Streaming data service time calculation method suitable for security transaction monitoring service

Abstract

The invention relates to the technical field of data processing, in particular to a streaming data business time calculation method suitable for security transaction supervision business, which comprises the following steps: S1, business data persistence: subscribing a real-time transaction data theme of an upstream transaction platform, docking each channel data of the theme and performing batch persistence, recording the latest data time of each channel; s2, business time calculation: periodically performing time calculation on the basis of the information record after data persistence, and storing periodical calculation details; and S3, the monitoring platform triggers the execution of the monitoring instance according to the latest theme service time. The method has the advantages that data omission is avoided; missing calculation of a monitoring algorithm due to the fact that certain theme data is not ready in a single theme dependence mode is avoided; a 24-hour service time calculation service is provided; the data persistence condition is monitored in real time, and the business time is accelerated.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a method for calculating streaming data service time applicable to securities trading monitoring services. Background Art

[0002] The securities trading monitoring algorithm (hereinafter referred to as securities monitoring), as the core tool for securities market supervision and control, plays a vital role in ensuring the fairness and stability of the market. It is mainly responsible for monitoring the trading behavior of the entire market, accounts, securities and cross-products. These monitoring activities involve multiple securities businesses (such as stocks, options, bonds, etc.) in multiple securities markets. Each type of business has different subject data (such as orders, transactions, market conditions, etc., hereinafter referred to as business subject data), and the data between each business subject are independent of each other.

[0003] The business subject data of securities monitoring comes from multiple real-time securities trading platforms. It has the characteristics of streaming data order, large volume, fast and continuous arrival. After cleaning, sorting and other processing, it forms unique and ordered data, which is persisted to the monitoring platform and provided to the monitoring algorithm. The real-time transaction data of each business subject carries information such as products, accounts, transaction volume, and transaction time. It is divided into multiple independent data areas by product and transmitted downstream, namely data channels {Di|i∈[1, n]}. The monitoring platform connects to the data channels of all upstream subjects, sorts them by time and transaction number, and ensures that the data of each channel is persisted in a strictly orderly manner.

[0004] Securities monitoring has the characteristics of timeliness and periodic execution. Securities monitoring combines periodic natural time to form securities monitoring instances (hereinafter referred to as monitoring instances). The monitoring platform schedules monitoring instances to execute in a specific period, that is, the algorithm is driven by natural time to run, calculate the data in the specified natural time range, and require the analysis results to be obtained as soon as possible. Securities monitoring can be divided into product monitoring and account monitoring from the perspective of monitoring objects. Product monitoring analyzes product transaction data {SECi|i∈[1, n]} distributed in multiple business subject channels. Since the products between channels are independent, only single channel data can be monitored, that is, the execution is triggered immediately when the natural time reaches the monitoring running time. Monitoring is only for single channel data Msec=Monitor(Di), where i∈[1, n]. Account monitoring analyzes accounts across multiple channel products, and requires the combination of account transaction data {ACTi|i∈[1, n]} of multiple channels. Monitoring needs to obtain all channel data to form a data set Macc={Monitor_ALL(Di) | i∈[1, n]}, so it is necessary to ensure that the data of each channel is ready at the monitoring running time.

[0005] Since business data is distributed in different channels and each channel has differences in data volume, transmission delay and persistence delay, it is difficult for the data of securities account monitoring at runtime to be fully ready at the same natural time point, affecting the accuracy of account monitoring results.

[0006] To solve this problem, securities monitoring needs to introduce business time applicable to securities streaming business data, that is, to calculate the latest data time that is ready according to all channels of the business subject, to ensure that the data that the monitoring depends on at all times is ready. The scheduling of securities monitoring instances depends on the ready status of multiple channel data. Once the data is ready, the operation of the monitoring instance will be triggered. The execution of the monitoring instance is directly linked to the business time, so an accurate business time calculation method is required to ensure the accuracy and timeliness of the algorithm.

[0007] In order to ensure that the monitoring instance can process all the ready securities trading data when it is triggered, the business time must be calculated according to the readiness of the current data. In this way, the monitoring instance can be triggered at the right time, thus ensuring the accuracy of its results.

[0008] Existing securities monitoring uses the latest time of transaction and declaration data of business subjects as business time to trigger the execution of monitoring instances. Due to the differences in data activity and persistence time of each channel of business subjects, there may be a mismatch between business time and data persistence. On the other hand, as the monitoring business develops, the monitoring algorithm needs more real-time data sources, and the business time of a single subject can no longer meet the monitoring requirements. Summary of the invention

[0009] The purpose of the present invention is to overcome the shortcomings of the prior art, solve the problem that the existing business time calculation method leads to inaccurate monitoring algorithm results, and provide a streaming data business time calculation method suitable for securities trading monitoring business, which is used to improve the accuracy of the monitoring algorithm by coordinating with the monitoring instance scheduling of the securities monitoring platform.

[0010] In order to achieve the above purpose, a streaming data business time calculation method suitable for securities trading monitoring business is designed, and the method steps include: S1 business data persistence, subscribing to the real-time trading data topic of the upstream trading platform, connecting to the channel data of the topic and persisting it in batches, and recording the latest data time of each channel; S2 business time calculation, the business time calculation is based on the information record after the data is persisted, periodically calculates to obtain the business time, and stores the business time details, the business time calculation method includes: S2.1 single-topic business time calculation, business data is persisted in batches, and the persistence time, topic, source channel and other information of each batch are recorded in the storage area. Through the business time calculation method, the latest business time of the topic channel is obtained, and the inactive data channels are filtered, and the latest time when the topic data is ready is obtained, that is, the business time of the topic, and the data is persisted; S2.2 multi-topic business time calculation, based on the persistence data of the single-topic business time, the business time when multiple topic data are ready is calculated, that is, the multi-topic business time, and the topic business time data is persisted; S3 The monitoring platform triggers the execution of the monitoring instance according to the latest topic business time.

[0011] The streaming data business time calculation method applicable to securities trading monitoring business provided by the present invention has more technical features, among which step S1 is specifically as follows: the securities monitoring platform connects to the business subject data channel of the upstream trading platform and processes each channel data independently; the business subject data is persisted in batches, and the time and subject channel information of each batch are recorded for business time calculation, and the specific steps include: S1.1 constructing a business data persistence batch storage block, the batch storage block is used to record the details of each persistent batch, the batch storage block includes the batch occurrence time T, channel ID, business subject, batch number; S1.2 the securities monitoring platform receives the channel data of the business subject of the upstream trading platform, the data includes the data occurrence time t, serial number, source subject channel, transaction related information, and sorts the data according to the channel, serial number, and time field; S1.3 the ordered data is divided into multiple batches according to a fixed number of items, recorded as a data set Dataset, each batch of data is persisted, and the transaction occurrence time of each batch of data is recorded to form a set recorded as D = {ti | i∈Dataset}, traverse the set D to obtain the minimum transaction occurrence time T=min(ti) of this batch; S1.4 Business subject data persistence At the same time, the minimum data occurrence time T, channel ID, business subject, and batch number information of each batch are persisted to the batch storage block.

[0012] The streaming data business time calculation method applicable to securities trading monitoring business provided by the present invention has more technical features, wherein the step S2.1 single-topic business time calculation is specifically as follows: S2.1.1 construct a subject business time storage block to record the latest time details of the business subject, including the business subject and the latest business time Tp; S2.1.2 poll and read the batch storage block at a fixed frequency, obtain the business time T of n channels under the business subject P from the storage block, and form a business time set recorded as DT = { Ti | i∈[1, n]}, which indicates that the business data of the corresponding channel i before Ti has been persisted and provides services to the outside world; S2.1.3 obtain the latest business time of all channels, calculate the latest business time set of each channel of the subject from the business time set DT, and record it as Tcur = { max(Ti) | i∈[1, n]}, calculate the lag time difference between the latest business time of each channel in this round and the latest business time Tpre in the previous round, and form a set recorded as { max(Ti) - Tpre| i∈[1, n]}}; S2.1.4 Eliminate inactive subject channels, and the lag time difference threshold is recorded as S1. If max(Ti)-Tpre>= S1, the business time Ti of channel i is not included in this round of calculation, and is recorded as the excluded channel set {j | j∈(channels not included in the calculation)}. The subject channel set CT for this round of business time calculation is CT = { max(Ti) | i∈[1, n] ∩ i≠j}; S2.1.5 Get the latest time of the ready data, and take the minimum value from the set CT participating in the business time calculation in this round, that is, take the minimum time from the latest time set of all channels, and record it as the quasi-business time. This time indicates that all channel data of the subject at the current moment have completed persistence, and is recorded as Tnew=min[max(Ti)], where i∈[1, n] and i ≠ j ; S2.1.6 Solve the problem of data persistence delay in the same batch. Since the subject channel data is persisted in batches, data with the same data occurrence time may be persisted in multiple batches, that is, the proposed business time of the subject channel reaches Tnew, but the data at that time point is divided into multiple batches and is not fully persisted at this time, resulting in omission of data dependent on the monitoring instance. Subtract the specific persistence delay time c from the proposed business time Tnew of the subject channel to avoid the problem of channel data persistence in batches. The single subject business time Tp =Tnew-1 =min[max(Ti)]-c, where i∈[1, n] and i ≠ j and c is a specified constant; S2.1.7 The single subject business time Tp is persisted to the business time storage block, and the Tpre value is updated.

[0013] The streaming data business time calculation method suitable for securities trading monitoring business provided by the present invention has more technical features, among which the step S2.2 multi-topic business time calculation is specifically as follows: S2.2.1 polling and reading the business time storage block at a fixed frequency, the number of topics is m, and the business time Tp of all single business topics is obtained to form a set MT = {Tpm |pm∈[1, m]}; S2.2.2 according to a specific configuration, 2 or more topics are combined to form multi-topic combination configuration information, and the minimum value set of multiple topic business times is taken out from the business time set MT, recorded as Tm= {min(Tpmi,Tpmj)|i, j∈[1, n] and i ≠ j}; S2.2.3 the multi-topic business time Tm is persisted to the business time storage block.

[0014] The streaming data business time calculation method suitable for securities trading monitoring business provided by the present invention has more technical features, among which step S3 is specifically as follows: a. During non-trading periods, the server system time deployed by the time calculation service is taken as the business time driver, and the server needs to enable the clock synchronization mechanism; b. During the trading preparation period, the securities market time is taken as the business time driver. There is no securities trading during this period, and the upstream system still needs to send the latest market snapshot on time; c. During the trading period, the latest time of various securities themes is taken as the business time.

[0015] Compared with the prior art, the present invention has the following advantages: a. The subject business time calculation method of the present invention can accurately obtain the business time of the ready data of each business subject, and triggering the monitoring algorithm can avoid data omission.

[0016] b. The subject business time of the present invention supports the business time calculation of multiple subject business data combinations, avoiding the situation in which a certain subject data is not ready in a single subject dependency mode, resulting in omissions in the monitoring algorithm.

[0017] c. The time calculation method of the present invention provides different calculation methods for different time periods according to the characteristics of securities trading running throughout the day, and provides business time calculation services for all time periods.

[0018] d. The present invention designs a subject data persistence batch table and a business timetable, which can monitor the data persistence status and business time advancement progress in real time. DETAILED DESCRIPTION

[0019] In order to make the purpose, principle and structure of the present invention clearer, it is further described below in conjunction with specific embodiments.

[0020] The existing securities business monitoring platform receives data from the upstream trading platform and calculates the latest business time based on the persisted stock transaction and declaration data to trigger task execution. The execution process is as follows: a. The securities business monitoring platform receives stock transaction and declaration records from the upstream trading platform according to the transaction set file. Each record has a corresponding data time Ti. The data of each transaction set is sorted separately, and the data in each transaction set is strictly ordered; b. The securities business monitoring platform traverses each transaction set file Fj, and collects the data to be persisted into a batch data(Fj) with a fixed number of entries each time. Data from different sets and different times may be merged into one persistent batch; c. After each batch of data is persisted, the latest business time max(Ti) corresponding to the batch of data data(Fj) is taken and sent to the scheduling time queue; d. The monitoring algorithm scheduling system consumes the time queue to obtain the latest time T of the persisted data and obtains the monitoring algorithm cycle instance that can be executed at that time; e. Pass the latest time as a parameter to the monitoring algorithm cycle instance, limit the time interval for the algorithm instance calculation, and trigger the instance execution.

[0021] During the execution of the above-mentioned securities business monitoring platform of the prior art, due to different activity levels, the number of records of each transaction set sent by the upstream platform is not exactly the same. The latest business time taken after each batch is persisted is not completely consistent with the data time of each transaction set that has been persisted. It may happen that the monitoring algorithm instance has been running according to the latest business time, but the data of some sets at that time is not completely persisted, or the activity of some sets is low but the time is uniformly pushed forward. And when there are many records of a certain set at the same time, it is divided into multiple persistent batches. It may happen that the corresponding data time of the batch has been sent to the scheduling time queue, but the records of the same time are not completely persisted, resulting in inaccurate calculation.

[0022] On the other hand, with the development of monitoring business, monitoring algorithms not only rely on stock transactions and declarations, but also rely on other real-time transaction-related data, such as options and bond transactions. Taking only single business type data as the monitoring algorithm scheduling data time may cause other types of data to not be fully persisted and inaccurate calculations.

[0023] Detailed description of the technical solution of this application The purpose of the present invention is to solve the problem that the existing business time calculation method leads to inaccurate monitoring algorithm results, and to provide a business time calculation method suitable for transaction monitoring business monitoring algorithm, which improves the accuracy of the monitoring algorithm by cooperating with the instance scheduling system.

[0024] To achieve the above purpose, a method for calculating business time is designed. Through the latest data time, channel and other information persisted in batches by each business topic, the latest time when the data of each business topic is ready is calculated, thereby triggering the execution of the monitoring algorithm. This method mainly consists of two parts: business topic data persistence channel record and business time calculation method.

[0025] 1. Business topic persistent channel records The data of each business topic is stored in the message queue. The data in the topic is divided into multiple channels according to the specified fields (such as securities product codes). The message queue data is consumed by the persistence program in batches to the database for use by the monitoring algorithm, and the data time, channel and other information of each batch of data are recorded for subsequent business time calculation. The specific process is as follows: A1: Initialize the business data batch table, which includes the batch data time T, channel number n, business topic, batch number, and records the detailed information of each persistent batch.

[0026] A2: The securities business monitoring platform receives real-time data from the upstream trading platform according to the business topic. Each piece of data is sorted according to the serial number and time field and then appended to the message queue. The data is stored in 1 to n channels modulo the securities code.

[0027] A3: Consume the topic data of the message queue in micro-batches with a fixed number of records. The time set of each batch of data is recorded as {ti |i∈[1, records]}. Get the minimum data time T=min(ti) corresponding to each channel n in each batch.

[0028] A4: After data persistence, the minimum data time, channel, batch number and other information of each batch are stored in the business data batch table.

[0029] 2. Business time calculation method Business time calculation is based on the information record after data persistence, periodically calculates time, and stores periodic calculation details. Business time calculation consists of single-topic business time and multi-topic business time, which are used for periodic instance triggering of single-topic dependency and multi-topic dependency of the monitoring algorithm of the securities business monitoring platform, respectively. Multi-topic business time is calculated based on single-topic business time.

[0030] (1) Calculation method for single-topic business time The securities business monitoring platform obtains the latest business time of the subject channel based on the persistent business data batch table through the business time calculation method, filters the inactive channels, obtains the latest time when the subject data is ready, and records it in the business subject timetable. The specific process is as follows: B1: Initialize the business schedule table, which contains fields such as business theme and latest business time, and is used to record the latest time after calculation for each business theme; Initialize the latest business time Tpre = 0 after calculation for each theme, which is updated after each round of calculation and used as the time calculation benchmark for the next round.

[0031] B2: The securities business monitoring platform reads the business data batch table at a fixed frequency. In each round, it obtains the set of business times for each channel of business theme P by theme and channel, denoted as {Ti | i ∈ [1, n]}. This time indicates that the data before Ti in the corresponding channel i has been persisted and is ready.

[0032] B3: Obtain the latest time for each channel. From the set of business times, obtain the latest business time for each channel of the theme, denoted as {max(Ti) | i ∈ [1, n]}.

[0033] B4: Calculate the lag time difference between the business time of each channel in this round and the latest business time Tpre in the previous round, denoted as {max(Ti) - Tpre | i ∈ [1, n]}.

[0034] B4: Exclude inactive channels. The lag time difference threshold is denoted as S1. If max(Ti) - Tpre >= S1, the channel business time Ti is not included in this time calculation and is recorded in {j | j ∈ (channels not included in the calculation)}; If max(Ti) - Tpre < S1, the channel business time Ti is included in this time calculation and is excluded from the set j. The set of channels participating in the business time calculation in this round is {max(Ti) | i ∈ [1, n] ∩ i ≠ j}.

[0035] B5: Obtain the latest time of the ready data. Take the minimum value from the set of channels participating in the business time calculation in this round, that is, take the minimum time among the latest channel times, indicating that all data of all channels before this time has been persisted and is ready, denoted as Tnew = min[max(Ti)], where i ∈ [1, n] and i ≠ j.

[0036] B6: Solve the problem of data persistence delay for the same batch. Data with the same business time may be written into channels in multiple batches of fixed numbers. That is, when the latest business time in the channel reaches Tnew, the data at this time point has not been fully persisted, resulting in data miscalculation. Subtract 1 ms from the Tnew time to represent the latest ready business time, ensuring that all data before this time is ready. The final single - theme business time is denoted as Tp = Tnew - 1 = min[max(Ti)] - 1, where i ∈ [1, n] and i ≠ j.

[0037] B7: The single-topic business time Tp is stored in the business time table, and the Tpre value is updated.

[0038] (2) Calculation method for multi-topic business time Based on the single-topic business time, calculate the business time when multiple topic data are ready, and obtain the multi-topic business time. The specific process is as follows: C1: Initialize the multi-theme combination configuration table, which is used to combine multiple single-theme times to form multi-theme times.

[0039] C2: The securities business monitoring platform reads the business schedule at a fixed frequency and obtains the business time set of all single business topics P in each round, which is recorded as {Tm | m∈[1, n]}.

[0040] C3: Obtain multiple topics to be merged and calculated according to the multi-topic combination configuration table, take out the topic time to be merged and calculated from the business time set, and take the minimum value of multiple topics, recorded as Tm'=min(Tmi, Tmj), where i, j∈[1,n] and i ≠ j.

[0041] C4: Multi-topic business time Tm' is stored in the business timetable.

[0042] The securities monitoring algorithm has the characteristics of running during designated periods throughout the day, and there is still a need for operation during non-trading periods. Due to the characteristics of the securities trading data time period, the non-trading period requires special time point processing, as follows: D1: During non-trading hours [0:00, 8:30) and (15:30, 24:00], the server system time deployed by the time calculation service is used as the business time driver. The server needs to enable the clock synchronization mechanism.

[0043] D2: The trading preparation period [8:30, 9:15] and lunch break period (11:30, 13:00) use the securities market time as the business time driver. There is no securities trading during this period, but the upstream system still needs to send the latest market snapshot on time.

[0044] D3: Trading hours [9:15-11:30] and [13:00-15:30] take the latest time of various securities topics as business hours.

[0045] <Example 1> The implementation of a securities business monitoring platform is used as a patent example description, and the following involves table structure, business time algorithm implementation and practical application.

[0046] 1. Database storage design (1) Business time batch data table (hereinafter referred to as detailed time table) This table is used to store the business time corresponding to each batch of business data persistence. Business data and business time are stored separately. Business data in the securities trading stage is persisted in real time in micro-batches. After each batch is persisted, the latest business time of this batch is written into the time batch table. The data volume of this table is related to the number of transaction reports, with an average of 1 million records per day.

[0047] Table 1 Detailed timetable (2) Schedule of each theme promotion time (hereinafter referred to as the theme schedule) This table is used to store the latest business time of each subject after processing. The time calculation service calculates the latest business time of each subject from the detailed timetable by subject channel at a frequency of 5 seconds / time, and stores it in the subject timetable, retaining the whole day's history, with an average of 100,000 records per day.

[0048] Table 2 Theme timetable Field Name type describe data_name Varchar(40) Business data type, corresponding to the data_name field of the detailed timetable event_time Numeric(16) The latest business time for each business type, which is aggregated from the detailed timetable by partition_num and event_time 2. Middleware Kafka storage design Each business topic is stored independently, corresponding to the storage topic of Kafka. Kafka stores the detailed data of each business topic in an appending manner. The persistence program consumes data from Kafka in real time and writes it to the database in micro-batches. Each batch will record the latest persistence time and store it in the detailed time table.

[0049] Each business theme is stored by channel, and the channel uses securities products as the division field. The key business theme information is listed below.

[0050] 3. Business batch persistence program The program is a distributed cloud deployment architecture. It allocates a corresponding persistence processing thread to each Kafka topic, consumes business data from the corresponding Kafka topic in real time, pulls about 20,000 detailed business data each time, assembles them into a persistent batch package, and generates a latest business time record at the same time, including data_source, data_name, partition_num, event_time and other fields. After this batch is persisted normally, the latest business time record is written into the detailed time table.

[0051] 4. Time calculation program The program is a distributed stateless cloud service. Among multiple distributed program instances, only one program instance performs time calculation tasks at each processing moment. The program obtains the latest business time of each channel from the detailed timetable by subject at a frequency of 5 seconds per time, calculates the latest business time of each subject according to the single subject business time acquisition method, and inserts it into the subject timetable in an appended manner.

[0052] 5. Business time application Scheduling monitoring algorithm tasks that rely on single business topics or multiple business topics. During the securities trading stage, the original securities business monitoring platform had monitoring algorithm tasks that relied on multiple business topics such as transaction declarations and market conditions. When scheduling such tasks, the transaction declaration topic data had been persisted, and other data such as market conditions were not persisted in time due to upstream delays, resulting in inaccurate data in the task calculation.

[0053] The above-mentioned multi-theme business time - transaction declaration and market conditions are now applied to the monitoring algorithm task. When the business times of multiple business themes reach the Tp moment, that is, all the subject business data before the Tp moment is ready, the system schedules the tasks that have not been started at the Tp moment and before, to ensure that all data before the Tp moment are accurately calculated.

[0054] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent substitutions or changes according to the technical solutions and novel concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for calculating streaming data service time applicable to securities trading supervision services, characterized in that: The method steps include: S1 business data persistence, subscribe to the real-time transaction data topic of the upstream trading platform, connect to the channel data of the topic and persist them in batches, and record the latest data time of each channel; S2 business time calculation, business time calculation is based on the information record after data persistence, periodically calculates to obtain business time, and stores business time details. The business time calculation method includes: S2.1 Single-topic business time calculation. Business data is persisted in batches. The persistence time, topic, source channel and other information of each batch are recorded in the storage area. The latest business time of the topic channel is obtained through the business time calculation method, and the inactive data channels are filtered. The latest time when the topic data is ready is obtained, that is, the business time of the topic, and the data is persisted; S2.2 Multi-topic business time calculation: Based on the persistent data of single-topic business time, the business time for which multiple topic data are ready is calculated, i.e., multi-topic business time, and the topic business time data is persisted; The S3 monitoring platform triggers the execution of monitoring instances based on the latest subject business time.

2. A method for calculating streaming data service time applicable to securities trading supervision business as claimed in claim 1, characterized in that: The step S1 is specifically as follows: the securities monitoring platform connects to the business subject data channel of the upstream trading platform, and processes each channel data independently; the business subject data is persisted in batches, and the time and subject channel information of each batch are recorded for business time calculation. The specific steps include: S1.1 build a persistent batch storage block for business data, which is used to record the details of each persistent batch. The batch storage block includes the batch occurrence time T, channel ID, business subject, and batch number; S1.2 The securities monitoring platform receives channel data of business topics of the upstream trading platform, which includes the data occurrence time t, sequence number, source subject channel, and transaction related information, and sorts the data by channel, sequence number, and time field; S1.3 divides the ordered data into multiple batches according to a fixed number of entries, recorded as a dataset Dataset, persists each batch of data, and records the transaction time of each batch of data to form a set recorded as D = { ti | i∈Dataset}, traverses the set D to obtain the minimum transaction time T=min(ti) of this batch; S1.4 While persisting business subject data, the minimum data occurrence time T, channel ID, business subject, and batch number information of each batch are persisted to the batch storage block.

3. A method for calculating streaming data service time applicable to securities trading supervision business as claimed in claim 1, characterized in that: The specific calculation of the single-topic service time in step S2.1 is as follows: S2.1.1 Construct a subject business time storage block to record the latest time details of the business subject, including the business subject and the latest business time Tp; S2.1.2 polls and reads the batch storage block at a fixed frequency, obtains the business time T of n channels under the business topic P from the storage block, and forms a business time set recorded as DT = { Ti | i∈[1, n]}, which indicates that the business data of the corresponding channel i before Ti has been persisted and provides services to the outside world; S2.1.3 Obtain the latest business time of all channels, calculate the latest business time set of each channel of the subject from the business time set DT and record it as Tcur = { max(Ti) | i∈[1, n]}, calculate the lag time difference between the latest business time of each channel in this round and the latest business time Tpre in the previous round, and form a set recorded as { max(Ti) - Tpre | i∈[1, n]}; S2.1.4 Eliminate inactive subject channels, and the lag time difference threshold is recorded as S1. If max(Ti)-Tpre>= S1, the service time Ti of channel i is not included in this round of calculation, and is recorded as the excluded channel set {j | j∈(channel not included in the calculation)}. The subject channel set CT for this round of service time calculation is CT = { max(Ti) | i∈[1, n] ∩ i≠j}; S2.1.5 Obtain the latest time of the ready data, take the minimum value from the set CT participating in the business time calculation of this round, that is, take the minimum time from the latest time set of all channels, record it as the quasi-business time, which means that all channel data of the subject at the current moment have completed persistence, record it as Tnew=min[max(Ti)], where i∈[1, n] and i ≠ j; S2.1.6 solves the problem of data persistence delay in the same batch. Since the subject channel data is persisted in batches, data with the same data occurrence time may be divided into multiple batches for persistence. That is, the proposed business time of the subject channel reaches Tnew, but the data at that time point is divided into multiple batches. At this time, it is not fully persisted, resulting in the omission of data that the monitoring instance depends on; Subtract the specific persistence delay time c from the proposed business time Tnew of the subject channel to avoid the problem of batch persistence of channel data. The single subject business time Tp = Tnew-1 =min[max(Ti)] -c, where i∈[1, n] and i ≠ j and c is a specified constant; S2.1.7 The single-topic business time Tp is persisted to the business time storage block, and the Tpre value is updated.

4. A method for calculating streaming data service time applicable to securities trading supervision business as claimed in claim 1, characterized in that: The calculation of multi-topic business time in step S2.2 is as follows: S2.2.1 poll and read the business time storage block at a fixed frequency, the number of topics is m, and obtain the business time Tp of all single business topics to form a set MT = {Tpm | pm∈[1, m]}; S2.2.2 Combine two or more topics according to the specific configuration to form multi-topic combination configuration information, and extract the minimum value set of multiple topic business times from the business time set MT, denoted as Tm= {min(Tpmi, Tpmj) | i, j∈[1, n] and i ≠ j}; S2.2.3 Multi-topic business time Tm is persisted to the business time storage block.

5. A method for calculating streaming data service time applicable to securities trading supervision business as claimed in claim 1, characterized in that: The step S3 is specifically as follows: a. During non-trading periods, the system time of the server where the time calculation service is deployed is used as the business time driver, and the clock synchronization mechanism needs to be enabled on the server; b. During the transaction preparation period, the securities market time is used as the business time driver. During this period, there is no securities trading, but the upstream system still needs to send the latest market snapshot on time; c. During the trading period, the latest time of each securities theme is taken as the business time.

Images