TWI484335B - Cached data processing method, processing system, and means - Google Patents

Description

Processing method, processing system and device for cache data

The present application relates to the field of network technologies, and in particular, to a method, a processing system and a device for processing cache data.

Distributed Cache, which caches cached data in decentralized system memory, distributes cached content to physically isolated cache nodes, and can be event-driven third-party The application provides high availability, high performance and scalability. Among them, high availability means that event processing third-party applications can publish output events to a repetitive, decentralized cache to ensure high availability of their calculations; high performance and scalability refer to many event-driven third parties. Applications need to use both streaming and external data, such as data retrieved from persistent storage. Caches can be used to increase the speed of accessing non-streaming data, thereby dynamically improving the performance of third-party applications, and are widely used in Internet third-party applications and large-scale website architecture design.

Currently, distributed caches mainly include server-side decentralization and client-side decentralization.

The server-decentralized decentralized cache structure is shown in Figure 1. It consists of Cache Client (Cache Client), Cache Node A (Cache Node A), Cache Node B (Cache Node B), and Cache. Node C (cache service node C). It is characterized by decentralized operations performed by the server. Usually, JGroup's multicast communication is used as a means to perform operations and synchronize data through a three-phase transaction protocol. The main representatives of this type are: JBoss Cache and EhCache. The biggest problem with this model is the efficiency problem. Due to the unreliability of multicast communication itself and the complexity of the strategy of three-phase transaction, the efficiency of such cache is not high enough to be applied to the system architecture with frequent operations and high concurrent requests. .

The decentralized decentralized cache structure of the client is shown in Figure 2. It consists of Cache Client, Cache Node A, Cache Node B, and Cache Node C. The feature is that the cache server itself is a centralized cache, the server is not responsible for distributing data and synchronizing data, but through the client's algorithm settings, the data of different keys is distributed to different servers, forming an expandable Client-side decentralized cache system. The main representative of this type: Memcached Cache. This mode is highly scalable and efficient, but lacks a fault tolerance mechanism. When one of the data cache service nodes is unavailable, the system will lose some data and affect the normal use of the business system. Sina's open source Memcached DB, the data is solidified to the memory database or local hard disk to ensure the integrity of the data, but this does not have the effect of cluster fault tolerance and balanced load under high concurrency.

The prior art has the following problem: the failure of the cache service node causes the data loss to be unrecoverable. The data is distributed by the client to a cache service node through a fixed algorithm. The cache node exception will cause all data on the cache service node to be lost. Moreover, the failure of the cache service node prevents the normal service from continuing. The data is distributed by the client to a cache service node through a fixed algorithm. The cache service node exception will cause the requests distributed to the cache service node to be no longer processed.

The application provides a method, a processing system and a device for processing cache data, which are applied to a network of multiple cache service nodes in a cluster, and are used for solving data loss caused by failure of the cache service node and the cache service node is unavailable. Use problems to improve system reliability and availability.

The present application provides a method for processing cache data, which is applied to a cluster including a plurality of cache service nodes, including: the cache client sends an operation to the primary cache service node when the cache data needs to be operated. a request; the cache client receives an operation result of the operation request by the primary cache service node; and the cache client selects an alternate cache service node when the primary cache service node fails to perform an operation and Send an operation request to the alternate cache service node.

Before selecting an alternate cache service node and sending an operation request to the standby cache service node, the method further includes: when the primary cache service node performs an operation failure, the cache client accumulates the number of failures; the cache The client compares the number of failures with a threshold, and when the number of failures exceeds a threshold, the primary cache service node is marked as invalid.

After the primary cache service node is marked as invalid, the method further includes: setting an expiration validity period of the primary cache service node. The selecting an alternate cache service node includes: the cache client acquiring a cache service node available in the network according to the obtained cache service node marked as invalid; the cache client is in the available cache In the service node, a cache service node is selected as an alternate cache service node according to the algorithm.

The algorithm includes: dynamic load distribution selection algorithm and hash algorithm or consistent hash algorithm.

After the operation request is sent to the standby cache service node, the method further includes: the cache client receiving the operation result of the operation request by the standby cache service node, and processing according to the operation result.

Before the cache client sends an operation request to the active cache service node, the method further includes: the cache client acquiring the cache service node available in the network according to the obtained cache service node marked as invalid; The cache client selects a cache service node as the active cache service node according to the algorithm in the available cache service node.

The processing according to the operation result includes: determining, by the cache client, whether the operation is successful according to the operation result; when the operation is successful, the cache client performs asynchronous processing on the primary cache service node; Upon failure, the cache client invalidates the alternate cache service node and reselects the alternate cache service node.

The cache client performs asynchronous processing on the active cache service node, specifically: when the operation succeeds and is an acquisition operation, the cache client caches the operation result to the cache client, and The operation and operation result are saved to the cluster operation task queue and the operation and operation content is restored to the cluster by the task thread pool to remove other cache service nodes that perform the operation; when the operation is successful and the update operation is performed, The cache client saves the operation and operation results to the cluster operation task queue and copies the operation and operation content to the cluster by the task thread pool to remove other cache service nodes that perform the operation.

Before the cache client sends an operation request to the active cache service node, the method further includes: the cache client acquiring the pre-designated primary cache service node when the cache data needs to be operated.

The processing according to the operation result includes: determining, by the cache client, whether the operation is successful according to the operation result; when the operation is successful, the standby cache service node performs data synchronization on the primary cache service node; Upon failure, the cache client will reselect the alternate cache service node and send an operation request to the alternate cache service node.

The method further includes: the cache client performing heartbeat detection on the primary cache service node and the standby cache service node; the cache client updating the primary cache service node and/or standby fast according to a result of heartbeat detection Take the failure status of the service node.

It also includes: when the capability of the cache service node in the cluster is insufficient, dynamic expansion of the single node can be performed.

The application provides a processing system for cache data, including a plurality of cache service nodes, including: a cache client, configured to send an operation request to the primary cache service node, and receive the master cache service node pair. The operation result of the operation request; when it is determined that the operation fails according to the received operation result of the operation request by the active cache service node, selecting an alternate cache service node and sending an operation cache to the standby cache service node a data request; the main cache service node is configured to receive an operation cache data request sent by the cache client, and return an operation result to the cache client; the alternate cache service node, when the main cache service When the node fails, the operation cache data request sent by the cache client is received.

When multiple cache service nodes in the cluster are in the active cluster configuration mode, multiple cache service nodes in the cluster are in working state. When one cache service node fails, other cache service nodes share the request; the cluster When multiple cache service nodes are in the Standby cluster configuration mode, only one of the cache service nodes in the cluster is in the working state as the primary cache service node, and other cache service nodes are in the standby state. When the primary cache service node fails, one of the other cache service nodes replaces the primary cache service node as a backup cache service node and receives the request.

When the capacity of multiple cache service nodes in the cluster is insufficient, dynamic expansion of single nodes can be performed.

The application provides a cache client, which is applied to a cluster including a plurality of cache service nodes, and includes: a request sending unit, configured to send an operation to the active cache service node when the cache data needs to be operated. a result receiving unit, configured to receive an operation result of the operation request sent by the primary cache service node to the request sending unit, and a cache service node selecting unit, configured to determine, according to an operation result received by the result receiving unit When the primary cache service node fails to perform the operation, an alternate cache service node is selected and an operation request is sent by the request sending unit to the standby cache service node.

The method further includes: a failure indication unit, configured to mark the primary cache service node as invalid when it is determined that the primary cache service node fails to perform an operation according to the operation result received by the result receiving unit.

The failure indication unit specifically includes: a time accumulation subunit, configured to accumulate the number of failures of the cache client when the main cache service node fails to perform the operation; the failure flag subunit is used to accumulate the subunits The cumulative number of failures is compared with a threshold value. When the number of failures exceeds the threshold, the primary cache service node is marked as invalid; the invalidation validity period setting subunit is used for the primary use in the failure indication subunit. After the cache service node is marked as invalid, the expiration validity period of the active cache service node is set.

The cache service node selection unit specifically includes: an available cache service node acquisition subunit, configured to acquire a cache service node available in the network according to the cache service node indicated as the invalidation by the failure indication subunit; the cache service a node selection subunit, configured to select a cache service node as an alternate cache service node according to an algorithm, in the available cache service node obtained by the available cache service node acquisition subunit, the algorithm includes dynamic load distribution Choose an algorithm and a hash algorithm or a consistent hash algorithm.

The result receiving unit is further configured to receive an operation result of the operation request sent by the standby cache service node to the request sending unit; the cache client further includes: an operation result processing unit, configured to receive the The alternate cache service node processes the result of the operation request.

When multiple cache service nodes in the cluster are in the active cluster configuration mode, the cache service node selection subunit is further used to obtain available cache service nodes obtained by the subunits in the available cache service node, according to the calculation The method selects a cache service node as the primary cache service node.

The operation result processing unit further includes: a non-synchronization processing sub-unit, configured to perform non-synchronization processing on the main cache service node when the operation is successful according to the operation result received by the result receiving unit of the cache client; When the operation succeeds and the operation is fetched, the result of the operation is cached to the cache client, and the operation and operation result are saved to the cluster operation task queue and the operation and operation are performed by the task thread pool. The content is restored to the cluster and the other cache service nodes that perform the operation are removed; when the operation is successful and the update operation is performed, the operation and operation results are saved to the cluster operation task queue and the operation and operation are performed by the task thread pool. The content is copied to the cluster and the other cache service nodes that do this are removed.

The failure flag indicating unit is further configured to: when the operation failure is determined according to the result of the operation received by the result receiving unit, the standby cache service node is marked as invalid.

When the plurality of cache service nodes in the cluster are in the Standby cluster configuration mode, the operation result processing unit further includes: a data synchronization sub-unit, configured to determine, when the operation succeeds according to the operation result received by the result receiving unit of the cache client , data synchronization is performed on the standby cache service node.

The method further includes: a heartbeat detecting unit, configured to perform heartbeat detection on the primary cache service node and the standby cache service node, and update the failure of the primary cache service node and/or the standby cache service node according to the heartbeat detection result. status.

The present application includes the following advantages, because the cluster configuration mode is adopted, when the primary cache service node fails, the standby cache service node is selected to perform operations, thereby solving the data loss caused by the failure of the cache service node and the cache service node being unavailable. The problem is to improve system reliability and availability.

The main idea of the present application includes that when the primary cache service node fails to execute the operation cache data, the cache client marks the primary cache service node as invalid, selects an alternate cache service node and sends the standby cache node to the standby fast. The service node sends an operation request, thereby solving the problem of data loss caused by the failure of the cache service node and the unavailability of the cache service node, thereby achieving the effect of improving system reliability and availability.

The technical solutions in the present application are clearly and completely described in the following with reference to the drawings in the present application. It is obvious that the described embodiments are a part of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts are within the scope of the present application.

The present application provides a method for processing cache data, which is used in a network of multiple cache service nodes in a cluster, and solves the problem of data loss caused by failure of the cache service node and the cache service node being unavailable. The problem, as shown in FIG. 3, includes a cache client for receiving a cache operation request of a third-party application and selecting a cache service node; the cache service node is equivalent to a logical cache server for Cache the data in the database, including the following steps:

In step s301, the cache client sends an operation request to the active cache service node when the cache data needs to be operated.

In step s302, the cache client receives the operation result of the operation request by the active cache service node.

Step s303: The cache client selects an alternate cache service node and sends an operation request to the standby cache service node when the primary cache service node fails to perform the operation.

The cluster configuration mode includes an Active Cluster configuration mode and a Standby cluster configuration mode.

The active cluster configuration mode means that multiple cache service nodes in the cluster are in working state. When a cache service node fails, other cache service nodes share the request, and at the end of the operation, the data of each node in the cluster is performed. Synchronize.

The Standby cluster configuration mode means that only one cache service node in the cluster is in the working state as the primary cache service node, and other cache service nodes are in the standby state. When the primary cache service node fails. One of the other cache service nodes replaces the active cache service node as an alternate cache service node, receives the request, and periodically synchronizes the data of each node in the cluster.

The above two Active and Standby cluster configuration modes actually provide an automatic fault recovery capability for the entire network system, ensuring that the client can access the data without interruption without manual intervention.

When multiple cache service nodes in the cluster are in the active cluster configuration mode, as shown in FIG. 4, the specific steps of the method for processing the cache data are: step s401, when the cache client needs to operate on the cache data. Selecting an active cache service node through the algorithm in multiple cache service nodes in the cluster and sending an operation request to the master cache service node.

Specifically, if the cache service node in the cluster is in a failed state, the failure indication is saved, and the cache client obtains the cache service node marked as invalid according to the storage from the local storage or in another location, and further Obtaining a cache service node available in the network; the cache client selects a cache service node as the primary cache service node according to the algorithm in the available cache service node, and the algorithm includes dynamic load distribution selection calculation Method, hash algorithm or consistent hash algorithm.

In step s402, the cache client receives the operation result of the operation request by the active cache service node.

After the cache client selects a cache service node as the active cache service node according to the algorithm, the cache client sends a request for operating the cache data to the master cache service node, and the cache server uses the cache. After the service node performs a request to operate the cache data, the service result is sent to the cache client.

In step s403, when the main cache service node fails to perform the operation, the cache client marks the primary cache service node as invalid, selects an alternate cache service node, and sends an operation request to the standby cache service node.

Specifically, when the main cache service node fails to perform the operation, the cache client accumulates the number of operation failures; the cache client compares the number of failures with the threshold, and when the number of failures exceeds the threshold, the cache is used for the main cache. The service node is marked as invalid and the expiration date is set.

At this time, the cache client obtains the cache service node available in the network according to the obtained cache service node marked as invalid; the cache client selects a backup according to the algorithm in the available cache service node. The service node is cached and an operation request is sent to the alternate cache service node. The alternate cache service node performs the operation cache data request and sends the operation result to the cache client.

When the operation is successful according to the result of the operation, the cache client performs asynchronous processing on the primary cache service node. The asynchronous processing specifically includes: when the operation is successful and the operation is acquired, the cache client will The result of the operation is cached to the cache client, and the operation and operation results are saved to the cluster operation task queue. The operation and operation content is restored to the cluster by the task thread pool to remove other cache service nodes that perform the operation; when the operation is successful and is an update operation, the cache client saves the operation and operation result to the cluster operation. The task queues and copies the operations and operations from the task thread pool to the cluster to remove other cache service nodes that do the operation.

In the Active Cluster Configuration mode, when the primary cache service node fails, the standby cache service node takes over the primary cache service node; when the primary cache service node is restored, the cluster operation task queues the standby cache service. The operation and operation result of the node is restored or copied to the active cache service node. At this time, the primary cache service node will be in the working state again, and the standby cache service node will not take over the request from the primary cache service node.

The above asynchronous processing is further used to update the operation performed by the primary cache service node to all other cache service nodes in the cluster when the primary cache service node has not failed. That is to say, in the Active Cluster Configuration mode, whether the primary cache service node or the standby cache service node is operated, the operation and operation result are saved to the cluster operation task queue in the queue. Waiting for non-synchronization of data to all other cache service nodes in the cluster to ensure the consistency of data of all cache service nodes in the cluster.

When it is determined according to the operation result that the operation fails, the cache client invalidates the alternate cache service node identifier and reselects the alternate cache service node.

The dynamic load distribution selection algorithm in the above step s401 is as shown in FIG. 5, and the specific steps are as follows: in step s501, the Cache Client obtains parameters of the available cache service nodes from the available cache service nodes through the background timing. Including the cache service node processing capability index W[i], the cache server node current cache data total C[i], and the cache service node request response time R[i].

In step s502, the Cache Client performs an ascending or descending power on each parameter of the obtained available cache service node.

For example: F(K[i], {K[1]..K[N]}, ASC) means that the array {K[1]...K[N]} is first ranked as an ascending power, and then K is returned. i] The subscript of the location.

F(K[i], {K[1]..K[N]}, DESC) means that the array {K[1]...K[N]} is firstly arranged in power, and then K[i] is returned. The subscript of the location.

In step s503, the Cache Client calculates each parameter according to the formula of the selected cache service node.

Specifically, select the cache service node formula as:

Max(F(W[i],{W[1]...W[N]}, ASC)*25%+F(C[i],{C[1]...C[N]}, DESC)

*25%+F(R[i], {R[1]...R[N]}, DESC)*50%). The specific meaning is that the processing power accounts for 25% of the weight (the greater the processing power, the better), the total amount of stored content accounts for 25% of the weight (the less the storage content is better), the response time accounts for 50% of the weight (the shorter the response time, the better).

Step s504: The Cache Client selects the active cache service node according to the calculation result of selecting the cache service node formula, and obtains the last cached service node with the largest value as the selected cache service node.

When multiple cache service nodes in the cluster are in the Standby cluster configuration mode, as shown in FIG. 6, the specific steps of the method for processing the cache data are: Step s601, when the cache client needs to operate on the cache data. And sending an operation request to the primary cache service node specified in the cluster, wherein the determination of the primary cache service node is in the order of the cache service node entering the cluster, that is, the cache service node in the cluster enters the cluster according to the order. The order of the first order is the primary cache service node in the cluster; when the primary cache service node fails, the cache service node that enters the second cluster time becomes the primary service. The cache service node; when the failed original primary cache service node recovers, it will be rearranged at the end of the cache service node in the cluster.

Step s602: The cache client receives an operation result of the operation request by the active cache service node.

When the cache client sends a request for operating the cache data to the active cache service node, the master cache service node performs a request for operating the cache data, and then sends the operation result to the cache client.

In step s603, when the main cache service node fails to perform the operation, the cache client marks the primary cache service node as invalid, selects an alternate cache service node, and sends an operation request to the standby cache service node.

Specifically, when the main cache service node fails to perform the operation, the cache client accumulates the number of operation failures; the cache client compares the number of failures with the threshold, and when the number of failures exceeds the threshold, the cache is used for the main cache. The service node is marked as invalid and the expiration date is set.

The cache client obtains the cache service node available in the network according to the obtained cache service node marked as invalid; the cache client is in the available cache service node, according to the hash algorithm or consistency. The algorithm selects a cache service node as the alternate cache service node. The alternate cache service node performs the operation cache data request and sends the operation result to the cache client.

When it is determined according to the operation result that the operation fails, the cache client invalidates the alternate cache service node identifier and reselects the alternate cache service node.

In the Standby cluster configuration mode, when the primary cache service node fails and the standby cache service node takes over the primary cache service node, the standby cache service node will permanently replace the original primary cache service node as the primary cache service. Node; when the original primary cache service node is restored, it acts as the alternate cache service node of the current primary cache service node. Moreover, the alternate cache service node in the cluster periodically requests the data set from the current active cache service node to perform asynchronous synchronization of the data to ensure the consistency of data of all the cache service nodes in the cluster.

In the embodiment of the present application, the Active Cluster configuration mode is applied to the method in the present application to apply to a system with high data security requirements and a high read/write ratio. The network includes Application (third party application), Cache Client, Cache Node A, and Cache Node B. When a third party that needs to use the cache mechanism, that is, a third-party application sends a read or operation cache data request, the Cache Client fails to perform a failure operation according to the cache node A that performs the operation cache data selected by the dynamic load distribution algorithm. The Cache Client re-selects the Cache Node B to perform operation cache data according to the dynamic load distribution algorithm, and performs asynchronous processing when the Cache Node B operation succeeds.

Specifically, as shown in FIG. 7, the processing method of the cache data includes the following steps:

In step s701, the third-party application sends a read or operation cache request to the Cache Client.

In step s702, the Cache Client receives the request sent by the third-party application to determine the available cache service node.

If the cache service node in the cluster is in a failed state, a failure indication is saved. The Cache Client checks the status of all cache service nodes, determines whether the cache service node is available according to whether it is marked with a failure indication, and determines the cache service node without the invalidation identifier as the available cache service node. When the Cache Client does not check the available cache service node, the Cache Client sends an error message to the third-party application; when the Cache Client checks the available cache service node, it proceeds to step s703.

In step s703, the Cache Client selects the active cache service node.

Specifically, the step is the same as "selecting a cache service node as the primary cache service node according to the algorithm" in step s401.

In this procedure, the Cache Node A is selected by the Cache Client as an example.

In step s704, when the Cache Client selects the Cache Node A, the operation cache data request is sent to the Cache Node A.

In step s705, the Cache Node A receives the operation request sent by the Cache Client, and operates the cache data.

In step s706, when the Cache Node A completes the operation request, the operation result is returned to the Cache Client.

In step s707, the Cache Client determines whether the operation is successful according to the returned operation result.

In step s708, the Cache Client performs processing according to whether the operation is successful.

In this flow, the operation failure is taken as an example for description.

Specifically, when the operation fails, the operation may fail due to other reasons (for example, a network connection failure), so the Cache Node A failure is not directly indicated, but the failure counter is accumulated to the failure counter of the Cache Node A. Get the number of failures n. The failure counter is set with a threshold N. When the number of failures n is greater than the threshold N, the Cache Client will indicate that the state of the Cache Node A is invalid.

In step s709, the Cache Client receives the request sent by the third-party application to determine the available alternate cache service node.

If the cache service node in the cluster is in a failed state, a failure indication is saved. The Cache Client obtains the cache service nodes available in the network according to the obtained cache service node marked as invalid. When the Cache Client does not check the available cache service node, the Cache Client sends an error message to the third-party application; when the Cache Client checks the available cache service node, it proceeds to step s710.

Step s710: The Cache Client selects a cache service node as an alternate cache service node from the available cache service nodes according to the algorithm, and the algorithm includes a dynamic load distribution selection algorithm and a hash algorithm or a consistent hash calculation. law.

In this procedure, the Cache Node B is selected by the Cache Client as an example.

In step s711, the Cache Client sends an operation cache data request to the Cache Node B.

In step s712, the Cache Node B receives the operation request sent by the Cache Client, and operates the cache data.

In step s713, when the Cache Node B completes the operation request, the operation result is returned to the Cache Client.

In step s714, the Cache Client determines whether the operation is successful according to the received operation result.

When the operation fails, go to step s709; when the operation is successful, go to step s715; in step s715, the Cache Client performs processing according to the operation result.

Specifically, it includes any of the following two situations:

(a) When the operation is successful and the operation is fetched, if the cache is needed, cache the data to the Cache Client, set the validity period, increase the access speed of the hit data, and save the operation and operation result to the cluster operation task queue. .

(b) When the operation is successful and is an update operation, the Cache Client saves the operation and operation result to the cluster operation task queue.

In step s716, the Cache Client returns the operation result to the third-party application.

After the operation and the operation result are saved in the cluster operation task queue in the above step s715, the asynchronous processing and the non-synchronous data recovery and copying are performed by using the asynchronous mode and the queue combination.

Specifically, as shown in FIG. 8, the method for asynchronous processing includes the following steps:

In step s801, the Cache Client records the acquisition or update operation performed by the cache service node to the queue operation task queue.

In step s802, the task consumer sends a polling check to the cluster operation task queue to check whether a new task is in the queue. When it is found that there is a new task in the queue, go to step s803.

In step s803, the task consumer sends the task to the task thread pool.

In step s804, the cluster operation task queue deletes the task.

In step s805, the task thread pool analyzes the task according to the task command, the task target, and the task content, and creates a task thread according to the analysis result.

In step s806, the task thread pool performs a cache processing node asynchronous processing task.

When the operation is a get operation, the operation and operation content are restored to the cluster to remove other cache service nodes that perform the operation; when the operation is an update operation, the operation and operation content are copied to the cluster to remove the Other cache service nodes that operate;

In step s807, the cluster cache service node sends the operation result to the task thread pool.

In step s808, the task thread pool determines whether the task is successfully executed according to the operation result and performs processing according to the judgment result.

When the task is successfully executed, the task ends; when the task fails to execute, the failed task is processed, specifically one of the following:

(a) Simply discard.

(b) Limited retry.

(c) The task is solidified to the local cache.

In the embodiment of the present application, the method in the present application adopts the Standby cluster configuration mode, which is applied to a system with a high proportion of read operations and low requirements for data synchronization, but high reliability requirements. The network includes Application, Cache Client, Cache Node A, and Cache Node A'. When the primary cache service node Cache Node A is not invalid but the execution operation fails or the primary cache service node Cache Node A fails, the Cache Client issues a request to the alternate cache service node Cache Node A', and the Cache Node A' performs the operation. And synchronize the data yourself.

Specifically, as shown in FIG. 9, the processing method of the cache data includes the following steps:

In step s901, the third-party application reads or operates the cache data request to the sending Cache Client.

In step s902, the Cache Client receives the request sent by the third-party application, and checks the status of the Cache Node A of the primary cache service node in the local cache. If the invalidation status has expired, the invalidation flag is cleared.

When the primary cache service node has not expired, go to step s903; when the primary cache service node fails, go to step s908;

In step s903, the Cache Client sends an operation cache data request to the Cache Node A.

In step s904, the Cache Node A receives the Cache Client to issue an operation cache data request, and operates the cache data.

In step s905, after the Cache Node A executes the operation cache data, the operation result is returned to the Cache Client.

In step s906, the Cache Client determines whether the operation is successful according to the returned operation result.

In step s907, the Cache Client performs processing according to whether the operation is successful.

This process uses the failure of this operation as an example.

Since the operation may fail due to other reasons (for example, a network connection failure), the Cache Node A failure is not directly indicated, but the failure counter is accumulated to the failure counter of the Cache Node A, and the number of failures n is obtained. The failure counter is set with a threshold N. When the number of failures n is greater than the threshold N, the state of the Cache Node A is marked as invalid, and the expiration date is set.

In step s908, the Cache Client selects one of the available cache service nodes as the alternate cache service node.

Specifically, if the cache service node in the cluster is in a failed state, a failure indication is saved. The Cache Client obtains the cache service node available in the network according to the obtained cache service node marked as invalid, and selects one of them as the backup cache service node. When the Cache Client does not check the available cache service node, the Cache Client sends an error message to the third-party application.

This process uses the Cache Client to select the alternate cache service node Cache Node A' as an example.

In step s909, the Cache Client sends an operation cache data request to the alternate cache service node Cache Node A'.

In step s910, the Cache Node A' performs the operation.

In step s911, the Cache Node A' returns the operation result to the Cache Client.

In step s912, the Cache Client determines whether the operation is successful according to the received operation result.

When the operation fails, go to step s908; when the operation is successful, go to step s913.

In step s913, the Cache Client returns the operation result to the third-party application.

In step s914, the alternate cache service node Cache Node A's timing synchronization data.

It should be noted that, at this time, the standby cache service node takes over the original active cache service node, which is the current active cache service node, and becomes the standby of the current active cache service node when the original primary cache service node is restored. The service node is cached, so Cache Node A will request the data set as the current alternate cache service node to the current primary cache service node Cache Node A' for data asynchronous synchronization.

Specifically, the alternate cache service node Cache Node A' timing synchronization data includes the following steps:

(a) Cache Node A requests a data set from Cache Node A';

(b) Cache Node A' returns the requested data set result to Cache Node A, and Cache Node A completes the data synchronization with Cache Node A'.

In the Standby cluster configuration mode, the Cache Client performs periodic heartbeat detection on the cache service node in the cluster, and periodically updates the state marked with the failure indication and the node marked with the failure indication cache according to the result of the heartbeat detection.

When the cache service node is marked with a failure indication, the cache client performs heartbeat detection on the cache service node. When the response times exceed the threshold, the cache client clears the failure indication of the cache service node. Specifically, as shown in FIG. 10, the Cache Client performs a specific step of performing heartbeat detection on the cache service node marked with a failure indication: this process uses an example of the Cache Node C marked with a failure indication as an example.

In step s1001, the Cache Client sends a periodic heartbeat detection packet to the Cache Node C.

In step s1002, the Cache Node C returns the heartbeat detection result to the Cache Client.

In step s1003, the Cache Client determines, according to the received heartbeat detection result, whether the Cache Node C responds to the data packet.

In step s1004, if the Cache Node C responds to the data packet, the Cache Node C invalidation flag is cleared.

When the cache service node is not marked with a failure indication, the cache client performs heartbeat detection on the cache service node. When the number of unanswered times exceeds the threshold, the cache client indicates that the cache service node is invalid. Specifically, as shown in FIG. 11 , a specific step for the Cache Client to perform heartbeat detection on a cache service node that is not marked with a failure indication is as follows: This process uses an example of a Cache Node D that is not marked with a failure indication.

In step s1101, the Cache Client sends a periodic heartbeat detection packet to the Cache Node D.

In step s1102, the Cache Node D returns the heartbeat detection result to the Cache Client.

In step s1103, the Cache Client determines whether the Cache Node D responds to the data packet according to the received heartbeat detection result.

In step s1104, if the Cache Node D does not respond to the data packet, the Cache Client accumulates the number of failures to the failure counter of the Cache Node D.

In step s1105, when the number n of failures of the Cache C is greater than the threshold N, the state of the Cache Node D is updated, and the Cache Node D is invalidated.

It should be noted that when the operation of the cache service node fails to be detected by the heartbeat detection and the operation of the cache service node receiving the operation cache request fails, the cumulative number of failures reaches the same failure counter, and the number of failures n is the sum of the two, The threshold N is compared.

The embodiment of the present application further provides a processing system for cache data, which is structured as shown in FIG. 12, and includes: a cache client 1201, configured to send an operation request to the main cache service node 1202, and receive the main request. The operation result of the operation request by the cache service node 1202; when it is determined that the operation fails according to the received operation result of the operation request by the primary cache service node 1202, an alternate cache service node 1203 is selected and is fast to the standby. The service node 1203 sends an operation cache request.

The active cache service node 1202 is configured to receive an operation cache data request sent by the cache client 1201, and return an operation result to the cache client 1201.

The alternate cache service node 1203 receives the operation request sent by the cache client 1201 when the primary cache service node 1202 fails.

When the active cache service node 1202 and the standby cache service node 1203 are in the active cluster configuration mode, multiple cache service nodes in the cluster are in a working state. When a cache service node fails, other cache service nodes are inactive. When the primary cache service node 1202 and the standby cache service node 1203 are in the Standby cluster configuration mode, only one of the plurality of cache service nodes in the cluster is in the working state as the primary cache service node. The other cache service nodes are in the standby state. When the primary cache service node fails, one of the other cache service nodes replaces the primary cache service node as the standby cache service node, and receives the request.

In addition, when multiple cache service nodes in the cluster have insufficient capacity, dynamic expansion of single nodes can be performed.

As shown in FIG. 13 , a schematic structural diagram of a cache client 1300 in the embodiment of the present application includes: a request sending unit 1301, configured to send to the active cache service node when the cached data needs to be operated. Operation request.

The result receiving unit 1302 is configured to receive an operation result of the operation request sent by the primary cache service node to the request sending unit.

The cache service node selecting unit 1303 is configured to: when determining, according to the operation result received by the result receiving unit, that the main cache service node fails to perform an operation, select an alternate cache service node and request the sending unit to the standby cache The service node sends an operation request.

As shown in FIG. 14, the schematic diagram of the configuration of the cache client 1400 in the active cluster configuration mode in the embodiment of the present application includes: a request sending unit 1401, when the operation of the cached data is required, The primary cache service node sends an operation request.

The result receiving unit 1402 is configured to receive an operation result of the operation request sent by the primary cache service node to the request sending unit 1401.

The failure indication unit 1403 is configured to indicate that the primary cache service node is invalid when the operation fails according to the operation result received by the result receiving unit 1402.

The failure indication unit 1403 specifically includes: a number accumulation sub-unit 14031, configured to: when the main cache service node performs an operation failure, the cache client accumulates the number of failures; the failure indication sub-unit 14032 is configured to accumulate the sub-unit 14031 The cumulative number of failures is compared to the threshold, and when the number of failures exceeds the threshold, the primary cache service node is marked as invalid.

The expiration validity period setting sub-unit 10433 is configured to set the expiration validity period of the main cache service node after the failure flag sub-unit 14032 indicates that the main cache service node is invalid.

The cache service node selection unit 1404 is configured to select an alternate cache service node and send an operation request to the alternate cache service node.

The cache service node selection unit 1404 specifically includes: an available cache service node acquisition subunit 14041, configured to acquire a cache service node available in the network according to the cache service node indicated as invalid; the cache service node selection subunit 14042. The available cache service node obtained by the available cache service node acquisition subunit 14041 selects a cache service node as a backup cache service node according to an algorithm, and the algorithm includes a dynamic load distribution selection algorithm. And a hash algorithm or a consistent hash algorithm.

The cache service node selection unit 1404 is further configured to select the primary cache service node and send an operation request to the primary cache service node by the request sending unit 1401.

The request sending unit 1401 is further configured to send an operation request to the standby cache service node when the cached data needs to be operated.

The result receiving unit 1402 is further configured to receive an operation result of the operation request sent by the requesting sending unit 1401 by the standby cache service node.

The failure indication unit 1403 is further configured to: when the operation failure is determined according to the operation result received by the result receiving unit 1402, mark the standby cache service node as invalid.

The cache client 1400 further includes an operation result processing unit 1405 for processing the operation result of the operation request according to the alternate cache service node received by the result receiving unit 1402.

The operation result processing unit 1405 further includes: a non-synchronization processing sub-unit 14051, configured to perform a non-cache on the main cache service node when the operation is successful according to the operation result received by the result receiving unit 1402 of the cache client. Synchronous processing.

The non-synchronization processing sub-unit 14051 is specifically configured to: when the operation succeeds and is an acquisition operation, the cache client caches the operation result to the cache client, and saves the operation and the operation result to the cluster operation task queue. And the task thread pool restores the operation and operation content to the cluster and removes other cache service nodes that perform the operation; when the operation succeeds and is an update operation, the cache client saves the operation and operation result to the cluster operation. The task queues and copies the operations and operations from the task thread pool to the cluster to remove other cache service nodes that do the operation.

FIG. 15 is a schematic structural diagram of a cache client 1500 in a Standby cluster configuration mode according to an embodiment of the present application, including: a request sending unit 1501, configured to perform operations on cache data when The primary cache service node sends an operation request.

The result receiving unit 1502 is configured to receive an operation result of the operation request sent by the primary cache service node to the request sending unit 1501.

The failure indication unit 1503 is configured to indicate that the primary cache service node is invalid when the operation fails according to the operation result received by the result receiving unit 1502.

The failure indication unit 1503 specifically includes: a number accumulation subunit 15031, configured to: when the main cache service node performs an operation failure, the cache client accumulates the number of failures; the failure indication subunit 15032 is configured to use the number accumulation subunit 15031. The cumulative number of failures is compared to the threshold, and when the number of failures exceeds the threshold, the primary cache service node is marked as invalid.

The expiration validity period setting sub-unit 15033 is configured to set an expiration validity period of the main cache service node after the failure flag sub-unit 15032 indicates that the main cache service node is invalid.

The cache service node selection unit 1504 is configured to select an alternate cache service node and send an operation request to the alternate cache service node by the request sending unit 1501.

The request sending unit 1501 is further configured to send an operation request to the standby cache service node when the cached data needs to be operated.

The result receiving unit 1502 is further configured to receive an operation result of the operation request sent by the requesting sending unit 1501 by the standby cache service node.

The failure indication unit 1503 is further configured to: when the operation failure is determined according to the operation result received by the result receiving unit 1502, mark the standby cache service node as invalid.

The cache client 1500 further includes an operation result processing unit 1505, and the standby cache service node received by the result receiving unit 1502 processes the operation result of the operation request.

The operation result processing unit 1505 further includes: a data synchronization subunit 15051, configured to perform data synchronization on the alternate cache service node when the operation is successful according to the operation result received by the result receiving unit 1502 of the cache client; The cache client 1500 further includes: a heartbeat detecting unit 1506, configured to perform heartbeat detection on the primary cache service node and the standby cache service node, and update the primary cache service node and/or the standby cache according to the heartbeat detection result. The failure status of the service node.

For the convenience of description, each part of the Cache Client described above is divided into various modules by functions. Of course, in the implementation of the present application, the functions of each module can be implemented in the same software or multiple hardware or hardware.

When the capacity of multiple cache service nodes in the cluster in the above embodiment is insufficient, dynamic expansion of a single node can be performed. For a centralized Cache, in order to provide extensions, the Cache Node is actually a virtual cache service node, which is composed of cache Instances, as shown in Figure 16. Cluster A is composed of Cache Node A and Cache Node B. Cache Node A is composed of Cache Instance 1, Cache Instance 2 and Cache Instance 3. Cache NodeB is composed of Cache Instance4, Cache Instance5 and Cache Instance6. These cache instances can be on the same server or distributed on different servers, so when the cache service node is not enough, you only need to increase the cache instance in the virtual cache service node. Improve processing power without affecting the usage of the Cache Client.

However, when the number of cache instances increases or decreases, it will directly affect the hit of the data acquisition algorithm. For example, there are only three cache service nodes, then hash according to Key, and then get the hash result to the number of cache service nodes. Mice, and finally determine which instance to store the content to, or from which instance. When the Cache Node is added or decreased, it is found that the previously stored data will not be obtained normally. After the cluster configuration is adopted, data migration can be performed in two ways: 1. The Cache Client provides a data migration interface, which can move a cache service node data of the cluster to other cache service nodes. For example, if the Cache Node A is expanded, the data can be directly migrated from the Cache Node B to the Cache Node A. The migration process obtains all the data of the Cache Node B through the Cache Client, and then deposits the Cache Node A according to the new algorithm. 2. Automatically and slowly migrate. When the Active Cluster configuration mode is adopted, if the Cache Client cannot obtain the data after the Cache Node A is expanded, the data will be obtained from other cache service nodes in the cluster. If the data is obtained, the data will be updated according to the new one. The algorithm is saved to Cache Node A, which will slowly copy all the accessed data to Cache Node A.

For convenience of description, each part of the processing system of the cache data described above is divided into various modules by function. Of course, in the implementation of the present application, the functions of each module can be implemented in the same software or multiple hardware or hardware.

The present application includes the following advantages, because the cluster configuration mode is adopted, when the primary cache service node fails, the standby cache service node is selected to perform operations, thereby solving the data loss caused by the failure of the cache service node and the cache service node being unavailable. The problem is achieved by improving system reliability and availability. Of course, implementing any of the products of the present application does not necessarily require all of the advantages described above to be achieved at the same time.

Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of a software plus a necessary universal hardware platform, and of course, can also be through hardware, but in many cases, the former is more Good implementation. Based on such understanding, the technical solution of the present application may be embodied in the form of a software product in essence or in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for making one The terminal device (which may be a cell phone, a personal computer, a server, or a network device, etc.) performs the methods described in various embodiments of the present application.

A person skilled in the art can understand that the drawings are only a schematic diagram of a preferred embodiment, and the modules or processes in the drawings are not necessarily required to implement the present application.

The above description is only a preferred embodiment of the present application, and it should be noted that those skilled in the art can make several improvements and refinements without departing from the principles of the present application. And retouching should also be considered as the scope of protection of this application.

1201. . . Cache client

1202. . . Primary cache service node

1203. . . Alternate cache service node

1300. . . Cache client

1301. . . Request sending unit

1302. . . Result receiving unit

1303. . . Cache service node selection unit

1400. . . Cache client

1401. . . Request sending unit

1402. . . Result receiving unit

1403. . . Failure indicator unit

14031. . . Cumulative subunit

14032. . . Failure flag subunit

14033. . . Expiration validity period setting subunit

1404. . . Cache service node selection unit

14041. . . The sub-unit can be obtained by the cache service node.

14042. . . Cache service node selection subunit

1405. . . Operation result processing unit

14051. . . Asynchronous processing subunit

1500. . . Cache client

1501. . . Request sending unit

1502. . . Result receiving unit

1503. . . Failure indicator unit

15031. . . Cumulative subunit

15032. . . Failure flag subunit

15033. . . Expiration validity period setting subunit

1504. . . Cache service node selection unit

1505. . . Operation result processing unit

15051. . . Data synchronization subunit

1506. . . Heartbeat detection unit

In order to more clearly illustrate the present application or the technical solutions in the prior art, the drawings to be used in the present application or the prior art description will be briefly described below. Obviously, the drawings in the following description are only the present application. For some embodiments of the present invention, other drawings may be obtained from those skilled in the art without departing from the drawings.

1 is a schematic diagram of a distributed decentralized cache structure of a server in the prior art;

2 is a schematic diagram of a distributed decentralized cache structure of a client in the prior art;

FIG. 3 is a flowchart of a method for cache data according to an embodiment of the present application;

4 is a flowchart of a method for processing cache data in an Active Cluster configuration mode according to an embodiment of the present application;

FIG. 5 is a flowchart of a dynamic load distribution selection algorithm in an embodiment of the present application; FIG.

6 is a flowchart of a method for processing cache data in a Standby cluster configuration mode according to an embodiment of the present application;

FIG. 7 is a flowchart of a method for processing cache data in an Active Cluster configuration mode according to an embodiment of the present application;

FIG. 8 is a flowchart of a method for asynchronous processing in an Active Cluster configuration mode according to an embodiment of the present application;

FIG. 9 is a flowchart of a method for cache data in a Standby cluster configuration mode according to an embodiment of the present application;

10 is a flowchart of a heartbeat detection method in a Standby cluster configuration mode according to an embodiment of the present application;

11 is a flowchart of a heartbeat detection method in a Standby cluster configuration mode according to an embodiment of the present application;

12 is a schematic structural diagram of a cache data processing system in an embodiment of the present application;

FIG. 13 is a schematic structural diagram of a cache client according to an embodiment of the present application;

14 is a schematic structural diagram of a cache client in an Active Cluster configuration mode according to an embodiment of the present application;

15 is a schematic structural diagram of a cache client in a Standby cluster configuration mode according to an embodiment of the present application;

16 is a schematic structural diagram of a cluster in an embodiment of the present application;

Claims (16)

A method for processing cache data is applied to a cluster including a plurality of cache service nodes, and is characterized in that: the cache client sends an operation to the master cache service node when the cache data needs to be operated. a request; the cache client receives an operation result of the operation request by the primary cache service node; and the cache client selects an alternate cache service node when the primary cache service node fails to perform an operation The standby cache service node sends an operation request; after the operation request is sent to the standby cache service node, the method further includes: the cache client receiving the operation result of the operation request by the standby cache service node, and performing the operation result according to the operation result Processing, the processing according to the operation result includes: determining, by the cache client, whether the operation is successful according to the operation result; when the operation is successful, the cache client performs asynchronous processing on the primary cache service node; The asynchronous processing of the primary cache server by the cache client includes: when the operation is successful and the operation is acquired, The operating result cache client cache client to the cache, and save the results of operations and operations to cluster operations task queue task by the thread pool operations And the operation content is restored to the cluster to remove other cache service nodes that perform the operation; when the operation is successful and is an update operation, the cache client saves the operation and the operation result to the cluster operation task queue and is executed by the task. The pool pool copies the operation and operation content to the cluster to remove other cache service nodes that do the operation. The method of claim 1, wherein the selecting an alternate cache service node and sending the operation request to the standby cache service node further comprises: when the primary cache service node fails to perform the operation The cache client accumulates the number of failures; the cache client compares the number of failures with a threshold, and when the number of failures exceeds the threshold, the primary cache service node is marked as invalid. The method of claim 2, wherein after the primary cache service node is marked as invalid, the method further includes: setting an expiration validity period of the primary cache service node. The method of claim 1, wherein the selecting an alternate cache service node comprises: the cache client obtaining the cache available in the network according to the obtained cache service node marked as invalid. a service node; the cache client selects a cache service node as an alternate cache service node according to an algorithm in the available cache service node. The method of claim 4, wherein the calculation The method includes: dynamic load distribution selection algorithm, hash algorithm or consistent hash algorithm. The method of any one of claims 1 to 5, wherein, before the cache client sends an operation request to the active cache service node, the method further includes: the cache client according to the obtained identifier Obtaining a cache service node available in the network for the failed cache service node; the cache client selects a cache service node as the primary cache service node according to the algorithm in the available cache service node . The method of claim 1, wherein the processing according to the operation result further comprises: when the operation fails, the cache client invalidates the alternate cache service node identifier, and reselects the alternate cache. Service node. The method of any one of claims 1 to 5, further comprising: performing dynamic expansion of a single node when the capacity of the cache service node in the cluster is insufficient. A processing system for cache data, comprising a plurality of cache service nodes, comprising: a cache client, configured to send an operation request to the primary cache service node, and receive the pair of the primary cache service node The operation result of the operation request; when it is determined that the operation fails according to the received operation result of the operation request by the active cache service node, selecting an alternate cache service node and The standby cache service node sends an operation cache data request; the primary cache service node is configured to receive an operation cache data request sent by the cache client, and return an operation result to the cache client; and the standby fast Receiving the service node, when the primary cache service node fails, after receiving the operation cache data sent by the cache client, requesting to send an operation request to the standby cache service node, the method further includes: the cache client receiving The standby cache service node processes the result of the operation request and processes the result according to the operation result; the processing according to the operation result includes: the cache client determines whether the operation is successful according to the operation result; when the operation is successful, the The cache client performs asynchronous processing on the primary cache service node; and the cache client performs asynchronous processing on the primary cache service node, specifically: when the operation is successful and the operation is acquired, the fast The client fetches the result of the operation to the cache client, and saves the operation and operation result to the cluster operation task queue and is tasked by the task The thread pool restores the operation and operation content to the cluster to remove other cache service nodes that perform the operation; when the operation succeeds and is an update operation, the cache client saves the operation and operation result to the cluster operation task. Column and the task thread pool to copy the operation and operation content to the cluster to remove the operation Other cache service nodes. The system of claim 9, wherein when the capacity of the plurality of cache service nodes in the cluster is insufficient, dynamic expansion of the single node can be performed. A cache client is applied to a cluster including a plurality of cache service nodes, and is characterized in that: a request sending unit is configured to send an operation to the active cache service node when the cache data needs to be operated. a result receiving unit, configured to receive an operation result of the operation request sent by the main cache service node to the request sending unit; and a cache service node selecting unit, configured to determine, according to the result of the operation received by the receiving unit When the primary cache service node fails to perform the operation, selecting an alternate cache service node and sending an operation request to the standby cache service node by the request sending unit; wherein the result receiving unit is further configured to receive the standby cache An operation result of the operation request sent by the service node to the request sending unit; the cache client further includes: an operation result processing unit, configured to receive, according to the result, the operation result of the operation request by the standby cache service node received by the receiving unit Processing; the operation result processing unit further includes: a non-synchronization processing sub-unit, configured to The successful operation of the client takes result receiving unit receives the operation result of the determination, the client cache of the main non-synced with caching service node; specific: When the operation succeeds and the operation is obtained, the operation result is cached to the cache client, and the operation and operation result are saved to the cluster operation task queue and the operation and operation content are restored by the task thread pool. Remove other cache service nodes from the cluster to perform the operation; when the operation succeeds and is an update operation, save the operation and operation results to the cluster operation task queue and copy the operation and operation content by the task thread pool Remove other cache service nodes from the cluster that are doing this. The cache client according to claim 11, wherein the method further includes: a failure indication unit, configured to: when it is determined according to the operation result received by the result receiving unit, that the main cache service node fails to perform an operation, The primary cache service node is marked as invalid. The cache client according to claim 12, wherein the failure flag unit specifically includes: a time accumulation subunit, configured to accumulate the cache client when the main cache service node fails to perform an operation. The number of failures; the failure flag sub-unit is configured to compare the number of failures accumulated by the number of accumulated sub-units with a threshold, and when the number of failures exceeds the threshold, the main cache service node is marked as invalid; The expiration validity period setting subunit is configured to set an expiration validity period of the main cache service node after the failure signing subunit indicates that the main cache service node is invalid. The cache client according to claim 13 , wherein the cache service node selection unit specifically includes: The cache service node may be used to obtain a subunit, configured to acquire a cache service node available in the network according to the cached service node marked as invalid according to the failure indication subunit; and a cache service node selection subunit for The available cache service node obtained by the cache service node acquisition sub-unit may select a cache service node as the alternate cache service node according to the algorithm, and the algorithm includes a dynamic load distribution selection algorithm and a hash algorithm or Consistent hash algorithm. The cache client according to claim 14, wherein when the plurality of cache service nodes in the cluster are in an active cluster configuration mode, the cache service node selection subunit is further used in the available cache. The serving node obtains the available cache service node obtained by the subunit, and selects a cache service node as the active cache service node according to the algorithm. The cache client according to claim 12, wherein the invalidation indicator unit is further configured to: when the operation failure is determined according to the result of the operation received by the result receiving unit, the standby cache service node is marked as invalid. .