A cloud computing server data backup and rapid recovery system and method

Abstract

The application discloses a cloud computing server data backup and quick recovery system and method, relates to the technical field of cloud computing data protection, and comprises a service sensing agent, a metadata reconstructor and an intelligent recovery scheduler. The service sensing agent establishes an association graph of service core functions and underlying data blocks during application running, and maintains data block priority labels. The metadata reconstructor monitors and analyzes an incremental snapshot chain, combines early infrequently accessed snapshots to generate a virtual full-amount base point, and reconstructs metadata index relations. After receiving a recovery instruction, the intelligent recovery scheduler analyzes the reconstructed metadata chain, determines a recovery priority according to the association graph, preferentially recovers key data, and integrates a service preheating mechanism to start a computing instance in advance. The method corresponds to the above system to perform backup, metadata reconstruction and priority recovery steps. The application realizes quick service recovery, improves data recovery pertinence and user experience, and guarantees business continuity.

Description

Technical Field

[0001] This invention relates to the field of cloud computing data protection technology, specifically to a cloud computing server data backup and rapid recovery system and method. Background Technology

[0002] With the widespread application of cloud computing technology, data backup and rapid recovery of cloud computing servers have become key technologies for ensuring business continuity. In this field, incremental snapshot-based data protection methods are widely used due to their high storage efficiency. However, as system uptime increases, the incremental snapshot chain grows continuously, posing significant performance challenges in data recovery scenarios. How to effectively reduce recovery time targets and improve service availability is a pressing technical problem that needs to be solved in this field.

[0003] Existing technologies have made numerous attempts to optimize data recovery processes. For example, the invention patent publication number CN112380062A describes a method and system for rapid system recovery based on multiple system backup points. This method can quickly restore services using backup virtual machines when the source server fails. However, this technical solution mainly focuses on the rapid switching of computing resources. For the recovery of underlying storage data, especially when facing a lengthy incremental snapshot chain, the data recovery stage still needs to traverse the entire snapshot chain to locate and load data, and the latency bottleneck of the recovery process has not been fundamentally solved. Another example is the data backup method, apparatus, electronic device, and storage medium published by invention patent publication number CN120872688A, which aims to improve data retrieval efficiency. However, when processing massive amounts of data, the index reconstruction process itself is still time-consuming, and its recovery process is decoupled from the state of upper-layer services. Services must wait for critical data to be fully ready before responding, resulting in a still relatively long recovery time window for service availability.

[0004] In summary, existing technical solutions fail to effectively address a key technical problem: in incremental snapshot-based cloud server data recovery, how to overcome the latency bottleneck caused by lengthy snapshot chain traversal and indiscriminate data loading to achieve instant service recovery and seamless user experience. Specifically, existing technologies suffer from two main drawbacks: first, the lengthy incremental snapshot chain results in significant metadata traversal overhead, severely slowing down the startup and execution of the recovery process; second, the data recovery process lacks awareness of the core needs of upper-layer services and cannot distinguish the priority of data recovery, causing service instances to fail to provide timely and effective services even if they are started in advance due to missing critical data, resulting in a poor user experience.

[0005] Therefore, there is a need in this field for an innovative technical solution that can deeply integrate the snapshot management of the storage layer with the service recovery needs of the application layer, and fundamentally optimize the data recovery path through intelligent metadata management and data scheduling strategies, so as to achieve a seamless connection from data recovery to service availability. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a cloud computing server data backup and rapid recovery system and method. By establishing a relationship graph between business and data through a service-aware agent, optimizing the snapshot chain structure through a metadata reconstructor, and achieving priority recovery and service preheating through an intelligent recovery scheduler, this invention can effectively shorten data recovery time, achieve rapid service preheating and priority availability of core functions, and significantly improve business continuity and user experience.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution: In one aspect, a cloud computing server data backup and rapid recovery system and method, comprising a service-aware agent, a metadata reconstructor, and an intelligent recovery scheduler: The service-aware agent is used to collect and establish a graph of the relationship between the core functions of the service and the underlying data blocks during application operation. The metadata reconstructor is used to monitor and analyze the metadata of the incremental snapshot chain, identify the set of early incremental snapshots that are not frequently accessed based on the snapshot access pattern, and merge the metadata of the set in the background to generate a virtual full base point. At the same time, it reconstructs the metadata index relationship between subsequent incremental snapshots and the virtual full base point. The intelligent recovery scheduler is used to parse the reconstructed metadata chain corresponding to the target time point when it receives a data recovery instruction, and determine the priority order of data recovery based on the correlation graph, and schedule system resources to prioritize the recovery of high-priority data.

[0008] Furthermore, the service-aware agent specifically includes a data relationship graph construction module and a runtime monitoring module; The data relationship graph construction module identifies and records key data entities that support the core business functions of the service through a software development kit embedded in the application, and maps the key data entities to the data block address range of physical storage; The runtime monitoring module continuously collects access requests to the key data entities, statistically analyzes access frequency and access time patterns, and dynamically updates the priority labels of data blocks in the association graph based on the statistical results. When the system performs a backup operation, the service-aware agent persistently stores the relationship graph and its priority tags as additional metadata along with the snapshot data.

[0009] Furthermore, the metadata reconstructor specifically includes a snapshot chain analysis module and a metadata merging module; The snapshot chain analysis module periodically scans the incremental snapshot chain to obtain the creation timestamp, data change amount, and historical access records of each snapshot, and filters out an early incremental snapshot set that is not frequently accessed based on a preset access popularity threshold. The metadata merging module iterates through the metadata of the data blocks referenced in the selected early incremental snapshot set and creates the virtual full base point. The virtual full base point integrates the pointer information of all data blocks in the early incremental snapshot set that have not been modified by subsequent snapshots. The metadata merging module is responsible for updating the metadata index so that the incremental snapshots after the virtual full base point point to the virtual full base point as the backtracking benchmark, and records the merging operation in the metadata log to ensure that the process is traceable.

[0010] Furthermore, the intelligent recovery scheduler specifically includes a recovery parsing module and a priority scheduling module; The recovery parsing module receives the data recovery instruction and the target recovery time point, locates and loads the reconstructed metadata chain corresponding to the time point, and the metadata chain includes the virtual full base point and the incremental snapshots thereafter up to the target time point; The priority scheduling module queries the relationship graph to identify key data blocks related to service startup and core functions, and marks the key data blocks as the highest recovery priority; The priority scheduling module issues differentiated input / output instructions to the storage subsystem based on the data block priority, controls the storage resources to prioritize the transmission and recovery of high-priority data blocks, and allows the upper-layer application service to send a service availability signal even when some low-priority data blocks are not yet ready during the recovery process.

[0011] Furthermore, the intelligent recovery scheduler also integrates a service preheating coordination module; The service preheating coordination module sends an instruction to the cloud platform management interface after the priority scheduling module starts restoring the highest priority data block to start the computing instance of the target business service in advance. The service preheating coordination module configures the memory loading logic of the computing instance so that after receiving the service availability signal sent by the intelligent recovery scheduler, it prioritizes loading the recovered key data blocks into memory and initializes the core business thread, so that the business service interface and basic functional framework can be presented to the user in advance. For service functions corresponding to data that has not yet been fully recovered, the service preheating coordination module controls the return of an intermediate state response during data loading.

[0012] On the other hand, a cloud computing server data backup and rapid recovery method, applicable to the aforementioned cloud computing server data backup and rapid recovery system, includes: Step 1: During application runtime, the service-aware agent collects and establishes the relationship graph between the core service functions and the underlying data blocks. Step 2: Monitor and analyze the metadata of the incremental snapshot chain through the metadata reconstructor, identify the set of early incremental snapshots that are not frequently accessed based on the snapshot access pattern, and merge the metadata of the set in the background to generate a virtual full base point. At the same time, reconstruct the metadata index relationship between subsequent incremental snapshots and the virtual full base point. Step 3: When the intelligent recovery scheduler receives a data recovery instruction, it parses the reconstructed metadata chain corresponding to the target time point, determines the priority order of data recovery based on the correlation graph, and schedules system resources to prioritize the recovery of high-priority data.

[0013] Furthermore, step two specifically includes: Regularly scan the incremental snapshot chain to obtain the creation timestamp, data change amount, and historical access records for each snapshot; Based on a preset access popularity threshold, multiple consecutive incremental snapshots that were created earlier and whose access frequency is lower than the threshold are identified as a set of early incremental snapshots that are not frequently accessed. Iterate through the metadata pointers of all data blocks in the early incremental snapshot set to identify the set of data blocks that have not been modified by any subsequent snapshots since the earliest snapshot in the set was created; Create a unified metadata view for the set of unmodified data blocks, namely the virtual full base point; Update the system metadata database to adjust the reference relationship of subsequent incremental snapshots to the original early snapshot chain to the reference relationship of the virtual full base point and subsequent incremental snapshots.

[0014] Furthermore, step three specifically includes: Receive recovery instructions containing the target recovery time point; Based on the target recovery time point, locate the corresponding virtual full base point and the required incremental snapshot sequence in the reconstructed metadata chain; Load the metadata information of the virtual full base point and incremental snapshot sequence; Query the aforementioned relationship graph, and divide the data blocks into at least two priorities based on the importance of the business functions associated with the data blocks and historical access patterns. Among them, the core metadata and key business data required for service startup are assigned the highest priority. Generate a recovery task queue containing data block priority tags and send the queue to the storage recovery engine to perform priority recovery.

[0015] Compared with existing technologies, this cloud computing server data backup and rapid recovery system and method have the following advantages: I. This invention continuously monitors and analyzes the incremental snapshot chain through a metadata reconstructor, identifies infrequently accessed early incremental snapshot sets and merges them to generate a virtual full base point, reconstructs the metadata index relationship between subsequent incremental snapshots and this base point, significantly shortening the snapshot chain length and reducing the overhead of metadata traversal during the recovery process; at the same time, it establishes a relationship graph between the core service functions and the underlying data blocks with the help of a service-aware proxy, and the intelligent recovery scheduler determines the data recovery priority based on this graph, prioritizing the scheduling of resources to restore key data, realizing rapid service recovery and seamless user experience, effectively ensuring business continuity.

[0016] Second, this invention dynamically collects data access information through the runtime monitoring module of the service-aware agent and updates the priority labels of data blocks in the relational graph in real time, so that data recovery always meets the actual business needs. The service preheating coordination module integrated in the intelligent recovery scheduler starts the computing instance in advance after the critical data recovery is started and configures the memory loading logic. After the core data recovery is completed, the basic service function framework can be presented to the user. Functions corresponding to data that has not been fully recovered return an intermediate state response, thereby improving the targeting and efficiency of data recovery and optimizing the user experience.

[0017] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0019] Figure 1 This is a schematic diagram of the system architecture and data flow of the present invention; Figure 2 This is a schematic diagram illustrating the metadata reconstruction and optimization of the present invention; Figure 3 This is a flowchart of the steps of the present invention. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] Example 1 like Figures 1 to 3 As shown, this embodiment aims to elaborate on the specific implementation of the cloud computing server data backup and rapid recovery system and method of the present invention, ensuring the sufficiency and feasibility of the disclosure of this technical solution. This embodiment uses an e-commerce business system deployed on a cloud server as an application scenario. In this scenario, business continuity requirements are high, and core business data includes user order information, basic product attributes, service configuration parameters, etc. The technical solution of the present invention is needed to achieve efficient data backup and rapid recovery, avoiding business interruption due to data loss or recovery delays.

[0022] In this embodiment, the service-aware proxy is a key component connecting the application layer and the storage layer. Its core function is to establish the association between the core functions of the service and the underlying data blocks, and to dynamically maintain the priority information of the data blocks.

[0023] Specifically, the service-aware proxy includes a data relationship graph construction module and a runtime monitoring module. The data relationship graph construction module is implemented through a software development kit (SDK) embedded in the e-commerce business system application. The SSD is integrated into the application's runtime environment as a dynamic link library, requiring no modification to the application's core business code; it only needs to be loaded during the application initialization phase. This SSD can intercept function calls supporting core business functions in the application and identify key data entities. For example, in the e-commerce business system, the order creation function corresponding to the user's order placement function, the product information retrieval function corresponding to the product query function, and the configuration loading function corresponding to service startup are all function calls related to core business functions.

[0024] The data relationship graph construction module analyzes these function calls to extract the relevant data entities, including key data entities such as order data, product data, and configuration data. Furthermore, this module interacts with the cloud server's storage management interface to obtain the address range of these key data entities in the physical storage medium. The mapping of data block address ranges is achieved through the logical address to physical address conversion service provided by the storage management interface, ensuring that each key data entity accurately corresponds to a specific set of data blocks in physical storage. Based on this, the data relationship graph construction module constructs a relational graph. This graph uses nodes to represent core service functions and data blocks, and edges to represent the relationships between them, clearly defining which data blocks support each core service function.

[0025] The runtime monitoring module continuously collects access requests to critical data entities during application execution. Access request collection is achieved by monitoring storage I / O operations; whenever the application performs a read or write operation on a critical data entity, the runtime monitoring module records relevant information about the access event. When statistically analyzing access frequency and access time patterns, a sliding time window statistical method is used. A fixed time window length is set, and the number of accesses to each data block within each time window is counted, while the time period of each access is recorded. Based on the statistical results, the priority labels of data blocks in the relationship graph are dynamically updated; data blocks with higher access frequency and those frequently accessed during peak core business periods have higher priority labels.

[0026] When the system performs a backup operation, the service-aware agent writes the relationship graph and its priority tags as additional metadata, along with the snapshot data, to the storage medium for persistent storage. The additional metadata uses the same storage format as the snapshot data, ensuring the integrity and consistency of the backup data and providing a priority determination basis for subsequent data recovery.

[0027] In this embodiment, the metadata reconstructor is used to optimize the structure of the incremental snapshot chain and reduce the overhead of traversing the snapshot chain during the recovery process. It includes a snapshot chain analysis module and a metadata merging module.

[0028] The snapshot chain analysis module periodically scans the incremental snapshot chain according to a preset cycle. The scanning process is achieved by traversing the snapshot index list in the cloud server storage system, which records basic information about all incremental snapshots. The snapshot chain analysis module retrieves the creation timestamp data change amount and historical access records for each snapshot from the index list. The historical access records are derived from the storage system's access logs, which record the number of times and the duration of each snapshot was accessed by restore or query operations over a past period.

[0029] The system filters out infrequently accessed early incremental snapshots based on a preset access frequency threshold. This threshold, set according to the system's storage strategy and historical access statistics, reflects the system's minimum requirement for snapshot access frequency. Specifically, the snapshot chain analysis module identifies multiple consecutive incremental snapshots created earlier and accessed less frequently than the access frequency threshold as an infrequently accessed early incremental snapshot set. Here, "multiple consecutive incremental snapshots" refers to snapshots arranged sequentially in the snapshot chain according to their creation time, ensuring that subsequent metadata merging operations can effectively integrate unmodified data block information.

[0030] The metadata merging module iterates through the metadata of the data blocks referenced in the selected set of early incremental snapshots. The traversal proceeds in chronological order of snapshot creation, reading the metadata information of each snapshot one by one. This metadata information includes the data block's identifier, pointer, and modification status. The traversal identifies the set of data blocks that have not been modified by any subsequent snapshots since the earliest snapshot in the set was created. This identification method compares the modification status of each data block across all snapshots in the early incremental snapshot set. If a data block exists in the earliest snapshot and has not been marked as modified in any subsequent snapshots, it is included in the set of unmodified data blocks.

[0031] A unified metadata view, or virtual full base point, is created for the set of unmodified data blocks. The virtual full base point does not physically merge all unmodified data blocks; instead, it creates a unified metadata index table that integrates pointer information for all unmodified data blocks. The pointer information for each data block directly points to its storage location in the physical storage medium, ensuring fast location during subsequent accesses.

[0032] The metadata merging module is responsible for updating the metadata index, adjusting the reference relationship between incremental snapshots after the virtual full base point and the original earlier snapshot chain to a reference relationship between the virtual full base point and subsequent incremental snapshots. The update process is implemented by modifying the snapshot reference records in the system metadata database, ensuring that subsequent data recovery operations can be performed based on the reconstructed metadata chain. Simultaneously, the metadata merging module records relevant information about the merging operation in the metadata log, including the execution time of the merging operation, the snapshot identifiers involved in the merging, and the identifier of the virtual full base point, ensuring the merging process is traceable and facilitating subsequent system maintenance and troubleshooting.

[0033] In this embodiment, the intelligent recovery scheduler is used to realize priority recovery of data and rapid preheating of services. It includes a recovery parsing module, a priority scheduling module, and a service preheating coordination module.

[0034] The recovery parsing module receives data recovery commands triggered by users or the system, which contain information about the target recovery time point. The module interacts with the system's time service to verify the accuracy of the target recovery time point. Subsequently, the module locates the corresponding virtual full base point and the required incremental snapshot sequence in the reconstructed metadata chain. The location process is based on the timestamp index in the metadata chain, comparing the target recovery time point with the creation timestamps of each virtual full base point and incremental snapshot to determine the virtual full base point and incremental snapshot sequence containing the data state of the target recovery time point. After location is complete, the recovery parsing module loads the metadata information of the virtual full base point and incremental snapshot sequence, providing a data foundation for subsequent recovery scheduling.

[0035] The priority scheduling module queries the relationship graph of the service-aware proxy persistent storage. The query process is achieved by accessing additional metadata in the storage medium to obtain the relationship between data blocks and core service functions, as well as the priority tags of the data blocks. Based on the importance of the business functions associated with the data blocks and historical access patterns, the data blocks are divided into at least two priority levels. Core metadata and critical business data necessary for service startup are assigned the highest priority. Core metadata includes application configuration parameters and database connection information, while critical business data includes basic information about user orders and core attributes of products. Data blocks related to non-core business functions and with low access frequency are assigned a lower priority, such as detailed log data of historical orders and redundant description information of products.

[0036] The priority scheduling module issues differentiated input / output commands to the storage subsystem based on data block priority. For the highest priority data blocks, high-priority input / output commands are issued, and the storage subsystem allocates more transmission bandwidth and I / O resources to ensure rapid transmission and recovery. For lower priority data blocks, normal-priority input / output commands are issued, and the storage subsystem allocates resources reasonably for recovery while meeting the recovery needs of high-priority data. Simultaneously, the priority scheduling module allows sending a service availability signal to the upper-layer application service even when some low-priority data blocks are not yet ready during the recovery process. This signal is sent when the highest-priority data blocks are fully recovered and can support the normal operation of core business functions, ensuring that the upper-layer application service can provide services to the outside world in a timely manner.

[0037] After the priority scheduling module begins restoring the highest priority data block, the service preheating coordination module immediately sends a command to the cloud platform management interface. This command contains information such as the target business service's identifier and computing resource requirements. Based on the command, the cloud platform management interface pre-starts the computing instance of the target business service. The startup process of the computing instance is executed according to the cloud platform's resource scheduling policy, allocating the corresponding computing resources, memory resources, etc., to ensure that the computing instance can start quickly and be in a standby state.

[0038] The service preheating coordination module configures the memory loading logic for compute instances. This memory loading logic is pre-written into the compute instance's startup configuration. When the compute instance receives a service availability signal from the intelligent recovery scheduler, it prioritizes loading recovered critical data blocks into memory according to this logic and initializes the core business threads. The initialization of the core business threads includes thread pool creation and loading of core business functions, enabling the business service interface and basic functional framework to be presented to the user in advance, reducing user waiting time.

[0039] For service functions corresponding to data that has not yet been fully restored, the service preheating coordination module controls the return of an intermediate status response indicating that data is loading. This intermediate status response is implemented by setting interception logic at the application service's interface layer. When a user requests access to a function corresponding to data that has not been fully restored, the interface layer intercepts the request and returns a preset intermediate status prompt, informing the user that the data is loading, thus improving the user experience.

[0040] The cloud computing server data backup and rapid recovery method in this embodiment is applicable to the above system, and the specific execution process is as follows: Step 1: Constructing the Relationship Graph. During the operation of the e-commerce business system application, the service-aware agent initiates and executes the relationship graph construction operation. The data relationship graph construction module intercepts the application's core business function calls through an integrated software development kit, identifies key data entities such as order data, product data, and configuration data, and obtains the physical storage data block address ranges corresponding to these key data entities through the storage management interface, constructing a relationship graph between the service's core functions and data blocks. The runtime monitoring module continuously collects access requests to key data entities, analyzes access frequency and access time patterns using a sliding time window statistical method, and dynamically updates the priority labels of data blocks in the relationship graph. When the system performs a backup operation, the service-aware agent persists the relationship graph and its priority labels as additional metadata, along with the snapshot data.

[0041] Step 2: Reconstruct the metadata chain. The metadata reconstructor periodically monitors and analyzes the incremental snapshot chain. The snapshot chain analysis module scans the incremental snapshot chain at a preset cycle, obtaining the creation timestamp data change volume and historical access records for each snapshot. Based on a preset access popularity threshold, it filters out a set of infrequently accessed early incremental snapshots. The metadata merging module traverses the metadata pointers of this early incremental snapshot set, identifies the set of data blocks that have not been modified by subsequent snapshots, creates a virtual full base point for this set, updates the system metadata database, and adjusts the reference relationship of subsequent incremental snapshots to reference the virtual full base point and subsequent incremental snapshots, thus completing the reconstruction of the metadata chain.

[0042] Step 3: Priority Recovery and Service Warm-up. Upon receiving a recovery instruction containing the target recovery time point, the intelligent recovery scheduler begins the recovery operation. The recovery parsing module locates and loads the reconstructed metadata chain corresponding to the target time point. This metadata chain includes the virtual full base point and the required incremental snapshot sequence. The priority scheduling module queries the correlation graph, divides data blocks into different priorities, generates a recovery task queue containing priority tags, and sends it to the storage recovery engine. The storage recovery engine executes the recovery operation according to the queue priority, prioritizing the transmission and recovery of the highest priority data blocks. After the highest priority data blocks begin recovery, the service warm-up coordination module sends an instruction to the cloud platform management interface to start the compute instance and configure the memory loading logic. Once the highest priority data blocks have been recovered, the intelligent recovery scheduler sends a service availability signal to the upper-layer application service. The compute instance loads the critical data blocks and initializes the core business threads, presenting the service interface and basic functional framework in advance. For functions corresponding to data that has not been fully recovered, an intermediate status response indicating that data loading is in progress is returned.

[0043] In some optional implementations, the access pattern analysis of the runtime monitoring module can incorporate a time weighting factor to further optimize the accuracy of data block priority labels. Specifically, when calculating access frequency, different time weights are assigned to access events in different time periods, with more recent access events receiving higher weights and more distant access events receiving lower weights. The time weighting factor is calculated using an exponential decay function, as shown in the following formula:

[0044] in Time weighting Attenuation coefficient The time difference between the current time and the time the access event occurred. It is a natural constant (approximately 2.718) used to ensure that the time weight decays smoothly and continuously over time and that the result always remains within a reasonable range of 0-1; A negative sign indicates that the attenuation coefficient will be negative. With time difference The product ( (This is a multiplication operator used to quantify the total attenuation of the two) is converted to a non-positive value to implement the logic that "the older the access event, the lower the weight"; the three work together to accurately match the business requirements of dynamically updating the priority of data blocks, and can be achieved through regular floating-point operations on cloud servers.

[0045] The access frequency of each data block is weighted by this time weighting factor to obtain a weighted access frequency. The priority label of the data block is then updated based on the weighted access frequency. This optimization method can more accurately reflect the importance of the data block in the current business scenario, making priority scheduling more aligned with actual business needs and further improving the targeting and efficiency of service recovery.

[0046] In some optional implementations, the metadata merging module of the metadata refactorer can dynamically adjust the merging timing based on system load to avoid impacting normal system operations. Specifically, the metadata merging module acquires system load metrics such as CPU utilization, memory usage, and storage I / O utilization in real time and sets a load threshold. When the system load metrics are below the load threshold, the metadata merging module performs a metadata merging operation on the early incremental snapshot set; when the system load metrics are higher than or equal to the load threshold, the metadata merging module delays the merging operation until the system load drops below the threshold.

[0047] This optimization method ensures that the metadata merging operation is carried out when system resources are sufficient, reduces the resource consumption of core business services, guarantees the stability and reliability of system operation, and does not affect the overall effect of metadata reconstruction.

[0048] In some optional implementations, the memory loading logic of the service preheating coordination module can introduce a preloading priority mechanism. Specifically, the service preheating coordination module performs a secondary priority division on the recovered key data blocks based on the priority tags of the data blocks in the association graph, dividing them into core loading data blocks and secondary core loading data blocks. After receiving the service availability signal, the compute instance first loads the core loading data blocks into memory and initializes the corresponding core business threads. After the core business functions are running stably, the secondary core loading data blocks are then loaded gradually.

[0049] This optimization method can further speed up the startup of core service functions, shorten the service availability window, improve the user's perception of service recovery, and at the same time, rationally allocate memory resources to avoid memory shortage problems caused by loading too much data at once.

[0050] This embodiment establishes the association between core service functions and data blocks and dynamically maintains priorities through a service-aware proxy, enabling data recovery to accurately match business needs. The metadata reconstructor reconstructs the metadata chain by creating a virtual full base point, effectively shortening the snapshot chain length, reducing the overhead of metadata traversal during recovery, and improving recovery startup speed. The intelligent recovery scheduler schedules resources based on priorities, prioritizing the recovery of critical data, and combines this with a service preheating mechanism, enabling services to quickly provide core functions after critical data recovery, shortening the service downtime window.

[0051] The implementation of the optimization examples further enhances the flexibility and adaptability of the technical solution, ensuring stable and efficient data backup and recovery under different system loads and business scenarios. The overall technical solution does not require significant modifications to the existing cloud server architecture, has good compatibility and feasibility, and can improve business continuity and service availability to a certain extent while ensuring data security.

[0052] Example 2 like Figures 1 to 3 As shown, this embodiment aims to further verify the universality and practicality of the technical solution of the present invention, using a cloud office collaboration system as an application scenario. In this scenario, the cloud server needs to support core businesses such as multi-user real-time document collaboration, video conferencing data storage, and account permission management. Core data includes user account information, document editing history, collaboration permission configuration data, and meeting recording file indexes. Such businesses have extremely high requirements for service continuity; data recovery delays may lead to problems such as collaboration interruption and loss of meeting records. The technical solution of the present invention can achieve efficient data backup and rapid recovery, ensuring seamless business continuity.

[0053] In this embodiment, the service-aware proxy serves as a connecting component between the application layer and the storage layer. Its core function is to establish a precise association between the core functions of the service and the underlying data blocks, and to dynamically maintain the priority of the data blocks. It includes a data relationship graph construction module and a runtime monitoring module.

[0054] Specifically, the data relationship graph construction module implements its functionality through a software development kit (SDK) embedded in the cloud-based office collaboration system application. This SSD offers multi-language versions, compatible with commonly used cloud office systems such as Java and Python. It adopts a plug-in integration approach, completing configuration loading during the application deployment phase without requiring modifications to the application's core business code. The SSD can intercept critical operations supporting core business functions within the application, such as user login account verification, real-time document saving data writing, collaboration permission allocation configuration updates, and meeting recording file index creation. By parsing these operations, it identifies key data entities.

[0055] Furthermore, the data relationship graph construction module obtains the physical storage data block address range corresponding to the aforementioned key data entities by calling the cloud server storage management interface. The storage management interface provides a standard logical address mapping service, capable of converting logical indexes such as user account IDs and unique document identifiers into the starting address and length information of data blocks in the physical storage medium, ensuring a one-to-one correspondence between each key data entity and a physical data block. Based on this, the data relationship graph construction module constructs a relational graph, where nodes represent core business functions and data blocks respectively, and edges represent the dependencies between them, clearly defining the data block support required for each core function.

[0056] During the operation of the cloud-based office collaboration system, the runtime monitoring module continuously listens for storage I / O requests and collects access records for key data entities. These access records include information such as access operation type, access time, and data block identifier. When statistically analyzing access frequency and time patterns, a fixed-period statistical method is used, setting each day as a statistical period. The number of accesses for each data block within the period is accumulated, and peak access times are recorded, such as 9:00-18:00 on weekdays. Based on the statistical results, the priority labels of data blocks in the relationship graph are dynamically updated. Key data blocks with high access frequency and concentrated during peak hours have higher priority labels.

[0057] When the system performs a backup operation, the service-aware agent writes the relationship graph and its priority tags as additional metadata, along with the snapshot data, to the storage system for persistent storage. The additional metadata and snapshot data use the same verification mechanism to ensure data integrity and provide a reliable basis for priority determination during subsequent recovery.

[0058] In this embodiment, the metadata reconstructor is used to optimize the incremental snapshot chain structure and reduce the metadata traversal overhead during the recovery process. It includes a snapshot chain analysis module and a metadata merging module.

[0059] The snapshot chain analysis module periodically scans the incremental snapshot chain according to a preset 24-hour cycle. The scanning process is achieved by traversing the snapshot management directory in the storage system, which records basic information about all incremental snapshots. The snapshot chain analysis module extracts the creation timestamp, data change amount, and historical access records for each snapshot from the directory. The historical access records are derived from the storage system's operation logs, which record in detail the access time, access type, and initiator of each snapshot.

[0060] The system filters out infrequently accessed early incremental snapshots based on a preset access frequency threshold. This threshold is set according to the business characteristics of the cloud office system; for example, fewer than three accesses within 30 consecutive days is considered insufficient. The snapshot chain analysis module identifies multiple consecutive incremental snapshots created more than 90 days ago and accessed at frequencies below the threshold as infrequently accessed early incremental snapshots, ensuring that the filtered snapshot set has merging value.

[0061] The metadata merging module iterates through the metadata of the data blocks referenced in the selected set of early incremental snapshots. The iteration proceeds in chronological order of snapshot creation, reading the metadata entries for each snapshot sequentially. These metadata entries contain information such as the data block identifier, modification status, and storage pointer. By comparing the modification status of the same data block across all snapshots in the set, the module identifies the set of data blocks that have not been modified by any subsequent snapshots since the earliest snapshot in the set was created.

[0062] A unified metadata view, or virtual full base point, is created for this set of unmodified data blocks. The virtual full base point is essentially an integrated metadata index file that summarizes the storage pointer information of all unmodified data blocks, directly pointing to the actual location of the data blocks in physical storage, without requiring physical migration of the data blocks themselves.

[0063] The metadata merging module then updates the system metadata database, adjusting the reference relationships of all incremental snapshots after the virtual full base point to the original earlier snapshot chain, to reference relationships to the virtual full base point and subsequent incremental snapshots. This update operation is implemented by modifying the snapshot association field in the metadata database, ensuring that subsequent recovery operations can be efficiently executed based on the reconstructed metadata chain. Simultaneously, the metadata merging module records detailed information about the merging operation in the metadata log, including the merging time, the snapshot identifiers involved in the merging, and the virtual full base point identifier, ensuring the merging process is traceable.

[0064] In this embodiment, the intelligent recovery scheduler is responsible for realizing data priority recovery and rapid service preheating, and includes a recovery parsing module, a priority scheduling module, and a service preheating coordination module.

[0065] The recovery parsing module receives data recovery commands triggered by the system administrator or automatically, which contain a specific target recovery time point. The module interacts with the cloud server's time synchronization service to calibrate the target recovery time point, ensuring time accuracy. Subsequently, the module locates the virtual full base point and incremental snapshot sequence corresponding to the target time point in the reconstructed metadata chain. This location is based on the timestamp information of each node in the metadata chain; by comparing the inclusion relationship between the target time point and the timestamps of each node, the required virtual full base point and incremental snapshots are determined. After location is complete, the recovery parsing module loads the corresponding metadata information into memory, providing data support for recovery scheduling.

[0066] The priority scheduling module queries the relationship graph of the service-aware proxy persistent storage to obtain the association information and priority tags between data blocks and core business functions. Based on the importance of the business functions associated with the data blocks and their historical access patterns, the data blocks are divided into three priority levels. Core metadata necessary for service startup (such as account authentication configuration and basic system parameters) and key business data (such as current collaborative document data and active user account information) are assigned the highest priority; data with recent access records but not necessary for real-time collaboration (such as document data not edited in the last 7 days) are assigned medium priority; and data that has not been accessed for a long time and is not core (such as the index of meeting recording files from six months ago) are assigned low priority.

[0067] Based on the aforementioned priority division, the priority scheduling module issues differentiated input / output instructions to the storage subsystem. For the highest priority data blocks, the instruction configures the highest I / O queue priority, and the storage subsystem allocates dedicated transmission bandwidth and read / write resources to ensure rapid transmission and recovery. For medium priority data blocks, the instruction configures a normal I / O priority, executing sequentially without affecting the recovery of the highest priority data. For low priority data blocks, the instruction configures a low I / O priority, utilizing idle system resources for recovery. Simultaneously, when the highest priority data block is fully recovered and meets the operational requirements of core business functions, the priority scheduling module sends a service availability signal to the upper-layer application service, without waiting for the recovery of medium and low priority data blocks to complete.

[0068] After the priority scheduling module initiates the recovery of the highest priority data block, the service preheating coordination module immediately sends a resource scheduling instruction to the cloud platform management interface. This instruction includes information such as the identifier of the target business service, the required computing resource specifications, and memory configuration requirements. Upon receiving the instruction, the cloud platform management interface quickly allocates computing nodes according to the resource scheduling strategy, starts up the computing instance of the target business service, and shortens the startup time of the computing instance.

[0069] The service preheating coordination module pre-configures the memory loading logic for the compute instances, which is embedded in the compute instance's startup script. When the compute instance receives a service availability signal, it loads the highest priority data blocks that have been recovered into memory according to the preset logic and initializes core business threads, including user authentication threads, document collaboration threads, and permission management threads. After the core business threads are initialized, the business service interface and basic functional framework can be presented to the user, allowing the user to log in, view current collaborative documents, configure basic permissions, and perform other operations normally.

[0070] For service functions corresponding to data that has not yet been fully restored, such as viewing historical documents that have not been accessed for a long time or playing early meeting recordings, the service preheating coordination module controls the interface response logic of the application service to return an intermediate status prompt indicating that the data required for the current function is being loaded, thereby avoiding user confusion due to waiting and improving the user experience.

[0071] The method in this embodiment is applicable to the above system, and the specific execution process is as follows: Step 1: Constructing the Relationship Graph. After the cloud-based office collaboration system starts, the service-aware agent runs automatically. The data relationship graph construction module, through an integrated software development kit, intercepts operations corresponding to core business functions, identifies key data entities such as user account information, document data, and permission configuration data, and obtains the physical data block address ranges corresponding to these data entities through the storage management interface to construct the relationship graph. The runtime monitoring module continuously listens for storage I / O requests, statistically analyzes the frequency and time periods of data block access, and dynamically updates the priority labels of data blocks in the graph. When the system performs scheduled or manual backups, the service-aware agent persists the relationship graph and priority labels as additional metadata along with the snapshot data.

[0072] Step two, reconstruct the metadata chain. The metadata reconstructor monitors and analyzes the incremental snapshot chain on a 24-hour cycle. The snapshot chain analysis module scans all incremental snapshots, collecting creation timestamps, data change amounts, and historical access records. Based on a preset access popularity threshold, it filters out a set of early incremental snapshots that are not frequently accessed. The metadata merging module traverses the metadata of this set, identifies the set of data blocks that have not been modified by subsequent snapshots, creates a virtual full base point, updates the system metadata database, adjusts the reference relationships of subsequent incremental snapshots, and completes the metadata chain reconstruction.

[0073] Step 3: Priority Recovery and Service Warm-up. Upon receiving a recovery instruction containing the target recovery time point, the recovery parsing module locates and loads the corresponding reconstructed metadata chain. The priority scheduling module queries the relationship graph, divides the data blocks into three priority levels, generates a recovery task queue with priority tags, and sends it to the storage recovery engine. The storage recovery engine executes recovery operations according to priority, processing the highest priority data blocks first. After the highest priority data block recovery starts, the service warm-up coordination module sends an instruction to the cloud platform management interface to start the compute instance and configure the memory loading logic. After the highest priority data block recovery is complete, the intelligent recovery scheduler sends a service availability signal, the compute instance loads the critical data blocks and initializes the core business threads, opening basic service functions to users. Functions corresponding to data that has not been fully recovered return an intermediate status response until the data loading is complete.

[0074] In some optional implementations, the service-aware proxy's software development kit (SDK) can add a non-intrusive adaptation mode, adapting to legacy cloud office applications that do not require source code modification. Specifically, the SSD provides process injection functionality, injecting the adaptation module into the target application's process space through the system's underlying interface, without modifying the application's source code or performing secondary compilation. The adaptation module intercepts the interaction interface between the application and the storage system using Hook technology, identifies the storage operations corresponding to core business functions, and then extracts key data entities and related data block information.

[0075] This optimization method expands the scope of service-aware proxy, making it applicable to legacy cloud office applications that cannot be modified from source code. It improves the compatibility and practicality of the technical solution, and both process injection and Hook technologies are existing mature technologies with a reliable implementation foundation.

[0076] In some optional implementations, the snapshot chain analysis module of the metadata refactorer can add a data change threshold filtering condition to further optimize the filtering accuracy of the early incremental snapshot set. Specifically, in addition to the access popularity threshold, a data change threshold is set, and only early incremental snapshots whose single snapshot data change amount is lower than this threshold are included in the set to be merged. The data change threshold is set according to the block size of the storage system and the frequency of business data updates to ensure that the snapshots included in the set are all snapshots with small change amounts and high merging value.

[0077] This optimization method avoids including early snapshots with large data changes in the merge set, reduces the computational overhead of the metadata merging process, and ensures the stability of the virtual full base point, thereby improving the efficiency and reliability of metadata chain reconstruction.

[0078] In some optional implementations, the priority scheduling module of the intelligent recovery scheduler can be enhanced with real-time access monitoring to dynamically adjust the recovery priority of data blocks. Specifically, during the data recovery process, the priority scheduling module monitors service access requests initiated by users in real time. If it detects that a medium-priority or low-priority data block is frequently requested by users, it immediately raises its priority to the highest level and prioritizes its recovery.

[0079] This optimization method can respond to users' real-time needs during the recovery process, avoid delays in the recovery of urgently needed data due to fixed priorities, further improve the flexibility of service recovery and user experience, and real-time access monitoring can be achieved by listening to the request logs of application services, which is a mature and easy-to-implement technology.

[0080] This embodiment verifies the application scenario of a cloud-based office collaboration system, demonstrating that the technical solution of this invention possesses good versatility and practicality. The service-aware proxy achieves precise association between core business functions and data blocks, ensuring that data recovery aligns with business needs; the metadata reconstructor shortens the snapshot chain length by creating a virtual full base point, significantly reducing metadata traversal overhead during the recovery process; the intelligent recovery scheduler's priority recovery mechanism and service preheating mechanism enable core business functions to become available quickly, greatly shortening the service unavailability window.

[0081] The optimized example further expands the applicability of the technical solution, improves operational efficiency and flexibility, and verifies the stability and availability of the solution under different application scenarios and system conditions. The technical implementation of this embodiment does not require significant modifications to the existing cloud server architecture, possesses good compatibility and feasibility, and can effectively improve the business continuity and service availability of the cloud office collaboration system while ensuring data backup security.

[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A cloud computing server data backup and rapid recovery system, characterized in that, This includes a service-aware agent, a metadata refactorer, and an intelligent recovery scheduler: The service-aware agent is used to collect and establish a graph of the relationship between the core functions of the service and the underlying data blocks during application operation. The metadata reconstructor is used to monitor and analyze the metadata of the incremental snapshot chain, identify the set of early incremental snapshots that are not frequently accessed based on the snapshot access pattern, and merge the metadata of the set in the background to generate a virtual full base point. At the same time, it reconstructs the metadata index relationship between subsequent incremental snapshots and the virtual full base point. The intelligent recovery scheduler is used to parse the reconstructed metadata chain corresponding to the target time point when it receives a data recovery instruction, and determine the priority order of data recovery based on the correlation graph, and schedule system resources to prioritize the recovery of high-priority data.

2. The cloud computing server data backup and rapid recovery system according to claim 1, characterized in that, The service-aware agent specifically includes a data relationship graph construction module and a runtime monitoring module; The data relationship graph construction module identifies and records key data entities that support the core business functions of the service through a software development kit embedded in the application, and maps the key data entities to the data block address range of physical storage; The runtime monitoring module continuously collects access requests to the key data entities, statistically analyzes access frequency and access time patterns, and dynamically updates the priority labels of data blocks in the association graph based on the statistical results. When the system performs a backup operation, the service-aware agent persistently stores the relationship graph and its priority tags as additional metadata along with the snapshot data.

3. The cloud computing server data backup and rapid recovery system according to claim 1, characterized in that, The metadata reconstructor specifically includes a snapshot chain analysis module and a metadata merging module; The snapshot chain analysis module periodically scans the incremental snapshot chain to obtain the creation timestamp, data change amount, and historical access records of each snapshot, and filters out an early incremental snapshot set that is not frequently accessed based on a preset access popularity threshold. The metadata merging module iterates through the metadata of the data blocks referenced in the selected early incremental snapshot set and creates the virtual full base point. The virtual full base point integrates the pointer information of all data blocks in the early incremental snapshot set that have not been modified by subsequent snapshots. The metadata merging module is responsible for updating the metadata index so that the incremental snapshots after the virtual full base point point to the virtual full base point as the backtracking benchmark, and the merging operation is recorded in the metadata log.

4. The cloud computing server data backup and rapid recovery system according to claim 1, characterized in that, The intelligent recovery scheduler specifically includes a recovery parsing module and a priority scheduling module; The recovery parsing module receives the data recovery instruction and the target recovery time point, locates and loads the reconstructed metadata chain corresponding to the time point, and the metadata chain includes the virtual full base point and the incremental snapshots thereafter up to the target time point; The priority scheduling module queries the relationship graph to identify key data blocks related to service startup and core functions, and marks the key data blocks as the highest recovery priority; The priority scheduling module issues differentiated input / output instructions to the storage subsystem based on the data block priority, controls the storage resources to prioritize the transmission and recovery of high-priority data blocks, and allows the upper-layer application service to send a service availability signal even when some low-priority data blocks are not yet ready during the recovery process.

5. The cloud computing server data backup and rapid recovery system according to claim 1, characterized in that, The intelligent recovery scheduler also integrates a service preheating coordination module; The service preheating coordination module sends an instruction to the cloud platform management interface after the priority scheduling module starts restoring the highest priority data block to start the computing instance of the target business service in advance. The service preheating coordination module configures the memory loading logic of the computing instance so that after receiving the service availability signal sent by the intelligent recovery scheduler, it prioritizes loading the recovered key data blocks into memory and initializes the core business thread, so that the business service interface and basic functional framework can be presented to the user in advance. For service functions corresponding to data that has not yet been fully recovered, the service preheating coordination module controls the return of an intermediate state response during data loading.

6. A method for data backup and rapid recovery of a cloud computing server, applicable to the cloud computing server data backup and rapid recovery system described in any one of claims 1-5, characterized in that, The method includes: Step 1: During application runtime, the service-aware agent collects and establishes the relationship graph between the core service functions and the underlying data blocks. Step 2: Monitor and analyze the metadata of the incremental snapshot chain through the metadata reconstructor, identify the set of early incremental snapshots that are not frequently accessed based on the snapshot access pattern, and merge the metadata of the set in the background to generate a virtual full base point. At the same time, reconstruct the metadata index relationship between subsequent incremental snapshots and the virtual full base point. Step 3: When the intelligent recovery scheduler receives a data recovery instruction, it parses the reconstructed metadata chain corresponding to the target time point, determines the priority order of data recovery based on the correlation graph, and schedules system resources to prioritize the recovery of high-priority data.

7. The cloud computing server data backup and rapid recovery system and method according to claim 6, characterized in that, Step two specifically includes: Regularly scan the incremental snapshot chain to obtain the creation timestamp, data change amount, and historical access records for each snapshot; Based on a preset access popularity threshold, multiple consecutive incremental snapshots that were created earlier and whose access frequency is lower than the threshold are identified as a set of early incremental snapshots that are not frequently accessed. Iterate through the metadata pointers of all data blocks in the early incremental snapshot set to identify the set of data blocks that have not been modified by any subsequent snapshots since the earliest snapshot in the set was created; Create a unified metadata view for the set of unmodified data blocks, namely the virtual full base point; Update the system metadata database to adjust the reference relationship of subsequent incremental snapshots to the original early snapshot chain to the reference relationship of the virtual full base point and subsequent incremental snapshots.

8. A cloud computing server data backup and rapid recovery system and method according to claim 6, characterized in that, Step three specifically includes: Receive recovery instructions containing the target recovery time point; Based on the target recovery time point, locate the corresponding virtual full base point and the required incremental snapshot sequence in the reconstructed metadata chain; Load the metadata information of the virtual full base point and incremental snapshot sequence; Query the aforementioned relationship graph, and divide the data blocks into at least two priorities based on the importance of the business functions associated with the data blocks and historical access patterns. Among them, the core metadata and key business data required for service startup are assigned the highest priority. Generate a recovery task queue containing data block priority tags and send the queue to the storage recovery engine to perform priority recovery.

Images