RPO tuning strategies, computer systems and devices

Abstract

The present invention relates to an RPO tuning strategy, a computer system and a device, wherein the strategy includes: S1, periodically obtaining local storage parameter data; S2, calculating a replication queue time reference system table, calculating the running time of the replication queue family and the change in the capacity of the queue family in the previous replication cycle; S3, forming a startup sequence and a theoretical operation schedule for the next replication cycle; S4, calculating RPO-related data, and deciding whether to start the optimization again until the ideal RPO index is reached. The present invention directly relies on local storage parameters and realizes the optimization of the replication queue through a reverse engineering method. It is easy to implement in engineering, and effectively shields the time sensitivity of the storage parameters, has little impact on storage performance, and has the characteristics of high confidence and high sensitivity in optimizing the storage RPO index.

Description

Technical Field

[0001] The present invention relates to the field of computer data disaster recovery and security, and in particular to an RPO tuning strategy, a computer system and a device. Background Art

[0002] The vigorous development of cutting-edge technologies such as cloud computing, big data, and artificial intelligence has brought many conveniences to human life and production, but information systems are also facing more and more operational security risks. How to ensure the information security and business continuity of an organization and enable the organization to calmly deal with various disasters is a common challenge faced by most companies. Data replication technology plays a vital role in achieving the organization's business continuity goals and disaster recovery. Therefore, it is of great significance to study and explore data replication technology and summarize various practical and effective data replication efficiency improvement solutions.

[0003] In the field of data replication technology, the RPO (recovery point objective) indicator is a key indicator for measuring the data protection capabilities of a replication system. The RPO indicator directly determines the actual increment of data loss that an organization can tolerate. The data increment is equal to the amount of data that needs to be repaired or supplemented from the backup data copy when the business needs to be restored. RPO is so important that compressing and reducing the RPO indicator has become the most important task in the field of information system disaster recovery technology. Generally, the shorter the replication cycle, the smaller the RPO indicator and the better the data continuity.

[0004] In order to prevent data loss caused by human, natural, or technical reasons, organizations often separate active data from silent data (data copies), store them in different locations, and connect the two copies of data through operator lines. Due to the physical distance between active data and silent data, there is a certain time difference in the amount of data between active data and silent data. This data replication method is called asynchronous replication. Asynchronous replication often uses remote replication based on logical volumes, and usually there is more than one replication queue. When there is more than one replication queue, it becomes very difficult to adjust the queue and compress the replication cycle. Therefore, actively looking for ways to compress the running time of the replication queue has become a practical technical challenge faced by organizations.

[0005] Chinese patent CN111258816A discloses "RPO adjustment method, device and computer-readable storage medium". The method includes: obtaining the storage system load parameters of the terminal; calculating a new RPO value according to the storage system load parameters, and when it is judged that the storage system is under high load, the new RPO value is greater than the original PRO value; determining a new PRO position corresponding to the new PRO value, and setting the new PRO position.

[0006] Since the above disclosed method obtains the storage parameters such as CPU, memory, network bandwidth, etc. stored in the terminal, it is easy to understand that these parameters have great uncertainty, and the accuracy of its fitting scheme is necessarily low, and the operability is very low; in order to improve the optimization accuracy, the above invention disclosure must adopt the method of increasing the sampling frequency and improving the calculation efficiency, and compressing the RPO calculation step, but because the randomness of CPU, memory, and network bandwidth is too large, even if the sampling is increased to the second level, the storage performance cannot be accurately fitted. Obviously, increasing the sampling frequency will inevitably consume the storage performance. Therefore, the above method has certain limitations. Summary of the invention

[0007] To solve the existing technical problems, the present invention provides an RPO tuning strategy, a computer system and a device.

[0008] The specific content of the present invention is as follows: An RPO tuning strategy comprises the following steps:

[0009] S1, periodically obtain local storage parameter data;

[0010] S2, calculate the replication queue time reference system table, and calculate the running time and capacity changes of the replication queue family in the last replication cycle;

[0011] S3, forms the startup sequence and operation theoretical schedule for the next replication cycle;

[0012] S4, calculates RPO related data and decides whether to start optimization again until the ideal RPO indicator is reached.

[0013] Furthermore, in S1, the parameter data includes a locally stored replication queue ID, a locally stored local time, a start time, an end time, and an amount of data to be copied in the queue.

[0014] Furthermore, in S2, the time step and start time in the reference table are calculated by the number of queues; the start and end time of the replication queue family and the data increment in the previous cycle are calculated, the theoretical running time of each queue in the replication queue family is calculated, and the start and end time are calculated to form the RPO measurement value of the previous cycle.

[0015] Calculate the RPO theoretical value, RPO deviation value, and RPO deviation of the replication queue in the previous cycle, and refer to the RPO deviation to decide whether to restart the optimization.

[0016] Furthermore, in S3, the start and end time of the next cycle replication queue family and the data increment are tracked, and the running time of the queue family and the start and end time are repeatedly calculated to form the RPO measurement value of the next cycle.

[0017] Further, after obtaining the locally stored parameter data, the cumulative distribution function is calculated according to the probability density function based on the number of queues n, and the queue reference system table is calculated;

[0018] In the replication cycle, the replication queue family contains n replication queues, and the data increment generated by the replication of the i-th replication queue is m i , the amount of data to be copied ΔM T , the start time of queue i is t b , the end time is t e , running time t i , the next start time is t i + RPO measurement, the theoretical running time of the i-th replication queue is t i =m i / BW T , B.W. T The theoretical value of RPO is a fixed bandwidth between replication teams. RPO is measured as t e -t b , the deviation value of RPO is (t e -t b )-T n , RPO deviation is [(t e -t b )-T n ] / T n ;

[0019] The difference between the actual running time of the queue and the interval of the reference queue table is subtracted, and the absolute value pair of each difference is kept to be the minimum, which is the startup order of each queue in the queue family. The startup order table is saved in the queue storage table.

[0020] Furthermore, the queue reference system table and the queue storage table are loaded into the data storage environment to form a replication cycle startup sequence reference table, the reference table contains column information including: queue ID number, startup sequence, queue next startup interval, queue operation measurement time, queue theoretical operation time; the RPO theoretical value, RPO measurement value, RPO deviation, and deviation degree are saved in the table; whether to start the next optimization is determined based on the RPO parameter value saved in the table until an RPO indicator acceptable to the user is reached.

[0021] The present invention also provides an RPO tuning computer system, including a data acquirer, a computing processor and a storage display, wherein the data acquirer acquires operating parameters of a source storage device; after the data acquired by the data acquirer is stored in the storage display, the computing processor reads the data therein to execute any of the above-mentioned RPO tuning strategies.

[0022] Furthermore, the local storage data acquired by the data acquirer includes each replication queue ID, the local time stored locally, the start time, the end time, and the amount of data to be copied in the queue; the computing processor calculates the queue running time and calculates the reference table; the storage display includes ROM, RAM and a screen, wherein the memory and registers respectively permanently and temporarily store the computing processor intermediate data and result data, and the display part displays the result table and the human-computer interaction interface.

[0023] The present invention also discloses an RPO tuning device, in which a program for implementing the RPO tuning strategy is arranged, and the program can implement any of the above-mentioned RPO tuning strategies.

[0024] Beneficial effects of the present invention: The present invention directly relies on local storage parameters and optimizes the replication queue through a reverse engineering method. It is easy to implement in engineering and effectively shields the time sensitivity of storage parameters, has little impact on storage performance, and has the characteristics of high confidence and high sensitivity in optimizing storage RPO indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The specific implementation of the present invention will be further explained below in conjunction with the accompanying drawings.

[0026] Figure 1 A schematic diagram of the deployment of the storage replication queue and the tuning device of the present invention;

[0027] Figure 2 It is a schematic diagram of the cumulative remaining time envelope of the replication queue;

[0028] Figure 3 It is a brief schematic diagram of the RPO tuning method of the present invention;

[0029] Figure 4 It is a workflow diagram of the RPO tuning method of the present invention;

[0030] Figure 5 The figure is a schematic diagram of the composition of the RPO tuning computer system of the present invention. DETAILED DESCRIPTION

[0031] The present invention is further described in detail with reference to the embodiments and the accompanying drawings so that those skilled in the art can better understand and implement the present invention with reference to the description.

[0032] Unless otherwise specified, the mathematical symbols and formulas involved in the present embodiment all represent the same physical meaning. It is worth emphasizing that the data formulas used in the present embodiment are neither exclusive to the present invention nor the creation of the present invention. The application of mathematical formulas and symbols to the present invention and embodiments belongs to the specific content of the present invention, and the derivation method process and the conclusions generated therefrom also belong to the content of the present invention.

[0033] Obviously, all other embodiments obtained by those skilled in the art using other mathematical expressions and the same derivation method without any creative work shall fall within the scope of protection of the present invention.

[0034] The terms used in the disclosure of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the disclosure of the present invention. The attributive and quantifier descriptions used in the disclosure of the present invention and the appended claims are only for the purpose of clearly expressing the present invention and have no special restrictive meaning.

[0035] Combination Figure 1-Figure 5 The conditions discussed in the present invention are: when IP networking is used, the actual network bandwidth (such as 1000Mbps or 125MBps) is greater than the average write IO bandwidth of the business during the replication cycle. When the actual average write IO bandwidth of the local business system is greater than the actual network bandwidth, the solution is usually to extend the replication time or expand the bandwidth, which is not within the scope of the present invention.

[0036] Creating data copies and storing them off-site is a simple and effective measure for asynchronous data protection. Usually, off-site data replication uses circuits or optical fibers to achieve transmission based on IP networks, such as Figure 1 As shown, data replication between local storage and remote storage is often implemented and managed based on logical volumes.

[0037] Before the replication queue runs, you need to create the same volume group, logical volume, and file system in the local storage and the remote storage. The local storage transfers the data copy to the remote storage over the IP network in the form of a task.

[0038] Usually, replication work and tasks are saved in the local storage in the form of replication queues, and sent to the off-site storage in the form of queue sequence. The queue operation status is saved in the form of records in the local storage log file.

[0039] as follows Figure 2 As shown in the figure, as the queue family is started at time t1, the accumulated remaining time of the queue gradually increases, reaches a peak at time t2, and gradually decreases over time until the end at time t3, when a complete replication cycle ends. The start and end time of the queue family from t1 to t3 is the RPO measurement value of this replication cycle.

[0040] Easy to understand Figure 2 The total running time of queues 1 to 7 that have not been optimized on the timeline is usually not fixed, and the running time of the previous running queue will be affected by the subsequent queues. This replication model conforms to the typical memory grouping system model, which has no independence and stability, and requires optimization of system utilization to reduce the total running time of the queues.

[0041] Avoid using and copying Figure 2 When the phenomenon described in the above occurs, the local storage management queue will abnormally consume storage resources at time t2, which may easily cause congestion of the network and storage IO queues, resulting in a sharp increase in the local storage log space, causing replication delays or even restarts, thereby increasing the RPO value.

[0042] The tuning strategy disclosed in the present invention is Figure 2 The replication queue described in the optimization is optimized to make its startup and running time stable and independent. The optimization finally uses the startup time and the running end time as output to form a reference system table, and also gives the measured value, theoretical value and deviation of RPO.

[0043] The RPO tuning strategy, computer system and device of the present application use logical volume replication as the unit between the replication queues of the local data storage device and the remote data storage device. There is more than one replication queue, and the number of replication tasks n is created as the number of replication queues. Its goal is to make each replication task in the replication process have independent increments and balanced occupation of communication bandwidth. The sum of the replication time widths between multiple queues n evenly occupies the entire replication statistical cycle, and there is no bandwidth crowding or idleness. Make the total time that multiple queues occupy the channel as small as possible, and in the best case, tend to the theoretical RPO value. Ensure that the running time of multiple queues is consistent with the data increment ΔM generated during the replication cycle T Linearly related, the measured running time of the i-th replication queue approaches the theoretical value in the best case.

[0044] Specifically, in this embodiment, the queues eventually show stability and independence, and the running time of the replication queues conforms to the Erlang distribution with parameters n and λ, and its probability density function is:

[0045]

[0046] The cumulative distribution function for each replication queue is:

[0047]

[0048] Wherein, n refers to the number of replication queues, λ refers to the inverse of the number of queues, and this embodiment provides a reference table in which the parameter n is 1 to 7 and the parameter λ is the average value of the queue within 24 hours.

[0049] In order to more accurately express the reference table, the probability density function and cumulative probability distribution described in Formula 1 and Formula 2 need to be normalized. For example, the normalization coefficient of a queue group with a number of 7 is 6 / 21. It should be noted that the normalization system does not affect the final result of the reference table, but is only for the convenience of calculation. Those skilled in the art can still use other normalization systems for calculation, but they need to match the number of queue groups.

[0050] Then calculate the time distribution table of queues 1 to 7 one by one, and control the crossover probability between the same queues to not exceed the confidence interval, and you can get the reference system table shown in Table 1 below:

[0051] Table 1

[0052] Serial number Number of queues Normalization coefficient Queue start time 1 1 — 000 2 2 2 [11.01] 3 3 1 [8.55][13.49] 4 4 2 / 3 [8.00][12.67][16.00] 5 5 1 / 2 [7.50][11.85][15.55][18.50] 6 6 2 / 5 [7.16][11.36][15.00][18.00][20.24] 7 7 6 / 2 [6.00][9.49][12.57][15.37][17.71][19.89]

[0053] In Table 1, the number of queues is the total number of replication queues, the normalization coefficient is the weighted calculation index for forming control data, and the start time is expressed in absolute numbers. Professional and technical personnel in this field can convert it into relative time based on the time counting method. For example, 7.50 is counted in time, which means 7:30 with 0 o'clock as the reference, and the rest are deduced in the same way.

[0054] In Table 1, the two time intervals [] indicate the theoretical running time of the queues in the queue family. Usually, the running time of the replication queue must be less than this theoretical running time. Taking queue family number 7 as an example, the theoretical running time of the first queue should not exceed 6 hours, the running time of the second queue should not exceed 3.49 hours, and so on.

[0055] In the replication cycle, the replication queue family contains n replication queues, and the data increment generated by the replication of the i-th replication queue is m i , the theoretical running time of the i-th replication queue is t i =m i / BW T ,and

[0056] Normally, t i The values ​​of are different, and they are arranged according to the minimum difference according to Table 1. After the queue is set to start time, the running time rule of queue i is: starting time t a , running time t i , the next start time is t i + RPO measurement. In addition, it is well understood that the RPO measurement is t e -t b , the theoretical value of RPO is T n , RPO deviation is [(t e -t b )-T n ] / T n .

[0057] Figure 3 It is a specific logical step of the method adopted by the embodiment of the present invention:

[0058] Step 301 represents the use of a computer system and device to periodically obtain Figure 1The parameter data such as the ID number of the locally stored replication queue, the locally stored local time, the start time, the end time, the amount of data to be copied in the queue, etc. are obtained and stored in the computer device.

[0059] Step 302 represents the use of a computer system and device in combination with the parameter data saved in step 301 to calculate Formula 1 and Formula 2 to form a reference table, calculate the start and end time of each queue in the last replication cycle of the replication queue family, calculate the capacity data of each queue, and store these result data in the computer device.

[0060] Step 303 represents the use of a computer system and device to calculate RPO related parameter data in combination with the parameter data saved in step 301, calculate the sum of the theoretical running time of the queues based on the theoretical running time of each queue, and store these result data in the computer device.

[0061] Step 304 indicates that the computer system and the device save the result data to the storage component of the device and display the result data to the user. The user decides whether to restart the optimization according to the result data until the user's ideal index is reached.

[0062] Figure 3 The working principle of the method and device disclosed in the present invention is briefly described. Figure 4 The computer running implementation program of the method disclosed in the present invention is publicly displayed through the program principle diagram.

[0063] exist Figure 3 In the embodiment of the present invention, an RPO tuning method is proposed, which is mainly divided into three stages: obtaining parameters, injecting parameters into a calculation program, and forming a strategy table and RPO related values ​​with the calculation results and saving them to a storage device for users to review at any time.

[0064] Figure 3 The procedural steps demonstrated include:

[0065] Figure 3 Step 1 in the figure indicates that the computer system and the device periodically obtain Figure 1 The locally stored replication queue ID, the locally stored local time, the start time, the end time, the amount of data to be copied in the queue, and other parameter data are shown in . Then the following calculations are completed in the computer system and device in sequence:

[0066] S1. After the tuning device obtains the local storage parameter data, it calculates the start and end time of the i-th replication queue and obtains the data increment m generated by the i-th replication queue i , calculate the total amount of data to be copied ΔM T .

[0067] S2. After the tuning device obtains the locally stored parameter data, it calculates the RPO measurement time value t according to the start time and end time. e -t b .

[0068] S3. After the tuning device obtains the locally stored parameter data, it calculates and forms a queue reference system table Table 1 as shown in Table 1 according to the number of queues n based on Formula 1 and Formula 2.

[0069] S4. After the tuning device obtains the local storage parameter data, it calculates the amount of data to be copied in the queue of the previous cycle m i and ΔM T .

[0070] S5. Calculate the theoretical running time of each queue in all queue families, and combine with step 2 to calculate the RPO measurement value, RPO theoretical value, and RPO deviation.

[0071] S6. After calculating the above parameters, refer to the interval in Table 1 minus the difference in the actual running time of the queue, and keep the absolute value of each difference to the minimum, which is the startup order of each queue in the queue family. Save the startup order table to the queue storage table Table 2.

[0072] The computer system and device load Table1 and Table2 into the data storage environment to form a reference table for the startup sequence of the next replication cycle. The column information contained in the reference table specifically includes: queue ID number, startup sequence, queue next startup interval, queue operation measurement time, and queue theoretical operation time.

[0073] The RPO theoretical value, RPO measured value, RPO deviation, and deviation degree are saved in Table 3.

[0074] S7. The user decides whether to start the next optimization according to the RPO parameter value saved in Table 3 until an RPO indicator acceptable to the user is reached.

[0075] In summary of the above invention disclosure, the embodiments of the present invention directly rely on individual parameters of local storage, and use mathematical analysis tools and simple calculation procedures to achieve optimization of the replication queue, with minimal impact on local storage and no impact on off-site storage.

[0076] The parameters acquired by the computer system and device of the present invention have time stability, and the sampling and value taking shield the time sensitivity of the parameters, so the confidence and sensitivity are very high. The computer system and device are easy to implement in engineering.

[0077] The computer system and device (tuning device) described in the present invention are implemented by Figure 5 Show. Figure 5In the figure, the computer system and device are composed of three parts: a data acquirer 401, a computing processor 402, and a storage display 403.

[0078] The data acquirer 401 acquires the local storage data through the rest interface, and copies the queue ID, the local time of the local storage, the start time, the end time, the amount of data to be copied in the queue, and other parameters. Except for the amount of data to be copied in the queue, the other parameters can also be directly acquired on the web page or interactive interface.

[0079] After the data acquired by the data acquisition device is stored in the storage display 403, the computing processor 402 reads the data in 403 and executes Figure 4 The computer programs shown in 2 to 7.

[0080] The storage display is generally an electronic device composed of ROM, RAM and a screen, which stores and registers permanent and temporary intermediate data and result data of the computing processor 402, and the display part is used to display the result table and the human-computer interaction interface.

[0081] The network bandwidth used for long-distance data transmission is fixed for a long period of time, and its quality of service and service level BW T Remains unchanged for a fixed long time. Data replication tasks appear on the transmission network in the form of several queues according to the user-side settings, and the tuning strategy makes the replication queues present independent increments.

[0082] The method, computer system and device for remote asynchronous data replication RPO tuning of the present invention periodically obtains the time, capacity and number of queues in the local storage, firstly theoretically calculates the queue operation time interval through a random function, combines the queue capacity, forms a time fitting distribution table, and gives the start time and the next start time of each queue. Then, the RPO theoretical value and measured value are calculated by measuring the queue data capacity, and finally, the RPO deviation is provided to prompt the user to optimize the RPO effect. The present invention directly relies on local storage parameters and realizes the optimization of the replication queue through a reverse engineering method. It is easy to implement in engineering, and effectively shields the time sensitivity of the storage parameters, has little impact on storage performance, and has the characteristics of high confidence and high sensitivity in optimizing the RPO index.

[0083] It should be noted that the professional terms and quantitative descriptions involved in this article are only used to clearly present the methods and operating steps to those skilled in the art, and do not imply or restrict the specific scenarios of the disclosed invention. The terms and symbols used in the disclosed invention are not exclusive, and the random models used are not exclusive.

[0084] Many specific details are described in the above description to facilitate a full understanding of the present invention. However, the above description is only a preferred embodiment of the present invention. The present invention can be implemented in many other ways different from those described herein, so the present invention is not limited to the specific implementation disclosed above. At the same time, any person familiar with the art can make many possible changes and modifications to the technical solution of the present invention using the methods and technical contents disclosed above without departing from the scope of the technical solution of the present invention, or modify it into an equivalent embodiment of equivalent changes. Any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of protection of the technical solution of the present invention.

Claims

1. A RPO tuning method, characterized by: The steps include: S1, periodically obtain local storage parameter data; S2, calculate the replication queue time reference system table, and calculate the running time and capacity changes of the replication queue family in the last replication cycle; S3, forms the startup sequence and operation theoretical schedule for the next replication cycle; S4, calculates RPO related data and decides whether to start optimization again until the ideal RPO indicator is reached.

2. The RPO tuning method according to claim 1, characterized in that: In S1, the parameter data includes the locally stored replication queue ID, the locally stored local time, the start time, the end time, and the amount of data to be copied in the queue.

3. The RPO tuning method according to claim 1, characterized in that: In S2, the time step and start time in the replication queue time reference system table are used to calculate the start and end time and data increment of the replication queue family in the previous cycle, the theoretical running time of each queue in the replication queue family, and the start and end time to form the RPO measurement value of the previous cycle.

4. The RPO tuning method according to claim 3, characterized in that: Calculate the RPO theoretical value, RPO deviation value, and RPO deviation of the replication queue in the previous cycle, and refer to the RPO deviation to decide whether to restart the optimization.

5. The RPO tuning method according to claim 4, characterized in that: In S3, the start and end time of the next cycle of replication queue family operation and data increment are tracked, and the running time of the queue family and the start and end time are repeatedly calculated to form the RPO measurement value of the next cycle.

6. The RPO tuning method according to claim 5, characterized in that: After obtaining the local storage parameter data, the cumulative distribution function is calculated according to the probability density function based on the number of queues n, and the replication queue time reference system table is calculated; In the replication cycle, the replication queue family contains n replication queues, and the data increment generated by the replication of the i-th replication queue is m i , the amount of data to be copied ΔM T , the start time of queue i is t b , the end time is t e , running time t i , the next start time is t i +RPO measurement value, the theoretical running time of the i-th replication queue is , B.W. T The theoretical value of RPO is a fixed bandwidth between replication teams. , the measured value of RPO is t e- t b , the deviation value of RPO is (t e- t b )-T n , RPO deviation is [(t e- t b )-T n ] / T n ; Refer to the interval of the replication queue time reference table minus the difference of the actual running time of the queue, and keep the absolute value pair of each difference as small as possible, which is the startup order of each queue in the queue family, and save the startup order table to the queue storage table.

7. The RPO tuning method according to claim 6, characterized in that: The replication queue time reference system table and the queue storage table are loaded into the data storage environment to form a replication cycle startup sequence reference table. The reference table contains column information including: queue ID number, startup sequence, queue next startup interval, queue operation measurement time, and queue theoretical operation time. The RPO theoretical value, RPO measurement value, RPO deviation, and deviation degree are saved in the table. According to the RPO parameter value saved in the table, it is decided whether to start the next optimization until the RPO index acceptable to the user is reached.

8. An RPO tuning computer system, characterized in that: The method comprises a data acquirer, a computing processor and a storage display, wherein the data acquirer acquires the operating parameters of the source storage device; after the data acquired by the data acquirer is stored in the storage display, the computing processor reads the data therein and executes the RPO tuning method as described in any one of claims 1 to 7.

9. The RPO tuning computer system according to claim 8, characterized in that: The local storage data acquired by the data acquirer includes each replication queue ID, local time of local storage, start time, end time, and the amount of data to be copied in the queue; The computing processor calculates the running time of the calculation queue and calculates the reference table; the storage display includes ROM, RAM and screen, wherein the memory and register respectively permanently and temporarily store the intermediate data and result data of the computing processor, and the display part displays the result table and the human-computer interaction interface.

10. An RPO tuning device, characterized in that: The tuning device is provided with an implementation program of the RPO tuning method, and the program can implement the RPO tuning method as described in any one of claims 1 to 7.

Images