A centralized management method of a data center cluster

Abstract

This invention discloses a centralized management method for a data center cluster, relating to the field of data management technology. The method includes acquiring the original dataset from the database cluster; deduplicating the original dataset based on the number of samples in the important score dataset to obtain a deduplicated dataset; randomly dividing the data within the deduplicated dataset based on the number of samples to obtain multiple granular spheres; classifying the data using the granular spheres; compressing and encrypting the classified data based on importance scores to obtain encrypted data; and migrating and storing the encrypted data based on importance scores. By deduplicating a large amount of data in the database, redundant data is reduced; by using granular spheres and calculating dispersion, the data becomes more compact; and by classifying the data and storing it according to its importance level, efficient data management is achieved. Furthermore, data compression reduces storage costs, and encryption protects data security, thus enabling efficient data management.

Description

Technical Field

[0001] This invention belongs to the field of data management technology, specifically relating to a centralized management method for data center clusters. Background Technology

[0002] A data center cluster consists of multiple geographically dispersed but interconnected data centers that collaboratively provide computing, storage, and networking services. Through resource sharing, load balancing, and failover, it achieves high availability and scalability, supporting cloud services and enterprise applications. Centralized management of a data center cluster is an operational model that integrates, schedules, and monitors resources such as servers, storage, and networks distributed across multiple locations through a unified management platform. It abstracts physical resources into resource pools, combining virtualization and automated operations and maintenance to achieve dynamic allocation of computing power, unified control of data traffic, centralized fault alerts, and rapid recovery. Centralized management improves resource utilization, simplifies operational processes, reduces labor costs, and ensures system stability and data security through real-time monitoring and policy configuration, making it suitable for the business expansion needs of large-scale enterprises and cloud service providers.

[0003] The prior art (publication number: CN112052287A) discloses a centralized management method for a data center cluster, wherein the data center cluster includes multiple data centers, and the method includes: acquiring environmental control data of at least one data center in the data center cluster; converting the acquired environmental control data into standardized environmental control data according to a preset environmental control data standard; and classifying and storing the standardized environmental control data according to the importance of the data.

[0004] However, in practical use, this method does not perform deduplication on the data before data classification, which leads to some duplicate and redundant data consuming computing resources. Furthermore, the operation of classifying data according to its importance score is only briefly introduced without detailed solutions, and the method does not store data on different storage media according to different importance scores. This results in poor performance in hierarchical data storage, thereby affecting the efficiency of users in the process of centralized data management. Summary of the Invention

[0005] The purpose of this invention is to solve the problem of low efficiency of hierarchical storage in the centralized management of data center clusters, and to propose a centralized management method for data center clusters.

[0006] In this invention, a centralized management method for a data center cluster is first proposed, the method comprising:

[0007] Obtain the original dataset from the database cluster, and perform deduplication processing on the original dataset to obtain the deduplicated dataset;

[0008] Multiple granular spheres are obtained by randomly dividing the data in the deduplicated dataset according to the number of samples in the granular sphere algorithm.

[0009] Calculate the dispersion of the target particle size sphere. If the dispersion is greater than 0, perform a first split on the target particle size sphere to obtain two first sub-particle size spheres, minimizing the sum of the dispersions of the two first sub-particle size spheres. Add the first sub-particles of each particle size sphere to the first sub-particle set. The target particle size sphere is any one of multiple particle size spheres.

[0010] Calculate the mean radius and median radius of each first sub-sphere in the first sub-sphere set. If the target first sub-sphere meets the preset conditions, then split the target first sub-sphere a second time to obtain two second sub-spheres, minimizing the sum of the dispersion of the two second sub-spheres, and add the second sub-spheres of each first sub-sphere to the second sub-sphere set; the target first sub-sphere is any one of the first sub-spheres in the first sub-sphere set.

[0011] Calculate the importance score of each second sub-ball in the second sub-ball set, and perform hierarchical processing based on the importance score to obtain the hierarchical data;

[0012] The classified data is compressed and encrypted to obtain encrypted data; the encrypted data is then migrated and stored.

[0013] Optionally, the process for calculating the dispersion of the particle size sphere is as follows:

[0014] Calculate the distribution metric and weighted distribution metric of the particle size sphere, and subtract the distribution metric from the weighted distribution metric to obtain the dispersion;

[0015] The graded data obtained by grading the important scores includes:

[0016] Calculate the Euclidean distance between two granular spheres as the fuzzy similarity;

[0017] The data importance score is calculated based on the fuzzy similarity.

[0018] The importance scores of the data are used to classify the data into different levels, resulting in graded data.

[0019] Optionally, the data importance score is calculated based on the fuzzy similarity, including:

[0020] A fuzzy relationship matrix is ​​constructed using the fuzzy similarity between granular spheres; the information particle size distribution of each granular sphere is calculated based on the fuzzy relationship matrix; and the importance of each attribute is then calculated based on the information particle size distribution.

[0021] Arrange the attributes in ascending order of importance to obtain an ordered attribute sequence. Based on the ordered attribute sequence, construct a simplified attribute sequence and a complex attribute sequence.

[0022] For each subset of attributes in the attribute simplification sequence, information particles of the granular sphere are extracted through its fuzzy relation matrix to form an attribute simplification particle sequence;

[0023] For each subset of attributes in the complex attribute sequence, information particles of the granular sphere are extracted through its fuzzy relation matrix to form a complex attribute particle sequence;

[0024] The weight of each granular sphere is calculated based on the information granularity of a single attribute;

[0025] The mean of the information particle cardinality in the attribute-simplified granular sequence and the attribute-complex granular sequence is fused, and the importance score of the granular sphere is calculated.

[0026] Optionally, the mean of the information particle cardinality in the attribute-simplified granular sequence and the attribute-complex granular sequence is fused, and the importance score of the granularity sphere is calculated, specifically including:

[0027]

[0028] in, For the importance score of the i-th granular sphere, Let i be the weight of the i-th granular sphere. This represents the i-th granular sphere. The cardinality of the q-th information particle in the attribute simplification particle sequence is represented. The cardinality of the q-th information particle in the complex particle sequence of attributes is represented by m; the total number of attributes in the attribute set is represented by g; and the total number of particle size balls in the particle size ball set is represented by g.

[0029] Optionally, the data is graded based on its importance score to obtain the graded data, specifically including:

[0030] Compare the important score with the threshold interval:

[0031] If the importance score is less than the lower limit of the threshold range, the data importance is classified into the third level.

[0032] If the importance score is not less than the lower limit of the threshold interval and not greater than the upper limit of the threshold interval, then the data importance is classified into the second level.

[0033] The importance score is compared with the threshold range. If the importance score is greater than the upper limit of the threshold range, the data importance is classified into the first level.

[0034] Optionally, the graded data may be compressed, specifically including:

[0035] Step 1: Obtain the original binary sequence of the data after grading. For the original binary sequence to be compressed of length n, count the number of 1s h and convert the binary sequence into an h-balanced code combination.

[0036] Step 2: Starting from the least significant bit of the h-balanced code combination, remove consecutive 0s until the first 1 is encountered, resulting in a binomial Y of length y, where y <n;

[0037] Step 3: Convert the binomial number into a number of the natural number system with length t using the binomial numbering function;

[0038] Step 4: Determine whether to compress based on the compression ratio formula, only for the compression ratio. Sequences larger than the threshold are compressed; otherwise, the original binary sequence is transmitted directly.

[0039] Optionally, the graded data can be encrypted to obtain encrypted data, specifically including:

[0040] Sensitive data after classification is encrypted using the Advanced Encryption Standard (AES-256), and the encryption key is securely generated, distributed, and rotated using a Key Management Service (KMS) to obtain encrypted data. During data transmission, the Transport Layer Security (TLS) protocol is used to encrypt the data channel. The sensitive data includes user authentication information, critical business data, and system configuration data.

[0041] Optionally, the encrypted data may be migrated and stored, specifically including:

[0042] For the data in the first level, it is stored in the first storage space and multiple copies are redundantly backed up;

[0043] Data at the second level is stored in a second storage space and redundantly backed up.

[0044] The data at the third level is stored in the third storage space;

[0045] The read / write rate of the first storage space > the read / write rate of the second storage space > the read / write rate of the third storage space.

[0046] Optionally, after migrating and storing the encrypted data, the method further includes:

[0047] By periodically verifying the hash value of the data, we ensure that the integrity of the data is not compromised during storage and transmission.

[0048] Real-time monitoring of user and system data access behavior patterns; once an abnormal access pattern is detected, an alarm is immediately triggered and isolation is initiated.

[0049] Optionally, differentiated backup strategies can be developed based on importance:

[0050] Use object storage services to store backup data in multiple physical locations;

[0051] Establish an off-site disaster recovery center database and synchronously replicate critical data to the disaster recovery center database. In the event of a failure in the primary data center, the system can quickly switch to the disaster recovery center to restore business operations.

[0052] The beneficial effects of this invention are:

[0053] This invention proposes a centralized management method for data center clusters. By deduplicating a large amount of data in the database to reduce redundant data, and by using granular sphere partitioning and discreteness calculation to make the data more compact, the hierarchical data is stored on corresponding storage media according to its importance level, achieving high-efficiency data management. Furthermore, data compression reduces storage costs, encryption protects data security, and access control and disaster recovery improve data availability. This makes data processing efficient, secure, and reliable, meeting the high demands of data management. Attached Figure Description

[0054] The invention will now be further described with reference to the accompanying drawings.

[0055] Figure 1 A flowchart illustrating a centralized management method for a data center cluster provided in an embodiment of the present invention;

[0056] Figure 2 A block diagram illustrating another centralized management method for a data center cluster provided in an embodiment of the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and B can represent: A alone, A and B simultaneously, and B alone. Furthermore, descriptions involving "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" can explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0058] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0059] This invention provides a centralized management method for data center clusters. See also... Figure 1 , Figure 1 A flowchart illustrating a centralized management method for a data center cluster provided in an embodiment of the present invention. The method includes the following steps:

[0060] S101: Obtain the original dataset from the database cluster, and perform deduplication on the original data to obtain the deduplicated dataset;

[0061] S102, multiple granular spheres are obtained by randomly dividing the data in the deduplicated dataset according to the number of samples in the granular sphere algorithm;

[0062] S103, calculate the dispersion of the target particle size sphere. If the dispersion is greater than 0, perform the first split on the target particle size sphere to obtain two first sub-particle size spheres, minimize the sum of the dispersion of the two first sub-particle size spheres, and add the first sub-particles of each particle size sphere to the first sub-particle set.

[0063] S104, calculate the mean radius and median radius of each first sub-ball in the first sub-ball set. If the target first sub-ball meets the preset conditions, split the target first sub-ball a second time to obtain two second sub-balls, so that the sum of the dispersion of the two second sub-balls is minimized, and add the second sub-balls of each first sub-ball to the second sub-ball set.

[0064] S105, calculate the importance score of each second sub-ball in the second sub-ball set, and perform hierarchical processing based on the importance score to obtain the hierarchical data;

[0065] S106, Compress and encrypt the graded data to obtain encrypted data; migrate and store the encrypted data.

[0066] The target particle size sphere is any one of multiple particle size spheres; the target first sub-sphere is any one of the first sub-spheres.

[0067] The centralized management method for data center clusters provided by this invention reduces redundant data by deduplicating a large amount of data in the database, makes the data more compact through granular sphere partitioning and discreteness calculation, stores data according to importance levels after data classification, achieves high-efficiency data management, reduces storage costs by data compression, protects data security by encryption, and improves data availability through access control and disaster recovery. This makes data processing efficient, secure and reliable, meeting high data management requirements.

[0068] In one implementation, deduplication of the original data includes calculating hash values ​​for data blocks to identify duplicate data before writing the data to the storage medium; saving storage space and improving performance and data reliability by deduplication; and randomly dividing the data into multiple granular spheres based on the number of samples in the deduplicated dataset; wherein the specific number of samples is the optimal division method obtained from historical data, and can also be adjusted by relevant personnel.

[0069] In one implementation, the preset condition is the radius of the first target sub-sphere. ,in Let be the average radius of each first child ball. The median radius of each first sub-sphere; if the dispersion is not greater than 0, the target particle size sphere is not split for the first time, and the target particle size sphere that has not been split for the first time is added to the first sub-sphere set as the first sub-sphere; if the target first sub-sphere does not satisfy the radius... If the target first sub-ball is not split a second time, the target first sub-ball that has not been split a second time will be added to the second sub-ball set as the second sub-ball.

[0070] In one implementation, the dispersion of the granular spheres is used for splitting to optimize the data distribution and make the data more compact. The two splitting operations are combined with the mean and median of the radius to effectively control the scale of the granular spheres and avoid over-segmentation or insignificant processing. Finally, the fuzzy similarity between granular spheres is calculated by Euclidean distance to accurately reflect the relative relationship of the data and improve the efficiency and accuracy of data hierarchical processing.

[0071] In one embodiment, the process of calculating the dispersion of the particle size sphere is as follows:

[0072] Calculate the distribution measure and weighted distribution measure of the particle size spheres, and subtract the distribution measure from the weighted distribution measure to obtain the dispersion.

[0073] The graded data, obtained by classifying the data based on the importance scores, includes:

[0074] Calculate the Euclidean distance between two granular spheres as the fuzzy similarity;

[0075] The importance score of the data is calculated based on fuzzy similarity.

[0076] The data is classified into different importance levels based on the importance scores to obtain the graded data.

[0077] In one implementation, calculating the distribution metric and weighted distribution metric of the particle size sphere includes first performing a temporary trial partitioning of the target particle size sphere to generate two temporary subspheres. and ;

[0078]

[0079]

[0080]

[0081] in, Let be the distribution metric for the t-th target particle size sphere. A weighted distribution metric for the target particle size sphere. Let t be the target particle size sphere. Let be the number of samples within the t-th target particle size sphere. This represents the sum of the Euclidean distances from all samples within the t-th target granularity sphere to its center. This represents the i-th sample in the target granularity sphere. G represents the center of the t-th target particle size sphere; G is the target particle size sphere to be split, and G1 and G2 are two temporary subspheres. , , These represent the sample size of the target particle sphere and the two temporary sub-spheres, respectively.

[0082] Subtracting the distribution measure from the weighted distribution measure yields the dispersion μ:

[0083]

[0084]

[0085] Where c represents the average value of the samples within the sphere and is the center of the particle size sphere, and k is the total number of samples within the sphere;

[0086] First calculate the centers of the two balls and Euclidean distance:

[0087]

[0088] Calculate fuzzy similarity again The process is as follows:

[0089]

[0090] Where B represents the attribute set of the grain size sphere. Let represent the fuzzy similarity between the centers of the two spheres under the attribute set. Let Euclidean distance be the center-to-center distance between the two spheres under the attribute set. The center of the i-th granularity sphere under the q-th attribute. The center of the j-th granularity sphere under the q-th attribute, For the q-th attribute in the attribute set, ∈[0,1] is a hyperparameter, a threshold used to control similarity, set according to the experiment. The center of the i-th granular sphere, Let be the center of the j-th granular sphere.

[0091] In one implementation, the granular sphere is essentially a semantically meaningful information unit covering a local data region; the radius of the granular sphere is a measure of the uncertainty or generalization range of this information unit, quantifying the average dispersion of data samples within the sphere; by calculating the difference between the distribution metric and the weighted distribution metric of the granular sphere, the dispersion can be obtained, which can optimize the data distribution and improve the organization compactness; by using the Euclidean distance between the centers of the granular spheres as fuzzy similarity, the importance score of the data can be calculated and the importance level can be divided, which can accurately distinguish the data value, improve the classification accuracy and efficiency, and effectively handle complex target detection tasks.

[0092] In one embodiment, calculating the data importance score based on fuzzy similarity includes:

[0093] A fuzzy relation matrix is ​​constructed using the fuzzy similarity between granular spheres; the information particle size of each granular sphere is calculated based on the fuzzy relation matrix; and the importance of each attribute is then calculated based on the information particle size.

[0094] Arrange the attributes in ascending order of importance to obtain an ordered attribute sequence. Based on the ordered attribute sequence, construct a simplified attribute sequence and a complex attribute sequence.

[0095] For each subset of attributes in the attribute simplification sequence, information particles of the granular sphere are extracted through its fuzzy relation matrix to form an attribute simplification particle sequence;

[0096] For each subset of attributes in the complex attribute sequence, information particles of the granular sphere are extracted through its fuzzy relation matrix to form a complex attribute particle sequence.

[0097] The weight of each granular sphere is calculated based on the information granularity of a single attribute;

[0098] The mean of the information particle cardinality in the attribute-simplified granular sequence and the attribute-complex granular sequence is fused, and the importance score of the granular sphere is calculated.

[0099] In one implementation, a fuzzy relation matrix is ​​constructed. for:

[0100]

[0101] in, For granular spheres With granular spheres The fuzzy similarity, where g is the total number of granular spheres in the set of granular spheres;

[0102] Calculate each granular sphere based on the fuzzy relation matrix. Information particle size The importance of an attribute is calculated based on the number of information granularities. The process is as follows:

[0103]

[0104]

[0105] Based on a single attribute Information particle number calculation for each particle size sphere The weights include:

[0106]

[0107] in, Let i be the weight of the i-th granular sphere. For a single attribute Lower particle size balls The information particle cardinality is given by g, where g is the total number of particle spheres in the particle sphere set, and m is the total number of attributes in the attribute set B.

[0108] In one implementation, a fuzzy relation matrix is ​​constructed using fuzzy similarity, the cardinality of information particles and the importance of attributes are calculated, and simplified and complex sequences are constructed by ranking attributes. Information particles are extracted from different sequences to form granular sequences, and the importance score of the granular sphere is calculated by combining the cardinality mean of information particles. This can more accurately measure the importance of data, improve the efficiency and accuracy of data processing, and enhance the interpretability of data.

[0109] In one embodiment, the mean of the information particle cardinality in the attribute-simplified granular sequence and the attribute-complex granular sequence is fused, and the importance score of the granularity sphere is calculated, specifically including:

[0110]

[0111] in, For the importance score of the i-th granular sphere, Let i be the weight of the i-th granular sphere. This represents the i-th granular sphere. The cardinality of the q-th information particle in the attribute simplification particle sequence is represented. The cardinality of the q-th information particle in the complex particle sequence of attributes is represented by m; the total number of attributes in attribute set B is represented by g; and the total number of particle size spheres in the particle size sphere set is represented by g.

[0112] In one implementation, by constructing a fuzzy relation matrix and calculating the information granularity cardinality, combined with the fusion of attribute simplification and complex sequences, the accuracy and efficiency of anomaly detection are effectively improved. It utilizes multi-granularity feature analysis to capture data anomalies from different dimensions, and the dependence on attribute importance measurement and the rationality of weight design become important operations for performance improvement.

[0113] In one embodiment, the data is graded based on its importance score to obtain the graded data, specifically including:

[0114] Compare the important scores with the threshold range:

[0115] If the importance score is less than the lower limit of the threshold range, the data importance is classified into the third level.

[0116] If the importance score is not less than the lower limit of the threshold interval and not greater than the upper limit of the threshold interval, then the data importance is classified into the second level.

[0117] The importance score is compared with the threshold range. If the importance score is greater than the upper limit of the threshold range, the data importance is classified as the first level.

[0118] In one implementation, the threshold range is the optimal range derived from historical data, which provides a good foundation for data classification. By setting the threshold range for importance scores, the data is divided into three levels, which can quickly and clearly filter out the data areas with high importance, realize the priority ranking of data, and provide hierarchical management for data governance. It is also convenient to take targeted measures according to different levels, thereby improving the efficiency of data management.

[0119] In one embodiment, the data after hierarchical classification is compressed, specifically including:

[0120] Step 1: Obtain the original binary sequence of the data after grading. For the original binary sequence to be compressed of length n, count the number of 1s h and convert the binary sequence into an h-balanced code combination.

[0121] Step 2: Starting from the least significant bit of the h-balanced code combination, remove consecutive 0s until the first 1 is encountered, resulting in a binomial Y of length y, where y <n;

[0122] Step 3: Convert the binomial number into a number of the natural number system with length t using the binomial numbering function;

[0123] Step 4: Determine whether to compress based on the compression ratio formula, only for the compression ratio. Sequences larger than the threshold are compressed; otherwise, the original binary sequence is transmitted directly.

[0124] In one implementation, see [link to implementation details]. Figure 2 , Figure 2 The diagram illustrates a centralized management method for a data center cluster, as provided in this embodiment of the invention. The threshold is based on experimental results. The compression ratio is calculated as follows:

[0125]

[0126] The compressed data will be restored, including:

[0127] Given the lengths t and h of the compressed binary number, the binary number is converted into the binomial number Y obtained in step two by using the inverse operation of the binomial numbering function;

[0128] Based on the fact that the binomial number Y ends with either 1 or 0, nh zeros or h ones are added to its least significant bit to restore the original binary sequence of length n.

[0129] In one implementation, the number of 0s in the binary sequence is counted and converted into h-balanced code combinations. Consecutive 0s or 1s are removed to form binomial numbers, which are then converted into natural numbers using a binomial numbering function, thus achieving data compression. Simultaneously, based on the compression ratio, only sequences with significant compression effects are compressed; otherwise, the original sequence is transmitted directly to ensure compression efficiency. The recovery process accurately restores the original sequence through inverse operations, achieving lossless compression. This approach effectively reduces data storage and transmission volume, improves efficiency, and is suitable for various data processing scenarios.

[0130] In one embodiment, the graded data is encrypted to obtain encrypted data, specifically including:

[0131] Sensitive data after classification is encrypted using the advanced encryption standard AES-256. The encryption key is securely generated, distributed, and rotated using the Key Management Service (KMS) to obtain encrypted data. During data transmission, the Transport Layer Security (TLS) protocol is used to encrypt the data channel. Sensitive data includes user authentication information, critical business data, and system configuration data.

[0132] In one implementation, AES-256 encryption is used to protect the confidentiality of sensitive data, KMS is used to achieve secure key generation, distribution and rotation, and TLS is used to encrypt the data transmission channel to ensure high security of data storage and transmission.

[0133] In one embodiment, migrating and storing encrypted data specifically includes:

[0134] For the first level of data, it is stored in the first storage space and multiple copies are redundantly backed up;

[0135] For the second level of data, it is stored in the second storage space and redundant backups are performed;

[0136] Data at the third level is stored in the third storage space;

[0137] First storage space read / write speed > Second storage space read / write speed > Third storage space read / write speed.

[0138] In one implementation, a differentiated storage strategy is adopted for data of different levels:

[0139] Tier 1 data is stored in tape libraries or cloud archive storage, and is deeply compressed and stored in multiple copies, which greatly reduces storage costs. At the same time, multiple copies prevent permanent data loss and meet the needs of archiving and long-term preservation.

[0140] The second level of data is stored on traditional hard disk drives or hybrid storage arrays, with appropriate redundancy backups to balance storage costs and reliability, ensuring that data is not lost under moderate performance conditions.

[0141] The third level of data is stored on high-performance storage media such as solid-state drives or memory, and is configured with low-latency data access paths, which can quickly respond to high-frequency access demands, reduce business latency, and improve the efficiency of critical tasks.

[0142] In one embodiment, after migrating and storing the encrypted data, the method further includes:

[0143] By periodically verifying the hash value of the data, we ensure that the integrity of the data is not compromised during storage and transmission.

[0144] Real-time monitoring of user and system data access behavior patterns; once an abnormal access pattern is detected, an alarm is immediately triggered and isolation is initiated.

[0145] In one implementation, abnormal access patterns include excessive data extraction, excessively fast access rates, login access from different locations, and abnormal IP addresses. By periodically verifying the data hash value, it is possible to effectively detect whether the data has been tampered with during storage and transmission, ensuring data integrity. At the same time, real-time monitoring of user and system data access behavior patterns can promptly detect abnormal access or suspicious operations, quickly trigger alarms, and take isolation or blocking measures, thereby enhancing the security of the data access process and reducing the impact and losses of security incidents.

[0146] In one embodiment, a differentiated backup strategy is defined:

[0147] Use object storage services to store backup data in multiple physical locations;

[0148] Establish an off-site disaster recovery center database and synchronously replicate critical data to the disaster recovery center database. In the event of a failure in the primary data center, the system can quickly switch to the disaster recovery center to restore business operations.

[0149] In one implementation, the physical location can be a data center in another city, such as the Qingpu District Data Center in Shanghai, the Xiaoshan District Cloud Computing Center in Hangzhou, Zhejiang Province, and the Suzhou Industrial Park Data Center in Jiangsu Province; differentiated backup strategies can effectively improve data security and business continuity; multi-location backup of object storage can prevent local physical damage, reduce costs, and is easy to expand; remote disaster recovery centers can synchronously replicate critical data, which can quickly restore business and reduce the impact of extreme situations; by combining the two, a defense in depth is formed to ensure high data availability.

[0150] The foregoing has described one embodiment of the present invention in detail, but this content is merely a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A centralized management method for a data center cluster, characterized in that, The method includes: Obtain the original dataset from the database cluster, and perform deduplication processing on the original dataset to obtain the deduplicated dataset; Multiple granular spheres are obtained by randomly dividing the data in the deduplicated dataset according to the number of samples in the granular sphere algorithm. Calculate the dispersion of the target particle size sphere. If the dispersion is greater than 0, perform a first split on the target particle size sphere to obtain two first sub-particle size spheres, minimizing the sum of the dispersions of the two first sub-particle size spheres. Add the first sub-particles of each particle size sphere to the first sub-particle set. The target particle size sphere is any one of multiple particle size spheres. Calculate the mean radius and median radius of each first sub-sphere in the first sub-sphere set. If the target first sub-sphere meets the preset conditions, then split the target first sub-sphere a second time to obtain two second sub-spheres, minimizing the sum of the dispersion of the two second sub-spheres, and add the second sub-spheres of each first sub-sphere to the second sub-sphere set; the target first sub-sphere is any one of the first sub-spheres in the first sub-sphere set. Calculate the importance score of each second sub-ball in the second sub-ball set, and perform hierarchical processing based on the importance score to obtain the hierarchical data; The graded data is compressed and encrypted to obtain encrypted data; the encrypted data is then migrated and stored. The process for calculating the dispersion of granular spheres is as follows: Calculate the distribution metric and weighted distribution metric of the particle size sphere, and subtract the distribution metric from the weighted distribution metric to obtain the dispersion; The graded data obtained by grading the important scores includes: Calculate the Euclidean distance between two granular spheres as the fuzzy similarity; The data importance score is calculated based on the fuzzy similarity. The importance scores of the data are used to classify the data into different levels, resulting in graded data. The importance score of the data is calculated based on the fuzzy similarity, including: A fuzzy relationship matrix is ​​constructed using the fuzzy similarity between granular spheres; the information particle size distribution of each granular sphere is calculated based on the fuzzy relationship matrix; and the importance of each attribute is then calculated based on the information particle size distribution. Arrange the attributes in ascending order of importance to obtain an ordered attribute sequence. Based on the ordered attribute sequence, construct a simplified attribute sequence and a complex attribute sequence. For each subset of attributes in the attribute simplification sequence, information particles of the granular sphere are extracted through its fuzzy relation matrix to form an attribute simplification particle sequence; For each subset of attributes in the complex attribute sequence, information particles of the granular sphere are extracted through its fuzzy relation matrix to form a complex attribute particle sequence; The weight of each granular sphere is calculated based on the information granularity of a single attribute; The mean of the information particle cardinality in the attribute-simplified granular sequence and the attribute-complex granular sequence is fused, and the importance score of the granular sphere is calculated.

2. The centralized management method for a data center cluster according to claim 1, characterized in that, The mean of the information particle cardinality in attribute-simplified granular sequences and attribute-complex granular sequences is fused, and the importance score of the granular sphere is calculated, specifically including: ； in, For the importance score of the i-th particle size sphere, Let i be the weight of the i-th granular sphere. This represents the i-th sphere. The cardinality of the q-th information particle in the attribute simplification particle sequence is represented. The cardinality of the q-th information particle in the complex particle sequence of attributes is represented; m refers to the total number of attributes in the attribute set; g represents the total number of particle spheres in the particle sphere set.

3. The centralized management method for a data center cluster according to claim 1, characterized in that, The graded data is obtained by classifying the data according to the importance scores of the data, specifically including: Compare the important score with the threshold interval: If the importance score is less than the lower limit of the threshold range, the data importance is classified into the third level. If the importance score is not less than the lower limit of the threshold interval and not greater than the upper limit of the threshold interval, then the data importance is classified into the second level. The importance score is compared with the threshold range. If the importance score is greater than the upper limit of the threshold range, the data importance is classified into the first level.

4. The centralized management method for a data center cluster according to claim 1, characterized in that, The data after classification is compressed, specifically including: Step 1: Obtain the original binary sequence of the data after grading. For the original binary sequence to be compressed of length n, count the number of 1s h and convert the binary sequence into an h-balanced code combination. Step 2: Starting from the least significant bit of the h-balanced code combination, remove consecutive 0s until the first 1 is encountered, resulting in a binomial Y of length y, where y <n; Step 3: Convert the binomial number into a number of the natural number system with length t using the binomial numbering function; Step 4: Determine whether to compress based on the compression ratio formula. Only compress sequences with a compression ratio φ greater than the threshold; otherwise, transmit the original binary sequence directly.

5. The centralized management method for a data center cluster according to claim 1, characterized in that, The classified data is then encrypted to obtain encrypted data, specifically including: Sensitive data after classification is encrypted using the Advanced Encryption Standard (AES-256), and the encryption key is securely generated, distributed, and rotated using a Key Management Service (KMS) to obtain encrypted data. During data transmission, the Transport Layer Security (TLS) protocol is used to encrypt the data channel. The sensitive data includes user authentication information, critical business data, and system configuration data.

6. The centralized management method for a data center cluster according to claim 3, characterized in that, Migrating and storing encrypted data specifically includes: For the data in the first level, it is stored in the first storage space and multiple copies are redundantly backed up; Data at the second level is stored in a second storage space and redundantly backed up. The data at the third level is stored in the third storage space; The read / write rate of the first storage space > the read / write rate of the second storage space > the read / write rate of the third storage space.

7. The centralized management method for a data center cluster according to claim 1, characterized in that, After migrating and storing the encrypted data, the method further includes: By periodically verifying the hash value of the data, we ensure that the integrity of the data is not compromised during storage and transmission. Real-time monitoring of user and system data access behavior patterns; triggering an alarm immediately upon detection of abnormal access patterns.

8. The centralized management method for a data center cluster according to claim 1, characterized in that, The method also includes developing differentiated backup strategies: Use object storage services to store backup data in multiple physical locations; Establish an off-site disaster recovery center database and synchronously replicate critical data to the disaster recovery center database. In the event of a failure in the primary data center, the system can quickly switch to the disaster recovery center to restore business operations.

Images