding to each computing cluster, the method includes: Distribute the batch tasks to the computing clusters and their disaster recovery clusters corresponding to the identification intervals according to the identification intervals to which the identification information of each task in the batch tasks belongs, wherein each computing cluster corresponds to one identification interval, and the identification information of the multiple computing clusters corresponding to the identification intervals is a continuous value; For each computing cluster, the computing cluster reads and executes a task with a first identifier from the distributed batch tasks, and the disaster recovery cluster corresponding to the computing cluster reads and executes a task with a second identifier from the distributed batch tasks, wherein one of the first identifier and the second identifier is identifier information of an odd number of bits in the identifier interval, and the other is identifier information of an even number of bits in the identifier interval; The task processing results of each task are stored in the computing cluster and its disaster recovery cluster.

2. The method according to claim 1, characterized in that The tasks include a first task and a second task different from the first task; The process of the computing cluster reading and executing the task of the first identification includes: in a continuous time period, the computing cluster sequentially and alternately reads and executes the first task of the first identification and the second task of the first identification; The process of the disaster recovery cluster reading and executing the task with the second mark includes: in a continuous time period, the disaster recovery cluster sequentially and alternately reads and executes the first task with the second mark and the second task with the second mark.

3. The method according to claim 2, characterized in that The computing cluster sequentially and alternately reads and executes the first task of the first identifier and the second task of the first identifier in the continuous time period, including: in the odd time period, the computing cluster reads and executes the first task of the first identifier; in the even time period, the computing cluster reads and executes the second task of the first identifier; The method comprises: in the continuous time period, the disaster recovery cluster reads and executes the first task with the second identifier and the second task with the second identifier alternately in sequence, including: in the odd time period, the disaster recovery cluster reads and executes the first task with the second identifier; in the even time period, the disaster recovery cluster reads and executes the second task with the second identifier.

4. The method according to any one of claims 1 to 3, characterized in that: The task includes a user behavior score calculation task and / or a user behavior parameter calculation task, wherein the user behavior score calculation task depends on the user behavior parameters.

5. The method according to claim 4, characterized in that In the case where the task includes a user behavior score calculation task, the process of executing the user behavior score calculation task includes: Acquire user behavior data, wherein the user behavior data includes historical behavior data and current behavior data, and the current behavior data includes at least part of user behavior parameters obtained in advance; A plurality of behavior factor parameters are generated according to the historical behavior data and the current behavior data, and the user behavior score is generated according to the plurality of behavior factor parameters, and the task processing result includes the user behavior score.

6. The method according to claim 5, characterized in that The acquiring of user behavior data comprises: reading the user behavior data from a database corresponding to the user representation according to the user identifier; The storing of the task processing result of each task in the computing cluster and its disaster recovery cluster includes: storing each user behavior score in a database corresponding to the user identifier.

7. A method for querying user behavior points, characterized in that: include: Acquire a user query instruction, wherein the user query instruction includes a user identifier; According to the user identifier, query the user behavior score corresponding to the current computing cycle from the database, the user behavior score of each computing cycle in the database is calculated by the batch task processing method according to any one of claims 1 to 6, and the task processing result includes the user behavior score; In response to the user behavior score corresponding to the user identifier in the current calculation cycle not being calculated, executing the user behavior score calculation task to obtain the user behavior score in the current calculation cycle; In response to the user behavior score corresponding to the user representation in the current calculation cycle having been calculated, reading the user behavior score; Display the user behavior score for the current calculation cycle.

8. A batch task processing device, characterized in that: Applied to a data processing system, the data processing system includes multiple computing clusters and a disaster recovery cluster corresponding to each computing cluster, the device includes: The task scheduling module is configured to distribute the batch tasks to the computing clusters and their disaster recovery clusters corresponding to the identification intervals according to the identification intervals to which the identification information of each task in the batch tasks belongs, wherein each computing cluster corresponds to one identification interval, and the identification information of the multiple computing clusters corresponding to the identification intervals is a continuous value; The task execution module is configured to, for each computing cluster, read and execute a task with a first identifier from the distributed batch tasks, and a disaster recovery cluster corresponding to the computing cluster reads and executes a task with a second identifier from the distributed batch tasks, wherein one of the first identifier and the second identifier is identifier information of an odd number of bits in an identifier interval, and the other is identifier information of an even number of bits in an identifier interval; The result storage module is configured to store the task processing result of each task in the computing cluster and its disaster recovery cluster.

9. A user behavior score query device, characterized in that: include: An instruction acquisition module is configured to acquire a user query instruction, wherein the user query instruction includes a user identifier; a points query module, configured to query a user behavior score corresponding to a current computing cycle from a database according to the user identifier, wherein the user behavior score of each computing cycle in the database is calculated by the batch task processing method according to any one of claims 1 to 6, and the task processing result includes the user behavior score; In response to the user behavior score corresponding to the user identifier in the current calculation cycle not being calculated, executing the user behavior score calculation task to obtain the user behavior score in the current calculation cycle; In response to the user behavior score corresponding to the user representation in the current calculation cycle having been calculated, reading the user behavior score; The points display module is configured to display the user behavior points in the current calculation cycle.

10. A computer program product, characterized in that When the computer program product is executed, the method according to any one of claims 1 to 7 is implemented.

Images