A High-Reliability Data Service System and Method

Abstract

A high-reliability data service system includes two or more token consumption node groups. Each token consumption node group includes token consumption node group A and token consumption node group B, a high-reliability service quality management cluster, and two or more storage nodes. Each token consumption node group includes a token generation node and at least one token consumption node. The token generation node includes a token generation module and a total token bucket. Each token consumption node includes a local token bucket. The token consumption node periodically retrieves tokens from the total token bucket and stores them in the local token bucket. The token consumption node also retrieves tokens from its local token bucket.

Description

Technical Field

[0001] This application belongs to the field of big data technology, and specifically relates to a highly reliable data service system and method. Background Technology

[0002] Typically, QoS (Quality of Service) provides the following three service models:

[0003] The best-effort service model is a single, simplest service model. This model provides no guarantees regarding latency, reliability, or other performance characteristics.

[0004] 2. Integrated service – This integrated service model can meet various QoS requirements. This system can clearly distinguish and guarantee the service quality of each service flow, providing the most granular service quality differentiation for the network.

[0005] 3. Differentiated Service – This model can meet different QoS requirements. Unlike Int-Serv, it does not require notifying the network to reserve resources for each service. Differentiated services are simple to implement and have good scalability.

[0006] The role of QoS technology in the network: Flow classification: It uses certain rules to identify packets that meet certain characteristics. It is the premise and foundation for differentiating network services.

[0007] Traffic policing: Monitoring specific traffic entering or leaving a device. When traffic exceeds set limits, restrictions or penalties can be imposed to protect network resources from damage. This can be applied to both inbound and outbound traffic at interfaces.

[0008] Traffic shaping: A flow control measure that actively adjusts the output rate of a flow to adapt the flow to the network resources available to downstream devices, avoiding unnecessary packet drop. It is usually applied at the interface.

[0009] Congestion management: This refers to how to formulate a resource scheduling strategy when congestion occurs to determine the processing order of packet forwarding, which usually applies to the interface direction.

[0010] Congestion avoidance: Monitors network resource usage and takes proactive packet dropping strategies when congestion shows signs of worsening. It relieves network overload by adjusting queue lengths and typically operates at the interface direction.

[0011] However, traditional QoS lacks metrics for evaluating node QoS performance, and even more so, it lacks technical means to assess the overall QoS performance of the network. Traffic policing of a single node cannot evaluate its QoS performance; when traffic congestion occurs, packets can only be discarded, without any technical means to adjust or improve the situation.

[0012] The token bucket management unit algorithm works by adding tokens to a bucket at a constant rate. If a request needs to be processed, it must first obtain a token from the bucket. If there are no tokens available, the request is rejected. When the bucket is full, newly added tokens are either discarded or rejected.

[0013] The token bucket management unit algorithm uses a bucket with a fixed capacity of tokens, adding tokens to the bucket at a fixed rate. Essentially, the token bucket management unit algorithm can be described by the following concepts: tokens are added to the token bucket at a fixed rate, such as 100 tokens per second. The bucket can hold a maximum of b tokens; when the bucket is full, newly added tokens are discarded or rejected.

[0014] The token bucket has several important parameters:

[0015] Token generation rate: This represents the number of tokens generated per second, i.e., the number of tokens generated per unit of time. This parameter affects the system's maximum throughput and is a core parameter for rate limiting.

[0016] Token bucket capacity: Indicates the maximum number of tokens that can be stored in the token bucket. When a request arrives, if there are not enough tokens in the bucket, the request is rejected.

[0017] Warm-up time: This refers to the period after system startup during which the token bucket management unit will gradually increase the token generation rate until it reaches the set maximum rate. This parameter helps prevent a large number of tokens from being generated during system startup.

[0018] In distributed networks, there are many nodes in equal positions. For example, in a distributed storage network, there are many gateway nodes. Token buckets can be run on these gateway nodes for QoS management, but the QoS of a single gateway node has no effect on improving the overall network performance. Token bucket technology can prevent subsequent network crashes due to traffic overload; it is essentially a rate-limiting technique and cannot improve the QoS capability of individual nodes, nor can it improve the overall network traffic balance or overall network QoS.

[0019] Definitions of abbreviations and key terms

[0020] Quality of Service (QOS)

[0021] ID Identity document

[0022] TPS Transactions Per Second (TPS) Summary of the Invention

[0023] A high-reliability data service system includes two or more token-consuming node groups, each comprising token-consuming node group A and token-consuming node group B, a high-reliability service quality management cluster, and two or more storage nodes. The token-consuming node groups are signal-connected to the high-reliability service quality management cluster. Each token-consuming node group includes a token-generating node and at least one token-consuming node. The token-generating node includes a token-generating module and a main token bucket. Each token-consuming node includes a local token bucket. The token-consuming node periodically retrieves tokens from the main token bucket and stores them in its local token bucket. The high-reliability service quality management cluster is used to configure the token-generating speed of the token-generating module. The token-generating node and the token-consuming node report their own status or the status of other nodes to the high-reliability service quality management cluster. Based on the status reports, the high-reliability service quality management cluster establishes a node status table. The high-reliability service quality management cluster republishes the updated node status table after a node status change.

[0024] According to the aforementioned high-reliability data service system, the status report includes the percentage of remaining token capacity; the percentage of remaining token capacity is equal to the ratio of the current number of tokens in the token bucket to the capacity of the token bucket management unit.

[0025] Based on the aforementioned high-reliability data service system, the high-reliability service quality management cluster collects the operating parameters or operating status of the token-consuming nodes in the token-consuming node group. According to the operating status of the token-consuming nodes, the high-reliability service quality management cluster sets traffic weights in the node status table. Based on the node status table and traffic weights, the token-consuming nodes in each token-consuming node group can determine the token-generating node of their respective token-consuming group through HASH calculation. The service access end selects the corresponding token-consuming node to process the request based on the node status table, each token-consuming node group, and traffic weights.

[0026] According to the aforementioned high-reliability data service system, the operating parameters include network bandwidth and CPU type; the operating status includes transmission latency, CPU utilization, and network bandwidth percentage.

[0027] According to the aforementioned high-reliability data service system, token-consuming node group A is used to process business A, and token-consuming node group B is used to process business B.

[0028] Based on the aforementioned high-reliability data service system, the high-reliability service quality management cluster outputs a visual operation status diagram. The visual operation status diagram includes the operation status of the token generation node, and the status report of the token generation node includes the percentage of remaining token capacity.

[0029] According to the aforementioned high-reliability data service system, the operating status of the token generation node includes the token generation rate and the percentage of remaining token capacity.

[0030] According to the aforementioned high-reliability data service system, the token-consuming nodes or token-generating nodes in the token-consuming node group include a heartbeat management module; the high-reliability service quality management cluster includes a node status module; the heartbeat management module monitors the heartbeat status of other nodes, identifies nodes without heartbeat messages as faulty nodes, and reports the faulty nodes to the high-reliability service quality management cluster.

[0031] According to the aforementioned high-reliability data service system, if the difference in the remaining token capacity ratio between token consumption node group A and token consumption node group B is greater than a set value, and the remaining token capacity ratio of token consumption node group A is higher than that of token consumption node group B, the high-reliability service quality management cluster will relocate at least one token consumption node from token consumption node group A to token consumption node group B; after the relocation, the high-reliability service quality management cluster may or may not modify the token generation speed of the token generation node.

[0032] Based on the aforementioned high-reliability data service system, and by combining the load status of token-consuming nodes and the remaining token capacity ratio of the token-consuming node group, the high-reliability service quality management cluster can dynamically adjust the traffic weight of each token-consuming node and generate and publish a new version of the node status table.

[0033] According to the aforementioned high-reliability data service system, if the remaining token capacity of the total token bucket is greater than the set value, the high-reliability service quality management cluster will reduce or adjust the weight of at least one token consumption traffic associated with the token consumption node group to 0; the corresponding token consumption node will reduce the processing of requests related to the token consumption node group.

[0034] According to the aforementioned high-reliability data service system, if the remaining token capacity of the total token bucket is less than the set value, the high-reliability service quality management cluster will increase the traffic weight of the token-consuming node group among all token-consuming nodes.

[0035] According to the aforementioned high-reliability data service system, when the traffic weight of a token-consuming node in all token-consuming node groups is 0, it enters a low-power waiting state.

[0036] According to the aforementioned high-reliability data service system, if the remaining token capacity of the total token bucket is less than the set value, the high-reliability service quality management cluster will reassign at least one token-consuming node from the backup token-consuming node group C to the token-consuming node group; after the reassignment, the high-reliability service quality management cluster may or may not modify the token generation speed of the token-generating node.

[0037] According to the aforementioned high-reliability data service system, if the remaining token capacity of the total token bucket is greater than the set value, the high-reliability service quality management cluster will allocate at least one token-consuming node to the backup token-consuming node group C; after the allocation, the high-reliability service quality management cluster may or may not modify the token generation speed of the token-generating node.

[0038] Based on the aforementioned high-reliability data service system, the high-reliability service quality management cluster establishes a table of business traffic allocation percentages for token-consuming nodes; the high-reliability service quality management cluster periodically publishes the business traffic allocation percentage table to the business access end.

[0039] A high-reliability data service method is provided for managing a high-reliability quality of service (HSS) management cluster, token generation nodes, and at least two token consumption nodes. A token generation node and at least one token consumption node form a token consumption node group. The HSS management cluster records token bucket configuration information, including a token generation speed used to set the token generation speed of the token generation node. Each token generation node includes a total token bucket, which generates a specified number of tokens within a specified time according to the token generation speed and places them into the total token bucket. Each token consumption node includes a local token bucket, which retrieves tokens from the total token bucket of the token generation nodes in the same group and places them into its local token bucket. Upon receiving a service request, each token consumption node retrieves a token from its local token bucket. The token generation node and token consumption node report their own status and the status of other nodes to the HSS management cluster. The HSS management cluster establishes a node status table based on these status reports. The HSS management cluster periodically publishes the node status table through a broadcast channel. The node status table includes the token generation node and token consumption node members of the token consumption node group.

[0040] According to the aforementioned high-reliability data service method, when a token generating node fails, the token consuming node selects a new token generating node from among the token consuming nodes based on the updated node status table. The high-reliability service quality management cluster then publishes the updated node status table through a broadcast channel.

[0041] According to the aforementioned high-reliability data service method, the status report includes the percentage of remaining token capacity; the percentage of remaining token capacity is equal to the ratio of the current number of tokens in the token bucket to the capacity of the token bucket management unit.

[0042] According to the above high-reliability data service method, the token consumption node groups are token consumption node group A and token consumption node group B. The remaining token capacity ratio of token consumption node group A is higher than that of token consumption node group B. At least one token consumption node from token consumption node group A is reassigned to token consumption node group B. After the reassignment, the high-reliability service quality management cluster may or may not modify the token generation speed of the token generation node.

[0043] According to the above high-reliability data service method, if the remaining token capacity ratio of the total token bucket is greater than the set value, at least one token consuming node in the token consuming node group will be reassigned to the backup token consuming node group C; the token consuming nodes in the backup token consuming node group C will enter a low-power waiting state; after the reassignment, the high-reliability service quality management cluster may or may not modify the token generation speed of the token generating node.

[0044] According to the above high-reliability data service method, if the remaining token capacity of the total token bucket is less than the set value, the high-reliability service quality management cluster will allocate at least one token consumption node from the backup token consumption node group C to the token consumption node group; after the allocation, the high-reliability service quality management cluster may or may not modify the token generation speed of the token generation node.

[0045] Based on the aforementioned high-reliability data service method, the high-reliability service quality management cluster establishes a token consumption node business traffic allocation ratio table; the high-reliability service quality management cluster periodically publishes the business traffic allocation ratio table to the business access end.

[0046] One of the technical advantages of the above solution is that in large-scale data service systems, different services require different network resources. By forming token consumption node groups with varying numbers of servers, and having one server responsible for token generation, compared to the traditional method where each server manages its own token, token issuance for a single service can be effectively controlled through a single token consumption node. Each service corresponds to multiple token consumption nodes, preventing service interruptions caused by a web page server failure if the web page server for a particular service fails.

[0047] One of the technical effects of the above solution is that by setting up independent token nodes for different services, the token consumption requirements of different services can be dynamically managed.

[0048] One of the technical effects of the above solution is that a high-reliability service quality management cluster is used to monitor the working status of each node in the token consumption node group. When a node fails, especially when a token node fails, the network node status can be published immediately through the node status table, which greatly improves the reliability of the system.

[0049] One of the technical advantages of the above solution is that in large data centers and large databases, where hundreds or thousands of services run simultaneously, each service corresponds to a different token consumption node group. The operational status of the token nodes in each token consumption node group reflects the workload and operational status of a service. By observing the percentage of remaining token capacity in each token consumption node group, data center administrators can intuitively observe the operational health of each service. A single parameter can be used to observe the operational health of the entire network.

[0050] One of the technical effects of the above solution is that by configuring the traffic weight of the business processing nodes through token consumption, servers with different processing capabilities can be coordinated to work together. Servers with stronger capabilities can process more business, while older servers can process less business.

[0051] One of the technical effects of the above solution is that although the percentage of remaining token capacity of the token-consuming node can be used to observe the overall operating status of the network, information such as network bandwidth, CPU type, transmission latency, CPU utilization, and network bandwidth percentage can be collected and used as a reference when analyzing the network status in detail.

[0052] One of the technical effects of the above solution is that it displays the percentage of remaining tokens in the token bucket. The percentage of remaining tokens can be the percentage of remaining tokens in the total token bucket. Since the total number of token buckets is small and directly corresponds to the business, network administrators can intuitively observe the health status of different business operations.

[0053] One of the technical effects of the above solution is that the remaining token capacity ratio can be the ratio of the remaining tokens in the local token buckets of different services. If there is a problem with the remaining token ratio of the total token bucket of a service, such as the ratio increasing, the network administrator can check whether the number of token consuming nodes is insufficient by checking the remaining token ratio of the local token bucket.

[0054] One of the technical effects of the above-mentioned technical solution is that in traditional technology, heartbeat monitoring runs in all nodes of the network. The scale of heartbeat monitoring is large, and phenomena such as heartbeat monitoring islands are easy to occur. By forming small heartbeat monitoring rings in the token consumption node group, the accuracy of heartbeat monitoring can be improved.

[0055] One of the technical effects of the above technical solution is that the traffic demand of business A is reduced, and the traffic demand of business B is reduced. By allocating token consumption node 13 in token consumption node group A to token consumption node group B, the processing capacity of business B can be improved. After the allocation, the token generation speed of the token generation node can be adjusted so that the proportion of remaining tokens in the adjusted token bucket runs between a reasonable 30% and 50%.

[0056] One of the technical effects of the above technical solution is that service node 31 runs the token node for service A, and service node 33 runs the token node for service B. Service nodes 31, 32, and 33 all run the token consumption nodes for services A and B. By adjusting the traffic weight of the token consumption nodes for services A or B in different service nodes, the resource consumption ratio of services A and B in the entire network can be changed.

[0057] One of the technical effects of the above-mentioned technical solution is that, according to the above-mentioned high-reliability data service system, when the traffic weight of the token-consuming node in all token-consuming node groups is 0, it enters a low-power waiting state, which can save the power consumption of the service cluster.

[0058] One of the technical effects of the above solution is that by setting up a token consumption node group C as a backup support group, idle servers can be centrally allocated to this group for management. In the backup support group, a certain number of services can be dynamically set to be in a power-on backup state, while other servers are in a motherboard power-off shutdown state. The motherboard power-off shutdown state can significantly reduce the power consumption of the servers. When business demand increases, the servers in the power-on backup state are first scheduled to the token consumption node group with demand. When the number of services in the power-on backup state decreases to a set value, the servers are gradually powered on through the network card power-on mode, gradually entering the backup support state. This method can significantly improve the efficiency of server operation and, during periods of low business activity, power off some servers to take them out of service, significantly saving the power consumption of the data center.

[0059] One of the technical effects of the above solution is that the token generation speed of the token generation node can be modified or not modified after the high-reliability service quality management cluster is adjusted. By modifying the token generation speed of the token generation node after adjustment, the proportion of remaining token capacity after adjustment can be brought into the normal monitoring range. When the business volume changes, the proportion of remaining token capacity will fluctuate and enter an abnormal state, increasing the number of nodes in the token consumption node group that need to be adjusted again.

[0060] One of the technical effects of the above solution is that the service access end obtains the service traffic allocation ratio table, which can allocate traffic to nodes with different processing capabilities and dynamically balance the traffic of each service node.

[0061] One of the technical effects of the above-mentioned technical solution is that by using a three-level token management model, the traditional token bucket technology is decomposed into a three-level model: token bucket performance parameter management level, token generation level, and token consumption level. This makes the token bucket function clearly divided in large-scale service clusters, with each node having a single function, high operating efficiency, and good fault tolerance. In the token consumption level, multiple nodes share the business and obtain tokens from a single token generation node, making token management intuitive and efficient.

[0062] One of the technical effects of the above technical solution is that by periodically obtaining the status of each token bucket, the QoS status of the entire network can be monitored, laying the foundation for scheduling the QoS of the entire network.

[0063] One of the technical effects of the above technical solution is that the ratio of the current number of tokens to the capacity of the token bucket management unit indicates the level of token consumption. If the ratio is high, it means that the current node has a high backlog of tokens and the node is not in good working condition. The periodic ratio of the current number of tokens to the capacity of the token bucket management unit can effectively evaluate the QoS performance of the node.

[0064] One of the technical effects of the above technical solution is that the node service quality monitoring module collects the QoS of each service quality, which can adjust the traffic allocation size or traffic allocation ratio of each gateway node and adjust the balance of the entire network service.

[0065] One of the technical effects of the above solution is that the dynamic capacity consumption ratio of the token bucket can intuitively assess the QoS status of a node; 50% can be considered healthy. If it is greater than 50%, it can be considered that the token generation rate is greater than the consumption rate; if it is less than 50%, it can be considered that the consumption rate is greater than the generation rate.

[0066] One of the technical effects of the above technical solution is that the dynamic capacity consumption ratio of the token bucket varies greatly. The token bucket parameter reporting module can reduce the fluctuation frequency of the dynamic capacity consumption ratio of the token bucket by averaging the dynamic capacity consumption ratio of the token bucket collected multiple times.

[0067] One of the technical effects of the above technical solution is that the main gateway node runs a token bucket to manage multiple sub-gateway nodes, which can significantly reduce the amount of information reported to the gateway management server. At the same time, gateway grouping can balance the speed and quantity of tokens distributed by sub-gateway nodes.

[0068] One of the technical effects of the above technical solution is that the health ranking of network management nodes, which sorts the token bucket management unit capacities that are the same or similar together, can intuitively obtain the QoS capabilities of each node.

[0069] One of the technical effects of the above technical solution is that the health sorting of network management nodes, which groups nodes with the same token generation rate together, allows for a direct understanding of the QoS capabilities of each node.

[0070] One of the technical effects of the above solution is that for nodes with the same token generation rate and the same token bucket management unit capacity, the nodes with a large proportion of dynamic capacity consumption of the token bucket are the most difficult to guarantee in terms of QoS, and it is necessary to prioritize reducing traffic allocation, so as to effectively schedule QoS capabilities. Attached Figure Description

[0071] Figure 1 This is one of the implementation methods for a highly reliable data service system;

[0072] Figure 2 A schematic diagram of the operating principle of a high-reliability data service system

[0073] Figure 3 This is a diagram illustrating the percentage of token buckets in operation.

[0074] Figure 4 This is one of the implementation methods for a highly reliable data service system;

[0075] Figure 5 This is a schematic diagram of node status monitoring in a highly reliable data service system.

[0076] Figure 6 This is a diagram illustrating a token generation node selection method.

[0077] Figure 7 One implementation method for a highly reliable data service system. Detailed Implementation

[0078] The contents of this application will be further described in detail below with reference to the accompanying drawings. It should be noted that the following description is of preferred embodiments of the present invention and does not constitute any limitation on the present invention. The description of the preferred embodiments of the present invention is merely an explanation of the general principles of the invention. The designations "first," "second," "A," and "B" used in this invention are for ease of explanation only and do not represent a temporal or spatial order. The combinations of letters and numbers "TA," "TB," and "H" used in this invention are for ease of explanation only, and their specific meanings are determined by the specific terms they represent.

[0079] like Figure 1 A high-reliability data service system includes multiple service access terminals, such as service A and service B access terminals as shown in the figure; two or more token consumption node groups, each including token consumption node group A and token consumption node group B; a high-reliability service quality management cluster; two or more storage nodes, such as storage node 11, storage node 12, storage node 13, storage node 21, storage node 22, and storage node 23 as shown in the figure; the token consumption node groups are signal-connected to the high-reliability service quality management cluster; each token consumption node group includes token generation nodes, such as token node 10 and token node 20 as shown in the figure; multiple token consumption nodes, such as token consumption nodes 11, 12, and 13 in token consumption node group A; and token consumption nodes 21, 22, and 23 in token consumption node group B; as shown in the figure. Figure 2The token generation node includes a token generation module and a total token bucket; the token consumption node includes a local token bucket; the token consumption node periodically retrieves tokens from the total token bucket and stores them in its local token bucket; the token consumption node retrieves tokens from its local token bucket; the high-reliability service quality management cluster is used to configure the token generation speed of the token generation module; such as Figure 5 ,like Figure 6 Token generating nodes and token consuming nodes report their own or other nodes' status reports to the high reliability service quality management cluster. The high reliability service quality management cluster builds a node status table based on the status reports. The high reliability service quality management cluster republishes the updated node status table after a node's status changes.

[0080] Token consumption nodes can be web page servers. Token consumption node group A corresponds to processing business A, and token consumption node group B corresponds to processing business B.

[0081] The token consumption node can be a gateway server. Token consumption node group A corresponds to the data storage business of processing business A, and token consumption node group B corresponds to the data storage business of processing business B.

[0082] In large-scale data service systems, different services require different network resources. By grouping servers into token consumption node groups with a varying number of servers, and having one server responsible for token generation, compared to the traditional method where each server manages its own token bin, token issuance for a single service can be effectively controlled through a single token consumption node. Each service corresponds to multiple token consumption nodes, preventing situations where a single web page server fails, causing service interruption for that service.

[0083] By setting up independent token nodes for different business functions, the token consumption requirements of different business functions can be managed dynamically.

[0084] Using a high-reliability service quality management cluster to monitor the working status of each node in the token consumption node group, when a node fails, especially when a token node fails, the node status table can publish the network node status in the first instance, which greatly improves the reliability of the system.

[0085] According to the aforementioned high-reliability data service system, the status report includes the percentage of remaining token capacity; the percentage of remaining token capacity is equal to the ratio of the current number of tokens in the token bucket to the capacity of the token bucket management unit.

[0086] In large data centers and large databases, hundreds or thousands of services run simultaneously. Each service corresponds to a different token consumption node group. The operational status of the token nodes in each token consumption node group reflects the workload and operational status of a service. By observing the percentage of remaining token capacity in each token consumption node group, data center administrators can intuitively observe the operational health of each service. A single parameter can be used to observe the operational health of the entire network.

[0087] Based on the aforementioned high-reliability data service system, the high-reliability service quality management cluster collects the operating parameters or operating status of the token-consuming nodes in the token-consuming node group. According to the operating status of the token-consuming nodes, the high-reliability service quality management cluster sets traffic weights in the node status table. Based on the node status table and traffic weights, the token-consuming nodes in each token-consuming node group can determine the token-generating node of their respective token-consuming group through HASH calculation. The service access end selects the corresponding token-consuming node to process the request based on the node status table, each token-consuming node group, and traffic weights.

[0088] By configuring the traffic weight of the business processing nodes through token consumption, servers with different processing capabilities can be coordinated to work together. Servers with stronger capabilities can handle more business, while older servers can handle less business.

[0089] According to the aforementioned high-reliability data service system, the operating parameters include network bandwidth and CPU type; the operating status includes transmission latency, CPU utilization, and network bandwidth percentage.

[0090] While the percentage of remaining token capacity of a token-consuming node can provide insight into the overall network operating status, collecting information such as network bandwidth, CPU type, transmission latency, CPU utilization, and network bandwidth percentage can serve as a relevant reference when conducting detailed analysis of the network status.

[0091] According to the aforementioned high-reliability data service system, token-consuming node group A is used to process business A, and token-consuming node group B is used to process business B.

[0092] like Figure 3 Based on the aforementioned high-reliability data service system, the high-reliability service quality management cluster outputs a visual operation status diagram. The visual operation status diagram includes the operation status of the token generation node, and the status report of the token generation node includes the percentage of remaining token capacity.

[0093] like Figure 3 The image above shows the percentage of remaining tokens in the token bucket. The percentage of remaining tokens can be the percentage of remaining tokens in the total token bucket. The number of total token buckets is small and directly corresponds to the business. Network administrators can intuitively observe the health status of different business operations.

[0094] The percentage of remaining token capacity can be the percentage of remaining tokens in the local token buckets of different services. If there is a problem with the percentage of remaining tokens in the total token bucket of a service, such as an increase in the percentage, the network administrator can check the percentage of remaining tokens in the local token bucket to see if the number of token consuming nodes configured is insufficient.

[0095] According to the aforementioned high-reliability data service system, the operating status of the token generation node includes the token generation rate and the percentage of remaining token capacity.

[0096] like Figure 5 , Figure 6 The token-consuming nodes or token-generating nodes in the token-consuming node group include a heartbeat management module; the high-reliability service quality management cluster includes a node status module; the heartbeat management module monitors the heartbeat status of other nodes, identifies nodes without heartbeat messages as faulty nodes, and reports faulty nodes to the high-reliability service quality management cluster. In traditional technology, heartbeat monitoring runs on all nodes in the network, resulting in a large-scale heartbeat monitoring system that is prone to heartbeat monitoring silos. By forming small heartbeat monitoring rings within the token-consuming node group, the accuracy of heartbeat monitoring can be improved.

[0097] According to the aforementioned high-reliability data service system, if the difference in the remaining token capacity ratio between token consumption node group A and token consumption node group B is greater than a set value, and the remaining token capacity ratio of token consumption node group A is higher than that of token consumption node group B, the high-reliability service quality management cluster will relocate at least one token consumption node from token consumption node group A to token consumption node group B; after the relocation, the high-reliability service quality management cluster may or may not modify the token generation speed of the token generation node.

[0098] like Figure 4 If the traffic demand of business A decreases, and the traffic demand of business B also decreases, the processing capacity of business B can be improved by relocating token consumption node 13 from token consumption node group A to token consumption node group B. After the relocation, the token generation speed of the token generation node can be adjusted so that the proportion of remaining tokens in the adjusted token bucket is between 30% and 50%.

[0099] like Figure 7 In actual physical servers, multiple token-consuming nodes run concurrently. Based on the aforementioned high-reliability data service system, and considering the load status of the token-consuming nodes and the remaining token capacity percentage of the token-consuming node group, the high-reliability service quality management cluster can dynamically adjust the traffic weight of each token-consuming node, generating and publishing a new version of the node status table. When network service traffic changes, traffic scheduling between different services can be achieved by adjusting the node traffic weights.

[0100] like Figure 7 Service node 31 runs the token node for service A, and service node 33 runs the token node for service B. Service nodes 31, 32, and 33 all run the token consumption nodes for services A and B. By adjusting the traffic weight of the token consumption nodes for services A or B in different service nodes, the resource usage ratio of services A and B in the entire network can be changed.

[0101] According to the aforementioned high-reliability data service system, if the remaining token capacity percentage of the total token bucket exceeds a set value, the high-reliability service quality management cluster will reduce or adjust the token consumption traffic weight associated with at least one token consumption node group to 0; the corresponding token consumption node will reduce the processing of requests related to that token consumption node group. For example, if a service's remaining token capacity percentage is greater than 90%, it means that the current service traffic demand is relatively low. By reducing or adjusting the weight of one token consumption traffic to 0, the token consumption node will be taken out of service for server maintenance.

[0102] According to the aforementioned high-reliability data service system, if the remaining token capacity of the total token bucket is less than the set value, the high-reliability service quality management cluster will increase the traffic weight of the token-consuming node group among all token-consuming nodes.

[0103] According to the aforementioned high-reliability data service system, when the traffic weight of the token-consuming node in all token-consuming node groups is 0, it enters a low-power waiting state, which can save power consumption of the service cluster.

[0104] According to the aforementioned high-reliability data service system, if the remaining token capacity of the total token bucket is less than the set value, the high-reliability service quality management cluster will reassign at least one token-consuming node from the backup token-consuming node group C to the token-consuming node group; after the reassignment, the high-reliability service quality management cluster may or may not modify the token generation speed of the token-generating node.

[0105] By setting up token consumption node group C as a backup support group, idle servers are centrally allocated to this group for management. Within the backup support group, a certain number of services can be dynamically configured to be in a power-on backup state, while other servers are in a motherboard power-off shutdown state. The motherboard power-off shutdown state can significantly reduce server power consumption. When business demand increases, servers in the power-on backup state are first scheduled to the token consumption node group with demand. When the number of services in the power-on backup state decreases to a set value, the servers are gradually powered on through network card power-on mode, gradually entering the backup support state. This method can significantly improve server operating efficiency and, during periods of low business activity, power off some servers to take them out of service, significantly saving data center power consumption.

[0106] According to the aforementioned high-reliability data service system, if the remaining token capacity percentage of the total token bucket exceeds a set value, the high-reliability service quality management cluster will reassign at least one token-consuming node to the backup token-consuming node group C. After reassignment, the high-reliability service quality management cluster may or may not modify the token generation speed of the token-generating nodes. By modifying the token generation speed of the token-generating nodes after adjustment, the adjusted remaining token capacity percentage can be brought back into the normal monitoring range. When business volume changes, the remaining token capacity percentage will fluctuate, entering an abnormal state, thus increasing the number of nodes in the token-consuming node group to be adjusted again.

[0107] Based on the aforementioned high-reliability data service system, the high-reliability service quality management cluster establishes a business traffic allocation ratio table for token-consuming nodes; the high-reliability service quality management cluster periodically publishes this business traffic allocation ratio table to the business access end. Upon receiving the business traffic allocation ratio table, the business access end can distribute traffic to nodes with different processing capabilities, dynamically balancing the traffic across all business nodes.

[0108] like Figure 2 This is a three-tiered token management method used to manage a high-reliability quality of service (HSS) management cluster, token generation nodes, and at least two token consumption nodes. A token generation node and at least one token consumption node form a token consumption node group. The HSS management cluster, as the first tier, records token bucket configuration information, including the token generation speed used to set the token generation speed of the token generation nodes. The token generation node, as the second tier, includes a total token bucket and generates a specified number of tokens within a specified time according to the token generation speed, placing them into the total token bucket. The token consumption node, as the third tier, includes a local token bucket, which retrieves tokens from the total token bucket of the token generation nodes in the same group and places them into its local token bucket. When a token consumption node receives a service request, it retrieves a token from its local token bucket. Token generation nodes and token consumption nodes report their own status and the status of other nodes to the HSS management cluster. The HSS management cluster establishes a node status table based on these status reports. The HSS management cluster periodically publishes the node status table through a broadcast channel. The node status table includes the token generation nodes and token consumption node members of the token consumption node group.

[0109] By adopting a three-level token management model, the traditional token bucket technology is decomposed into three levels: token bucket performance parameter management level, token generation level, and token consumption level. This makes the token bucket function clearly divided in large-scale service clusters, with each node having a single function, high operating efficiency, and good fault tolerance. In the token consumption level, multiple nodes share the business and obtain tokens from a single token generation node, making token management intuitive and efficient.

[0110] According to the aforementioned high-reliability data service method, when a token generating node fails, the token consuming node selects a new token generating node from among the token consuming nodes based on the updated node status table. The high-reliability service quality management cluster then publishes the updated node status table through a broadcast channel.

[0111] According to the aforementioned high-reliability data service method, the status report includes the percentage of remaining token capacity; the percentage of remaining token capacity is equal to the ratio of the current number of tokens in the token bucket to the capacity of the token bucket management unit.

[0112] According to the above high-reliability data service method, the token consumption node groups are token consumption node group A and token consumption node group B. The remaining token capacity ratio of token consumption node group A is higher than that of token consumption node group B. At least one token consumption node from token consumption node group A is reassigned to token consumption node group B. After the reassignment, the high-reliability service quality management cluster may or may not modify the token generation speed of the token generation node.

[0113] According to the above high-reliability data service method, if the remaining token capacity ratio of the total token bucket is greater than the set value, at least one token consuming node in the token consuming node group will be reassigned to the backup token consuming node group C; the token consuming nodes in the backup token consuming node group C will enter a low-power waiting state; after the reassignment, the high-reliability service quality management cluster may or may not modify the token generation speed of the token generating node.

[0114] According to the above high-reliability data service method, if the remaining token capacity of the total token bucket is less than the set value, the high-reliability service quality management cluster will allocate at least one token consumption node from the backup token consumption node group C to the token consumption node group; after the allocation, the high-reliability service quality management cluster may or may not modify the token generation speed of the token generation node.

[0115] Based on the aforementioned high-reliability data service method, the high-reliability service quality management cluster establishes a token consumption node business traffic allocation ratio table; the high-reliability service quality management cluster periodically publishes the business traffic allocation ratio table to the business access end.

[0116] While the present invention has been described and illustrated with reference to preferred embodiments and several alternatives, the invention is not limited to the specific descriptions herein. Other alternatives or equivalent components may also be used to practice the invention.

Claims

1. A high-reliability data service system, characterized by: The high-reliability service quality management cluster, two or more storage nodes, two or more token consumption node groups, the token consumption node groups including token consumption node group A and token consumption node group B; The token consumption node groups are in signal connection with the high-reliability service quality management cluster; The token consumption node groups include a token generation node and at least one token consumption node; the token generation node includes a token generation module and a total token bucket; The token consumption node includes a local token bucket; The token consumption node periodically takes tokens from the total token bucket and stores the tokens in the local token bucket; The token consumption node takes tokens from the local token bucket; The high-reliability service quality management cluster is configured to configure the speed of the token generation module to generate tokens; The token generation node and the token consumption node report a state report of themselves or other nodes to the high-reliability service quality management cluster, the high-reliability service quality management cluster establishes a node state table according to the state report, and the high-reliability service quality management cluster reissues an updated node state table after a change in the node state; The high-reliability service quality management cluster collects running parameters or running states of the token consumption nodes in the token consumption node groups, sets a traffic weight in the node state table according to the running states of the token consumption nodes, and each token consumption node in each token consumption node group can learn the token generation node of the token consumption group through HASH calculation based on the node state table and the traffic weight; A service access end selects a corresponding token consumption node to process a request based on the node state table and each token consumption node group and the traffic weight.

2. The high-reliability data service system of claim 1, wherein, The state report includes a residual token capacity proportion, and the residual token capacity proportion is equal to a ratio of a current token number of the token bucket to a capacity of a token bucket management unit.

3. The high-reliability data service system of claim 2, wherein, The running parameters include network bandwidth and CPU type, and the running states include transmission delay, CPU utilization, and network bandwidth proportion. Further include any one or more of the following features: Feature TA1: the token consumption node group A is used to process service A, and the token consumption node group B is used to process service B; Feature TA2: the high-reliability service quality management cluster outputs a visual running state diagram, the visual running state diagram includes a running state of the token generation node, and the state report of the token generation node includes a residual token capacity proportion; Feature TA3: the running state of the token generation node includes a token generation rate and a residual token capacity proportion; Feature TA4: the token consumption node or the token generation node in the token consumption node group includes a heartbeat management module; The high-reliability service quality management cluster includes a node state module; The heartbeat management module monitors the heartbeat state of other nodes, identifies a node without a heartbeat message as a fault node, and reports the fault node to the high-reliability service quality management cluster.

4. The high-reliability data service system of claim 3, wherein, Further include any one or more of the following features: Feature TB1: the difference between the remaining token capacity proportion of the token consumption node group A and the token consumption node group B is greater than a set value, the remaining token capacity proportion of the token consumption node group A is higher than that of the token consumption node group B, and the high-reliability service quality management cluster allocates at least one token consumption node of the token consumption node group A to the token consumption node group B; after the allocation, the high-reliability service quality management cluster modifies or does not modify the token generation speed of the token generation node; Feature TB2: in combination with the load state of the token consumption node and the remaining token capacity proportion of the token consumption node group, the high-reliability service quality management cluster can dynamically adjust the traffic weight of each token consumption node, generate and publish a new version of the node state table.

5. The high-reliability data service system of claim 3, wherein, Further comprising any one or more of the following features: Feature TC1: the remaining token capacity proportion of the total token bucket is greater than a set value, the high-reliability service quality management cluster reduces or adjusts to 0 at least one token consumption traffic weight associated with the token consumption node group; the corresponding token consumption node reduces processing of requests related to the token consumption node group; Feature TC2: the remaining token capacity proportion of the total token bucket is less than a set value, the high-reliability service quality management cluster increases the traffic weight of the token consumption node group in each token consumption node; Feature TC3: when the traffic weight of the token consumption node in all token consumption node groups is 0, the token consumption node enters a low-power waiting state; Feature TC4: the remaining token capacity proportion of the total token bucket is less than a set value, the high-reliability service quality management cluster allocates at least one token consumption node of the backup token consumption node group C to the token consumption node group; after the allocation, the high-reliability service quality management cluster modifies or does not modify the token generation speed of the token generation node; Feature TC5: the remaining token capacity proportion of the total token bucket is greater than a set value, the high-reliability service quality management cluster allocates at least one token consumption node to the backup token consumption node group C; after the allocation, the high-reliability service quality management cluster modifies or does not modify the token generation speed of the token generation node; Feature TC6: the high-reliability service quality management cluster builds a token consumption node business traffic distribution proportion table; The high-reliability service quality management cluster periodically publishes the business traffic distribution proportion table to the business access end.

6. A high-reliability data service method, characterized by: For managing a high-reliability service quality management cluster, a token generation node, and at least two token consumption nodes; One token generation node and at least one token consumption node form a token consumption node group; The high-reliability service quality management cluster is used for recording token bucket configuration information, and the token bucket configuration information includes a token generation speed for setting a token generation speed of a token generation node; The token generation node includes a total token bucket, and generates a specified number of tokens in a specified time according to the token generation speed and places the tokens into the total token bucket; The token consumption node includes a local token bucket, and takes tokens from the total token bucket of the token generation node in the same group and places the tokens into the local token bucket; The token consumption node receives a business request and takes tokens from the local token bucket; The token generation node and the token consumption node report a state report of themselves or other nodes to the high reliability service quality management cluster, and the high reliability service quality management cluster establishes a node state table according to the state report; and the high reliability service quality management cluster periodically publishes the node state table through a broadcast channel; The node state table includes a token generation node and token consumption node members of a token consumption node group; The high reliability service quality management cluster collects running parameters or running states of the token consumption nodes in the token consumption node group, and sets a traffic weight in the node state table according to the running states of the token consumption nodes; the token consumption nodes in each token consumption node group can learn the token generation node of the token consumption group through HASH calculation based on the node state table and the traffic weight; A service access end selects corresponding token consumption nodes to process a request based on the node state table and each token consumption node group and the traffic weight.

7. The high-reliability data service method of claim 6, wherein, When the token generation node fails, the token consumption nodes select a node as a new token generation node in the token consumption nodes based on an updated node state table, and the high reliability service quality management cluster publishes the updated node state table through a broadcast channel.

8. The high-reliability data service method of claim 7, wherein, The state report includes a residual token capacity proportion; and the residual token capacity proportion is equal to a ratio of a current token number of a token bucket to a capacity of a token bucket management unit.

9. The high-reliability data service method of claim 8, wherein, Further comprising any one or more of the following features: Feature TD1: The token consumption node group is a token consumption node group A and a token consumption node group B, a residual token capacity proportion of the token consumption node group A is higher than that of the token consumption node group B, and at least one token consumption node of the token consumption node group A is allocated to the token consumption node group B; the high reliability service quality management cluster modifies or does not modify a token generation speed of the token generation node after the allocation; Feature TD2: In combination with a load state of the token consumption node and the residual token capacity proportion of the token consumption node group, the high reliability service quality management cluster can dynamically adjust the traffic weight of each token consumption node, generate and publish a new version of the node state table; Feature TD3: When the residual token capacity proportion of the total token bucket is greater than a set value, the high reliability service quality management cluster reduces or adjusts at least one token consumption traffic weight associated with the token consumption node group to 0; and the corresponding token consumption node reduces processing of a request related to the token consumption node group; Feature TD4: When the residual token capacity proportion of the total token bucket is less than a set value, the high reliability service quality management cluster increases the traffic weight of the token consumption node group in each token consumption node; Feature TD5: When the traffic weight of the token consumption node in all token consumption node groups is 0, the token consumption node enters a low-power waiting state; Feature TD6: When the residual token capacity proportion of the total token bucket is greater than a set value, at least one token consumption node of the token consumption node group is allocated to a backup token consumption node group C; The token consumption nodes of the backup token consumption node group C enter a low-power waiting state; and The token consumption nodes of the backup token consumption node group C enter a low-power waiting state. The high-reliability service quality management cluster modifies or does not modify the token generation speed of the token generation node after deployment; Feature TD7: The remaining token capacity proportion of the total token bucket is less than a set value, and the high-reliability service quality management cluster deploys at least one token consumption node of the backup token consumption node group C to the token consumption node group; the high-reliability service quality management cluster modifies or does not modify the token generation speed of the token generation node after deployment; Feature TD8: The high-reliability service quality management cluster builds a token consumption node service traffic allocation proportion table; The high-reliability service quality management cluster periodically publishes the service traffic allocation proportion table to the service access end.

Images