A heterogeneous cloud platform resource management device and method based on a hybrid synchronization mechanism

Abstract

This invention belongs to the field of cloud computing and multi-cloud management technology, and discloses a heterogeneous cloud platform resource management device and method based on a hybrid synchronization mechanism. Addressing the problems of large synchronization latency, high interface pressure, data inconsistency, and tight coupling in existing centralized polling schemes, this invention achieves unified resource management by deploying a proxy middleware layer between the multi-cloud management platform and each cloud platform, and employing a hybrid synchronization mechanism combining event triggering and periodic comparison. The cloud platform proxy module converts raw data into a unified model and maintains a local cache; the event triggering unit monitors internal operations and pushes changes in real time; and the periodic comparison unit periodically identifies differences and synchronizes them through hash comparison. This invention revolutionizes the traditional polling and retrieval mode, achieving near real-time synchronization of resource status; significantly reducing API access pressure and improving data consistency; and, through modular decoupling design, greatly reducing the complexity of system integration and maintenance, and improving scalability and stability.

Description

Technical Field

[0001] This invention relates to the field of cloud computing and multi-cloud management technology, and in particular to a heterogeneous cloud platform resource management device and method based on a hybrid synchronization mechanism. Background Technology

[0002] With the rapid evolution of cloud computing technology, the cross-deployment of public, private, and hybrid clouds has become the norm, and business systems increasingly rely on heterogeneous cloud platforms to collaboratively provide computing power and resources. This hybrid model has created an urgent need for unified, agile, and efficient management and near real-time monitoring of multi-cloud environments.

[0003] Currently, multi-cloud management primarily relies on centralized cloud management platforms, which periodically poll various cloud platforms via APIs to obtain resource data. This mechanism has inherent drawbacks: First, data real-time performance is poor, and resource changes within the polling interval cannot be detected in a timely manner. Second, shortening the polling cycle in pursuit of timeliness significantly increases the call pressure on cloud platform interfaces, consuming substantial network and computing resources. Third, for resource changes triggered by "bypass operations" such as cloud platform consoles and CLIs, the platform cannot proactively capture them, relying solely on subsequent polling for passive discovery, leading to a high risk of data inconsistency and making it difficult to strike a balance between real-time performance, system overhead, and consistency. Furthermore, the aforementioned centralized polling architecture requires tight interface coupling between the cloud management platform and each heterogeneous cloud platform to be managed. Whenever a new cloud platform is connected or a cloud platform API is upgraded, extensive adaptation development is required on the operations and management platform side, resulting in a large workload and long cycle. Simultaneously, adjustments to the interface of one cloud platform may unexpectedly affect the stability of other connected platforms, making the system difficult to flexibly expand and maintain.

[0004] Even in existing improved solutions, such as the multi-level resource allocation method proposed in publication CN118245235A, the synchronization mechanism is still based on the traditional timed polling and full or incremental fetching mode. This fails to effectively solve the problems of interface coupling, large development workload, and difficulty in controlling the impact surface, and also fails to fundamentally resolve the contradiction between real-time performance, overhead, and consistency. Changes on the cloud platform side cannot be perceived in real time, and the inconsistencies caused by continuous query pressure and bypass operations persist. Therefore, current technology has not yet effectively overcome the core challenge of balancing real-time performance, low overhead, and strong consistency in resource state synchronization in a multi-cloud environment. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, the present invention aims to provide a resource management device and method that interfaces with multiple cloud platforms. By introducing a decoupled proxy middleware layer between the multi-cloud management platform and the cloud platform, and triggering intelligent synchronization based on resource status changes, the near real-time synchronization and agile management of resource status in a multi-cloud environment are achieved. This reduces system coupling from an architectural perspective and fundamentally reduces the interface pressure and resource consumption caused by traditional periodic polling methods.

[0006] To achieve the above objectives, the first aspect of the present invention provides a resource management device for interfacing with multiple heterogeneous cloud platforms, comprising: a multi-cloud management agent module deployed on the side of a multi-cloud management platform, and a cloud platform agent module deployed on at least one heterogeneous cloud platform.

[0007] The multi-cloud management agent module is communicatively connected to multiple cloud platform agent modules and includes a first memory storage unit for caching integrated resource information;

[0008] The cloud platform proxy module includes an interface adaptation unit and a second memory storage unit. The interface adaptation unit is used to convert the resource data returned by the native API of the corresponding cloud platform into a unified data model, and the second memory storage unit is used to cache the converted unified data model.

[0009] The multi-cloud management agent module is configured to: in response to a resource query request from the multi-cloud management platform, prioritize querying and returning resource information from the first memory storage unit;

[0010] The cloud platform agent module is configured to synchronize resource status with the multi-cloud management agent module through a hybrid synchronization mechanism.

[0011] The hybrid synchronization mechanism includes: a first synchronization method based on resource operation events; and a second synchronization method based on periodic difference detection.

[0012] The first synchronization method is implemented by an event-triggered synchronization unit. The event-triggered synchronization unit is used to listen for resource operation event notifications generated by the cloud platform, update the second memory storage unit when it receives the event notification, and push the resource change information to the multi-cloud management agent module.

[0013] The second synchronization method is implemented by a timed comparison and synchronization unit. The timed comparison and synchronization unit is used to periodically obtain the current cloud platform resource information through the interface adaptation unit, compare it with the historical resource information stored in the second memory storage unit to identify differences, and synchronize the difference resource information to the multi-cloud management agent module.

[0014] Furthermore, the event-triggered synchronization unit specifically includes:

[0015] The event listening subunit is used to subscribe to the resource operation audit log stream or change notification service provided by the cloud platform.

[0016] The parsing and updating subunit is used to parse the listened event messages, obtain the operation type and target resource identifier, and call the interface adaptation unit to obtain the latest unified data model of the resource to update the second memory storage unit.

[0017] The event push subunit is used to generate change events containing a unique resource identifier, change type, and timestamp, and push them to the multi-cloud management agent module in real time via a secure communication link.

[0018] Furthermore, the timing comparison and synchronization unit specifically includes:

[0019] The hash calculation subunit is used to calculate the hash value of the unique identifier and key attribute set of the resource based on the resource type, and generate the current resource hash list;

[0020] The comparison and analysis subunit is used to compare the current resource hash list with the historical resource hash list stored in the second memory storage unit, and determine whether the resource has changed based on the consistency of the hash values.

[0021] The difference synchronization subunit is used to obtain the complete unified data model of the changed resource, update the second memory storage unit, and synchronize the change information to the multi-cloud management agent module.

[0022] Furthermore, both the first memory storage unit and the second memory storage unit use a classified key-value pair structure to organize resource data, wherein: the key is a globally unique identifier for the resource, and the value is a structured data object containing resource specifications, status, tags and associations; the resource data is partitioned and stored according to computing, storage and network types.

[0023] Furthermore, it also includes a secure communication and control module, which includes:

[0024] The security authentication submodule is used to establish an encrypted communication channel based on a two-way TLS or token mechanism between the multi-cloud management agent module and the cloud platform agent module, and to perform two-way identity authentication.

[0025] The access control submodule is used to control the scope of operations that each agent module can perform on cloud platform resources based on pre-configured role-based access control policies.

[0026] Furthermore, it also includes a fault tolerance and state recovery module, which includes:

[0027] The message assurance submodule is used to automatically restore the connection and ensure reliable message transmission when communication between the multi-cloud management agent module and the cloud platform agent module is abnormal.

[0028] The status checkpoint submodule is used to generate and maintain consistency checkpoints. These consistency checkpoints are used to trigger an incremental data recovery process based on the checkpoint information after the agent module restarts, in order to ensure eventual data consistency. The information recorded by the consistency checkpoints includes one or more of the following: the last confirmed change version number, timestamp, and offset of pending messages.

[0029] Furthermore, it also includes a resource dependency management module, used for:

[0030] Use a directed acyclic graph model to describe and manage creation, startup, or network connection dependencies between resources on different cloud platforms;

[0031] When performing batch resource synchronization or operations, topological sorting is performed according to the directed acyclic graph to determine the synchronization or operation order of resources, avoiding errors caused by unmet dependencies.

[0032] Furthermore, the second memory storage unit supports tiered storage strategies and resource sharding synchronization strategies:

[0033] The hierarchical storage strategy is as follows: frequently accessed hot resource data is stored in memory, while infrequently accessed cold resource data is stored on disk or in an external cache database.

[0034] The resource sharding synchronization strategy is as follows: cloud platform resources are logically sharded according to the hash value of the resource's unique identifier. During the timed comparison and synchronization process, only shards whose resource status has changed are subject to full comparison and synchronization.

[0035] Secondly, the present invention provides a heterogeneous cloud platform resource management method based on a hybrid synchronization mechanism, comprising the following steps:

[0036] S1: The multi-cloud management platform initiates a resource query request;

[0037] S2: The multi-cloud management agent module receives the request and first queries its first memory storage unit; if a match is found, the resource information is returned directly; if a match is not found, the request is forwarded to the target cloud platform agent module.

[0038] S3: The cloud platform agent module maintains the data timeliness of its second memory storage unit through a hybrid synchronization mechanism and keeps synchronized with the multi-cloud management agent module; the execution of the hybrid synchronization mechanism includes:

[0039] S31: Event-triggered synchronization steps: When the cloud platform agent module listens to a resource operation event generated within its cloud platform, it triggers real-time updates and pushes of resource information.

[0040] S32: Periodic comparison and synchronization step: The cloud platform agent module periodically and actively queries cloud platform resources, identifies differences by comparing newly acquired resource information with historical information stored in the second memory storage unit, and performs synchronization.

[0041] Further, step S31 specifically includes:

[0042] The cloud platform proxy module receives and parses resource operation requests or event logs forwarded by the cloud platform's internal management components;

[0043] Update the resource status in the local second memory storage unit based on the parsing results;

[0044] Generate a resource change event and immediately push it to the multi-cloud management agent module through the event channel;

[0045] The comparison process in step S32 uses a hash comparison method, which calculates and compares the hash values ​​of key resource attributes to efficiently determine whether the resource has changed.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1. This invention optimizes the traditional fixed-period polling-based fetching mode into a hybrid synchronization mode that primarily uses event-triggered real-time push and supplements it with timed comparison. This significantly reduces the latency of resource status synchronization. By prioritizing cached responses and using differential synchronization, the frequency of direct calls to the cloud platform's native management interfaces is significantly reduced, lowering system overhead and external interface pressure.

[0048] 2. By deploying a proxy module on the cloud platform side to listen for and preprocess events locally, and combining it with the global cache view of the multi-cloud management terminal, this invention can directly respond to a large number of queries at the local cache level, avoiding frequent penetration to the cloud platform's native API, thereby alleviating the cloud platform's interface load and shortening the query response time.

[0049] 3. The event-triggered mechanism can instantly collect resource changes occurring within the cloud platform, including those initiated through the console, command line, or other means, and combine them with a periodic comparison mechanism as a compensatory verification. These two mechanisms complement each other, reducing the risk of data inconsistency due to synchronization delays or omissions, thereby improving the reliability of the multi-cloud resource view.

[0050] 4. By employing in-memory key-value pair caching, hash comparison based on key attribute summaries for rapid difference detection, and optimization strategies such as hierarchical resource storage and sharding synchronization, this invention improves the efficiency of data access, comparison, and synchronization. This solution abstracts the differences between cloud platforms, enabling the system to flexibly expand to support unified management of large-scale, multi-type heterogeneous cloud resources, exhibiting excellent scalability and adaptability.

[0051] 5. By introducing a proxy middleware layer and a unified data model between the multi-cloud management platform and various cloud platforms, modular design and decoupling between systems are achieved. This architecture greatly reduces the development workload when connecting to new cloud platforms or adapting to API upgrades, allowing each module to evolve independently. At the same time, the isolation design of the proxy layer effectively prevents the impact on the stability of the global system caused by changes in the interface of a single cloud platform, thereby improving the maintainability, scalability, and operational stability of the overall system. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the system architecture of the resource management device for interfacing with multiple cloud platforms according to the present invention;

[0053] Figure 2 This is a schematic diagram illustrating the interaction between the multi-cloud management platform side and the cloud platform side proxy module in this invention;

[0054] Figure 3 This is a flowchart illustrating the resource management method of the present invention. Detailed Implementation

[0055] The specific embodiments of the present invention will be described in detail below with reference to the claims and technical disclosure. It should be understood that the specific embodiments described are only for illustrating the technical concept and implementation principle of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art can make various equivalent modifications or substitutions to the specific implementations without departing from the spirit of the present invention.

[0056] like Figure 1-3 As shown, in one specific embodiment, the heterogeneous cloud platform resource management device based on a hybrid synchronization mechanism of the present invention includes a multi-cloud management agent module deployed on the multi-cloud management platform side, and cloud platform agent modules deployed on at least two heterogeneous cloud platforms respectively. The heterogeneous cloud platform refers to a cloud environment where the underlying resource model, management interface, or communication protocol differs.

[0057] The multi-cloud management agent module establishes a secure communication channel with the agent modules of each cloud platform, and the two together constitute a proxy middleware layer for multi-cloud resource synchronization. The multi-cloud management agent module has a first memory storage unit to cache integrated resource information from multiple cloud platforms, forming a global resource cache view; each cloud platform agent module has a second memory storage unit to cache the current resource status information of the corresponding cloud platform.

[0058] During system initialization, each agent module establishes a secure connection and performs identity authentication. The secure communication channel can be implemented using a security mechanism based on encrypted transmission and two-way identity verification. Specifically, the secure communication and management module is configured to execute the following process: First, an encrypted channel is established based on the two-way TLS (Transport Layer Security) protocol. Each agent module is injected with a unique X.509 certificate issued by a private Certificate Authority (CA) during deployment. During the handshake phase, both parties exchange and verify the certificates to ensure the legitimate identity of the communication endpoints. Second, as a supplement or alternative, the module supports a token authentication mechanism. The cloud platform agent module automatically obtains a short-lived access token using its hosted identity on the cloud platform (such as AWSIAMRole) and includes it in the Authorization header of all requests. The authentication service embedded in the multi-cloud management agent module verifies the validity and scope of this token. Finally, the module loads a pre-configured role and permission policy file (such as in YAML format). For example, the cloud platform agent module is assigned the Reporter role, whose permission policy stipulates that it can only "push resource change events"; while the multi-cloud management agent module is assigned the Orchestrator role, allowing it to "query resources" and "issue operation instructions". All cross-agent requests must pass the authentication check of the policy engine before routing. After the cloud platform agent module is authenticated by its host cloud platform, it registers its supported resource types and capability information with the multi-cloud management agent module, thereby completing system initialization.

[0059] The cloud platform proxy module is deployed on the cloud platform side to obtain the status of cloud platform resources, maintain its secondary memory storage unit, and trigger synchronization. The cloud platform proxy module includes at least an interface adaptation unit and a secondary memory storage unit.

[0060] The interface adaptation unit is used to interface with the cloud platform's native management interface and convert the acquired resource data into a unified data model. Different cloud platforms differ in interface protocols, parameter definitions, and returned data structures. The interface adaptation unit uses preset adaptation rules to mask these differences, allowing the upper-layer synchronization logic to be implemented independently of the specific cloud platform.

[0061] In one implementation, the interface adaptation unit can obtain resource information based on the management interface or development tools provided by the cloud platform and convert the native data into an internally unified resource data object. The unified data model is a structured data object that includes at least fields such as a unique resource identifier, resource type, cloud platform identifier, operating status, resource specification information, and resource dependencies. For example, the unified data model can be defined as a structured data object whose field set covers the common core attributes in resource descriptions across different cloud platforms, thereby providing a consistent access interface for the upper layer.

[0062] The second memory storage unit is used to store resource status information on the cloud platform side, and its internal management can adopt a categorized key-value mapping structure. Specifically, both the first and second memory storage units use a key-value pair structure. The key is a globally unique identifier for the resource, and an exemplary generation rule is: {cloud vendor}:{account ID}:{region}:{resource type}:{resource native ID}, for example, aws:123456789012:us-east-1:ec2:i-0a1b2c3d4e5f67890. The value is a structured data object corresponding to the unified data model. Taking JSON format as an example, it contains core fields such as id, name, type, provider, status, spec (resource specification), tags, createdAt, and relationships (a list of associated resource IDs). When storing resource data, logical partitions are made according to types such as computing (e.g., EC2, VM), storage (e.g., EBS, disk), and network (e.g., VPC, SG). For example, they are managed in different in-memory hash tables or database partitions to improve query efficiency.

[0063] The cloud platform agent module maintains the timeliness of data in the second memory storage unit through a hybrid synchronization mechanism and synchronizes resource change information to the multi-cloud management agent module. This hybrid synchronization mechanism includes event-triggered synchronization and timed comparison synchronization.

[0064] In one implementation, the cloud platform proxy module listens for resource management operation events within the cloud platform to obtain operation information such as resource creation, deletion, or attribute changes. When a resource operation event is detected, the cloud platform proxy module calls the interface adaptation unit to obtain the current status of the corresponding resource and updates the second memory storage unit.

[0065] One alternative implementation is that the cloud platform agent module subscribes to event message streams (such as operation audit log streams or resource change notification topics) published by the cloud platform service to listen for relevant events.

[0066] As a preferred implementation of the event-triggered synchronization of this invention, the monitoring function needs to be adapted to the architecture of different cloud platforms. In one specific embodiment:

[0067] For Amazon AWS cloud that supports CloudTrail, the cloud platform agent module is configured to create an AmazonSQS message queue to receive the operation audit log stream delivered by CloudTrail and filter out key resource operation events such as RunInstances, TerminateInstances, and CreateVolume.

[0068] For OpenStack-based private clouds, the cloud platform proxy module subscribes to its internal oslo.messaging message bus to listen for notification events such as compute.instance.create.end and volume.create.end issued by services such as Nova, Cinder, and Neutron.

[0069] For Alibaba Cloud, the cloud platform agent module delivers logs to the Log Service (SLS) through the configuration of ActionTrail and deploys the Logtail client to collect logs in real time.

[0070] Upon successfully capturing the aforementioned events, the cloud platform agent module extracts the unique identifier of the target resource (e.g., AWS instance IDi-xxxxx) and immediately invokes the interface adaptation unit. Using the cloud platform's native API (e.g., AWS EC2's DescribeInstances), it obtains the latest and complete status information of the resource, converts it into a unified data model, and updates it to the second in-memory storage unit. In one example implementation, a synchronization completion time of approximately 1–3 seconds was observed in most test scenarios; this latency is affected by factors such as the cloud platform's event delivery mechanism, network conditions, and system load.

[0071] After completing the local update, the cloud platform agent module generates a resource change message, which includes at least the resource identifier, change type, and time information. The message is then sent to the multi-cloud management agent module through a secure communication channel, thereby achieving near real-time synchronization of resource status changes.

[0072] In another implementation, the cloud platform agent module also periodically executes a timed comparison and synchronization process to capture resource changes not detected by the event-triggered mechanism. The timed comparison and synchronization is initiated at a preset time interval, which is a configurable parameter. This process is specifically executed by the timed comparison and synchronization unit, whose internal logic is divided into:

[0073] Hash Calculation Subunit: Periodically calls the interface adaptation unit to obtain the current full resource list of the cloud platform; for each resource, according to the predefined key attribute set of its type (such as virtual machine specifications, status, internal network IP, tags, etc.), sorts these attribute values ​​and concatenates them into a string, and uses the SHA-256 algorithm to calculate the hash value to generate the current resource hash list;

[0074] The comparison and analysis subunit reads the previously saved historical resource hash list from the metadata area of ​​the second memory storage unit and compares the hash values ​​of the same resource ID item by item. If they do not match, the resource ID is marked as changed.

[0075] Differential Synchronization Subunit: For all resources marked as changed, the interface adaptation unit is called to obtain their complete unified data model, the second memory storage unit is updated, and a change event is generated and pushed to the multi-cloud management agent module.

[0076] During the periodic comparison process, the cloud platform agent module obtains the current resource status list of the cloud platform and generates status summary information based on the key attributes of the resources. Subsequently, the status summary information is compared with the historical summary information stored in the second memory storage unit; when an inconsistency is detected, it is determined that the corresponding resource has changed, and the second memory storage unit is updated, while a synchronization message is generated and sent to the multi-cloud management agent module.

[0077] As an efficient and resource-saving implementation, the state summary information can be generated by calculating hash values ​​from key resource attributes. The specific hash comparison strategy is as follows:

[0078] Instead of calculating all resource attributes, the system extracts a set of key metadata fields to generate a signature. First, a set of key metadata fields is predefined for each resource type (e.g., for virtual machines, this includes instance_type, image_id, state, private_ip, security_group_ids, and all tags). Then, the values ​​of these fields are arranged lexicographically and concatenated into a string, and a hash value is generated using a cryptographic hash function (such as SHA-256). During each periodic query, the system generates or updates this list of hash values ​​only for the resource.

[0079] During the comparison, only the new and old hash values ​​corresponding to the same resource identifier need to be compared. If they are equal, it is determined that no critical changes have occurred to the resource; if they are not equal, it is marked as changed, thereby triggering full data acquisition and synchronization. Through the above hash comparison method, the system reduces the computational complexity of resource consistency comparison from a product of the number of resources and the number of attributes to a linear level that is only related to the number of resources, greatly reducing data processing and transmission overhead.

[0080] In one embodiment, the apparatus of the present invention further includes a fault tolerance and state recovery module. This module is specifically implemented as follows:

[0081] The message assurance submodule operates when communication between the multi-cloud management agent module and the cloud platform agent module fails. This submodule maintains a persistent sending queue on the sending end (e.g., the cloud platform agent module). Each change message to be pushed is first stored in this queue and marked as "sending" before being sent. Only after receiving an application-layer acknowledgment (ACK) from the receiving end (multi-cloud management agent module) is the message marked as "acknowledged" and removed from the queue. If the network connection is lost or an ACK is not received within a timeout period, the submodule will attempt to reconnect and automatically resend all messages in the "sending" state after the connection is restored. The receiving end uses message IDs for deduplication to ensure at least one reliable delivery.

[0082] State Checkpoint Submodule: This submodule periodically records consistency checkpoints during the synchronization process. A checkpoint is a lightweight data snapshot, including but not limited to: the last confirmed version number of the change synchronized to the multi-cloud management terminal (a monotonically increasing sequence), the timestamp corresponding to the change, and the offset of pending messages in the local persistent queue. Checkpoints are periodically persisted to disk. When the agent module restarts abnormally, this submodule first loads the latest checkpoint, and then, based on the version number or timestamp, requests incremental resource change data that occurred after that checkpoint only from the cloud platform API. Simultaneously, it resumes message processing from the corresponding offset in the message queue, thereby quickly and accurately completing incremental data recovery and ensuring eventual consistency of system data.

[0083] In one embodiment, the invention further includes a resource dependency management module. This module uses a directed acyclic graph (DAG) model to describe and manage complex dependencies between resources. Specifically, the module abstracts each cloud resource as a vertex in the graph and the dependencies between resources (such as "the creation of virtual machine A depends on the existence of subnet B") as directed edges. For example, in an application scenario requiring the creation of virtual machines (VMs), cloud disks (Volumes), and virtual private clouds (VPCs), the module constructs a DAG containing two creation dependency edges: "VPC->VM" and "Volume->VM". During batch resource synchronization or orchestration operations, the module executes a topology sorting algorithm (such as Kahn's algorithm) on this DAG to obtain a linear sequence of operations that satisfies all dependencies, such as: [Create VPC, Create Volume, Create VM]. The system strictly follows this sequence to perform operations, thereby avoiding failures due to unavailable dependent resources. For deletion operations, the topology sequence is executed in reverse order.

[0084] In a further implementation, the second memory storage unit supports tiered storage and resource sharding synchronization. The specific strategy is as follows:

[0085] Tiered storage strategy: The system monitors resource access frequency in real time. For hot resources (such as running virtual machines) that are frequently queried or modified within the recent time window, their complete data models are always kept in memory to ensure millisecond-level read speeds. For cold resources that have not been accessed for a long time (such as instance metadata that has been shut down for weeks), their data models are swapped to disk files or external cache databases (such as Redis), with only their indexes and key hash values ​​retained in memory, thereby freeing up memory space.

[0086] Resource Sharding Synchronization Strategy: The cloud platform agent module logically divides all resources into a fixed number of shards, such as 1024, based on the hash value of the resource's unique identifier (e.g., CRC32 calculation of the ID). Each shard is associated with a "dirty flag" bit. When an event triggers the synchronization mechanism to update a resource, the "dirty flag" of its shard is set. In subsequent timed comparison and synchronization cycles, the system only performs a full resource list retrieval and hash comparison operation on shards marked as "dirty." For "clean" shards, the full comparison process is skipped. This strategy concentrates the comparison overhead of large-scale resources on a small number of shards that are actually likely to change, thereby significantly reducing the computational and I / O load of the synchronization process.

[0087] The multi-cloud management agent module is deployed on the multi-cloud management platform side and is used to uniformly receive, cache, and respond to resource data from multiple cloud platforms. The multi-cloud management agent module includes at least a first memory storage unit and request processing logic.

[0088] The first memory storage unit is used to store integrated multi-cloud resource information. It can use a hierarchical index key-value mapping structure to support the rapid location of different cloud platforms and resources, thereby building a global resource cache view.

[0089] When the multi-cloud management platform initiates a resource access request, the multi-cloud management agent module first obtains the required resource information from the first memory storage unit; when the corresponding resource does not exist in the cache or the cached data is marked as invalid, it then interacts with the corresponding cloud platform agent module to obtain the latest data, and updates the first memory storage unit after returning.

[0090] The multi-cloud management agent module continuously receives resource change messages from agent modules of various cloud platforms, and updates the resource data in the first memory storage unit according to the resource identification information in the message, thereby ensuring the real-time or near real-time performance of the global resource view.

[0091] Based on the above-mentioned device, the resource synchronization method of the present invention includes: system initialization and establishment of a secure connection; the cloud platform agent module obtains and synchronizes resource changes based on a hybrid synchronization mechanism; the multi-cloud management agent module responds to resource access requests based on a global resource cache view, and interacts with the cloud platform agent module to obtain the latest resource data when necessary; and performs fault tolerance recovery and data compensation in the event of communication abnormalities or node abnormalities.

[0092] The above method steps correspond to the aforementioned device implementation in terms of technical logic, and will not be described again here.

[0093] The beneficial effects of this invention lie in its ability to effectively overcome the inherent data latency and high invalid query pressure of traditional fixed-time-interval polling synchronization methods through the aforementioned proxy intermediate layer and hybrid synchronization mechanism. Specifically, the synchronization latency for resource status changes is reduced from minutes in traditional polling to seconds in event-driven methods; by prioritizing cached responses and differential synchronization, the frequency of direct calls to the cloud platform's native management interfaces and the amount of network data transmission are significantly reduced, thereby improving overall system efficiency and scalability. Simultaneously, the modular decoupling design and detailed fault tolerance and dependency management mechanisms greatly reduce the complexity of system integration and maintenance, improving overall stability and maintainability.

[0094] In summary, this invention effectively improves the timeliness and consistency of resource status synchronization in a multi-cloud environment by introducing a proxy intermediary layer between the multi-cloud management platform and the cloud platform, and combining it with a hybrid synchronization mechanism based on resource status changes. Any equivalent modifications or substitutions made to the above embodiments by those skilled in the art without departing from the technical concept of this invention should fall within the protection scope of this invention.

Claims

1. A heterogeneous cloud platform resource management device based on a hybrid synchronization mechanism, characterized in that, include: A multi-cloud management agent module deployed on a multi-cloud management platform, and a cloud platform agent module deployed on at least one heterogeneous cloud platform; The multi-cloud management agent module is communicatively connected to multiple cloud platform agent modules and includes a first memory storage unit for caching integrated resource information; The cloud platform proxy module includes an interface adaptation unit and a second memory storage unit. The interface adaptation unit is used to convert the resource data returned by the native API of the corresponding cloud platform into a unified data model, and the second memory storage unit is used to cache the converted unified data model. The multi-cloud management agent module is configured to: in response to a resource query request from the multi-cloud management platform, prioritize querying and returning resource information from the first memory storage unit; The cloud platform agent module is configured to synchronize resource status with the multi-cloud management agent module through a hybrid synchronization mechanism. The hybrid synchronization mechanism includes: a first synchronization method based on resource operation events; and a second synchronization method based on periodic difference detection. The first synchronization method is implemented by an event-triggered synchronization unit. The event-triggered synchronization unit is used to listen for resource operation event notifications generated by the cloud platform, update the second memory storage unit when it receives the event notification, and push the resource change information to the multi-cloud management agent module. The second synchronization method is implemented by a timed comparison and synchronization unit. The timed comparison and synchronization unit is used to periodically obtain the current cloud platform resource information through the interface adaptation unit, compare it with the historical resource information stored in the second memory storage unit to identify differences, and synchronize the difference resource information to the multi-cloud management agent module.

2. The resource management device according to claim 1, characterized in that, The event-triggered synchronization unit specifically includes: The event listening subunit is used to subscribe to the resource operation audit log stream or change notification service provided by the cloud platform. The parsing and updating subunit is used to parse the listened event messages, obtain the operation type and target resource identifier, and call the interface adaptation unit to obtain the latest unified data model of the resource to update the second memory storage unit. The event push subunit is used to generate change events containing a unique resource identifier, change type, and timestamp, and push them to the multi-cloud management agent module in real time via a secure communication link.

3. The resource management device according to claim 1, characterized in that, The timing comparison and synchronization unit specifically includes: The hash calculation subunit is used to calculate the hash value of the unique identifier and key attribute set of the resource based on the resource type, and generate the current resource hash list; The comparison and analysis subunit is used to compare the current resource hash list with the historical resource hash list stored in the second memory storage unit, and determine whether the resource has changed based on the consistency of the hash values. The difference synchronization subunit is used to obtain the complete unified data model of the changed resource, update the second memory storage unit, and synchronize the change information to the multi-cloud management agent module.

4. The resource management device according to claim 1, characterized in that, Both the first memory storage unit and the second memory storage unit use a classified key-value pair structure to organize resource data, wherein: the key is a globally unique identifier for the resource, and the value is a structured data object containing the resource specifications, status, tags and associations; the resource data is partitioned and stored according to computing, storage and network types.

5. The resource management device according to claim 1, characterized in that, It also includes a secure communication and control module, which includes: The security authentication submodule is used to establish an encrypted communication channel based on a two-way TLS or token mechanism between the multi-cloud management agent module and the cloud platform agent module, and to perform two-way identity authentication. The access control submodule is used to control the scope of operations that each agent module can perform on cloud platform resources based on pre-configured role-based access control policies.

6. The resource management device according to claim 1, characterized in that, It also includes a fault tolerance and state recovery module, which includes: The message assurance submodule is used to automatically restore the connection and ensure reliable message transmission when communication between the multi-cloud management agent module and the cloud platform agent module is abnormal. The status checkpoint submodule is used to generate and maintain consistency checkpoints. These consistency checkpoints are used to trigger an incremental data recovery process based on the checkpoint information after the agent module restarts, in order to ensure eventual data consistency. The information recorded by the consistency checkpoints includes one or more of the following: the last confirmed change version number, timestamp, and offset of pending messages.

7. The resource management device according to claim 1, characterized in that, It also includes a resource dependency management module, used for: Use a directed acyclic graph model to describe and manage creation, startup, or network connection dependencies between resources on different cloud platforms; When performing batch resource synchronization or operations, topological sorting is performed according to the directed acyclic graph to determine the synchronization or operation order of resources, avoiding errors caused by unmet dependencies.

8. The resource management device according to any one of claims 1 to 7, characterized in that, The second memory storage unit supports tiered storage strategies and resource sharding synchronization strategies: The hierarchical storage strategy is as follows: frequently accessed hot resource data is stored in memory, while infrequently accessed cold resource data is stored on disk or in an external cache database. The resource sharding synchronization strategy is as follows: cloud platform resources are logically sharded according to the hash value of the resource's unique identifier. During the timed comparison and synchronization process, only shards whose resource status has changed are subject to full comparison and synchronization.

9. A method for managing resources in a heterogeneous cloud platform based on a hybrid synchronization mechanism, characterized in that, The method is based on a multi-cloud management agent module deployed on a multi-cloud management platform and a cloud platform agent module deployed on at least one heterogeneous cloud platform, and includes the following steps: S1: The multi-cloud management platform initiates a resource query request; S2: The multi-cloud management agent module receives the request and first queries its first memory storage unit; if a match is found, the resource information is returned directly; if a match is not found, the request is forwarded to the target cloud platform agent module. S3: The cloud platform agent module maintains the data timeliness of its second memory storage unit through a hybrid synchronization mechanism and keeps synchronized with the multi-cloud management agent module; the execution of the hybrid synchronization mechanism includes: S31: Event-triggered synchronization steps: When the cloud platform agent module listens to a resource operation event generated within its cloud platform, it triggers real-time updates and pushes of resource information. S32: Periodic comparison and synchronization step: The cloud platform agent module periodically and actively queries cloud platform resources, identifies differences by comparing newly acquired resource information with historical information stored in the second memory storage unit, and performs synchronization.

10. The resource management method according to claim 9, characterized in that, Step S31 specifically includes: The cloud platform proxy module receives and parses resource operation requests or event logs forwarded by the cloud platform's internal management components; Update the resource status in the local second memory storage unit based on the parsing results; Generate a resource change event and immediately push it to the multi-cloud management agent module through the event channel; The comparison process in step S32 uses a hash comparison method, which calculates and compares the hash values ​​of key resource attributes to efficiently determine whether the resource has changed.

Images