Elastic capacity expanding and shrinking system applied to nuclear industry
By using a monitoring and acquisition adaptation layer, a safety policy adapter, a multi-factor decision engine, and a safety constraint-driven actuator, combined with the K-ELI model and a circuit breaker coordination module, the shortcomings of existing technologies in nuclear industry applications are addressed, enabling precise flexible expansion and contraction decisions and safety compliance, thus meeting the special needs of the nuclear industry.
Patent Information
- Application Number
- CN202511964941.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-24
- Publication Date
- 2026-04-28
AI Technical Summary
Existing general-purpose flexible expansion and contraction technologies suffer from problems in nuclear industry applications, such as decision-making lag, disconnect between safety strategies and flexible mechanisms, poor platform adaptability, coarse expansion and contraction granularity, and lack of intelligent collaboration with processes specific to nuclear industry applications, thus failing to meet the special needs of nuclear industry applications.
It employs a monitoring and acquisition adaptation layer, a security policy adapter, a multi-factor decision engine, and a security constraint-driven executor, combined with the K-ELI model and a circuit breaker coordination module, to achieve precise elastic scaling decisions, support Pod-level and process group-level scaling operations, and perform real-time verification and optimization of security policies.
It achieves precise elastic response, improves the accuracy and timeliness of scaling decisions, ensures security and compliance, supports fine-grained resource control at the process group level, and improves resource utilization and system stability.
Smart Images

Figure CN121940414A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of nuclear power technology, specifically relating to a flexible expansion and contraction system for nuclear industry applications. Background Technology
[0002] With the widespread adoption of cloud computing and containerization technologies, automatic elastic scaling (auto-scaling) has become a core technology for improving resource utilization and application resilience. Container orchestration platforms, such as Kubernetes, primarily rely on Horizontal Pod Autoscaler (HPA) for their core elasticity capabilities. HPA makes decisions based on preset basic resource metric thresholds (such as average CPU and memory utilization) or custom metrics (such as requests per second - QPS) to dynamically adjust the number of Pod replicas.
[0003] Nuclear industry applications (such as reactor condition monitoring, safety control systems, radiation dose management, and simulation tasks) have significant unique characteristics: extreme safety and reliability requirements: service interruptions or performance degradation can have serious consequences; unpredictable load fluctuations: affected by real-time event triggers, periodic tasks, or analysis requests; highly sensitive business metrics: such as control command latency, sensor data queue depth, and simulation task queue counts; and stringent compliance and security policies: mandatory enforcement of access control (SELinux), security partitioning, and auditing requirements (mapped to the "three-person management" principle).
[0004] For nuclear industry applications, existing general-purpose elastic scaling technologies (represented by Kubernetes HPA) have the following serious shortcomings in meeting the requirements of nuclear industry applications in chemical environments:
[0005] 1. Scaling up and down decisions are delayed and inaccurate, making it impossible to quickly respond to sudden business peaks.
[0006] 2. The disconnect between security policies and resilience mechanisms can lead to instance startup failures, service unavailability, or, more seriously, introduce security vulnerabilities and even violate regulatory requirements.
[0007] 3. Poor platform adaptability and insufficient compatibility with container engines (iSula / Pouch) and hardware platform characteristics: It is difficult to efficiently collect chip-specific performance indicators or fine-grained operating status of container engines; performance is low or the failure rate is high when calling the scaling API of the container platform; the hardware architecture characteristics are not fully utilized for optimized scheduling; monitoring data collection is incomplete or inaccurate, affecting decision quality; and operating efficiency is low, affecting elastic response speed.
[0008] 4. Coarse-grained scaling and lack of coordination: It can only perform overall scaling at the Pod / container level, which cannot meet the needs of certain critical modules in the nuclear industry that require fine-grained resource allocation at the process group / thread pool level (such as optimizing the number of parallel computing tasks). It lacks intelligent coordination with nuclear industry-specific processes (such as circuit breaker degradation mechanisms): When a service triggers circuit breaker degradation (to protect the service), general scaling may still be performing scaling down operations, exacerbating service pressure; or it may waste resources due to decision delays when the load decreases. Summary of the Invention
[0009] To overcome the problems existing in related technologies, a flexible scaling system for nuclear industry applications is provided. The system includes: a monitoring and acquisition adaptation layer, a safety policy adapter, a multi-factor decision engine, and a safety constraint-driven actuator.
[0010] Monitoring and data collection adaptation layer: used to collect indicator data from heterogeneous data sources and aggregate the collected indicator data into the monitoring database.
[0011] The security policy adapter is used to retrieve node security label mapping rules, application security configuration file templates, and network policy restrictions from the central policy repository at preset intervals; after loading the retrieved data, the security policy adapter caches it locally and subscribes to change notifications.
[0012] The multi-factor decision engine dynamically determines the K-ELI based on the indicator values of each infrastructure and each nuclear industry business. Based on the calculated K-ELI and the strategy library, it determines the expansion and contraction operations to be sent to the safety constraint drive actuator. The expansion and contraction operations include expansion, contraction, or no operation, and also include the basic step size for expansion and contraction. The strategy library contains the mapping relationship between different K-ELIs and expansion and contraction suggestions.
[0013] The safety constraint-driven actuator determines the final change in the number of instances of a node using the following formula based on the base step size in the received scaling operation, and adjusts the number of instances of the node based on the final change in the number of instances:
[0014] ΔN1=min(ΔNbase,Smax-Ncurrent);
[0015] ΔM=round(ΔN1×Trend_Correction),N min ≤(ΔN+Ncurrent)≤N max ;
[0016] ΔNbase is the base step size, Smax is the maximum number of compliant nodes that can support instances, and Ncurrent is the current number of instances in the node; N min N is the preset minimum number of node instances. maxThe maximum number of nodes is preset; Trend_Correction is the trend correction factor, and Trend_Correction has a preset value range.
[0017] In one possible implementation, the multi-factor decision engine obtains the latest indicator data from the monitoring database, preprocesses the obtained indicator data, and determines the nuclear industry resilience load index K-ELI based on the preprocessed indicator data using the following formula:
[0018]
[0019] F(·) is a comprehensive calculation function used to perform a weighted summation of the terms of the input function; Winfra i For the i-th infrastructure, InfraMetric i Let N be the weight of the i-th infrastructure, and N be the total number of infrastructures; Wbiz j Let BizMetric be the j-th nuclear industry business indicator value. j is the weight of the j-th nuclear industry business indicator; M is the total amount of nuclear industry business; Security_Factor is the safety compliance factor.
[0020] In one possible implementation, the pre-defined mapping relationships between different K-ELI values and scaling suggestions in the strategy library include: when 0 ≤ K-ELI < 0.4, the scaling operation type is shrinking with a base step size of -1; when 0.4 ≤ K-ELI ≤ 1.2, the scaling operation type is no operation with a base step size of 0; when 1.2 < K-ELI ≤ 1.6, the scaling operation type is expanding with a base step size of 1; and when 1.6 < K-ELI ≤ 2.0, the scaling operation type is emergency expanding with a base step size of 2.
[0021] In one possible implementation, the scaling operations performed by the security constraint-driven executor include Pod-level scaling and process group-level scaling;
[0022] Pod-level scaling includes: calling the container engine's API, adjusting the number of Deployment / StatefulSet replicas, and injecting the security context obtained from the application security configuration file template into the Pod Spec;
[0023] Process group scaling includes sending HTTP commands to a pre-installed lightweight agent within the target Pod. The agent is responsible for starting or stopping a specified number of worker processes and synchronously updating the Cgroup resource limits of the process group.
[0024] In one possible implementation, the safety constraint-driven actuator employs a rolling update strategy during the scaling operation.
[0025] In one possible implementation, the system further includes a circuit breaker coordination module that subscribes to microservice circuit breaker state change messages.
[0026] When the circuit breaker coordination module changes the circuit breaker status to Open or Half-Open, it freezes all scaling-down operations of the corresponding service.
[0027] After the circuit breaker coordination module returns to Closed status, it unfreezes the elasticity policy of the corresponding service.
[0028] In one possible implementation, the monitoring and data acquisition adaptation layer is configured to perform the following metrics data acquisition tasks:
[0029] Deploy a customized Node-Exporter to collect hardware and operating system metrics of the infrastructure;
[0030] The deployment adapter reads performance data specific to the domestic operating system and converts the read performance data into data in a standard indicator format;
[0031] Collect container-level metrics by extending Metric-Exporter or calling the iSula / Pouch SDK;
[0032] Collect key business metrics for nuclear industry applications, including: embedding tracking points at key locations in the application code to collect and output the application's business metrics; using a sidecar to deploy an auxiliary container for the application container, and non-intrusively collecting the application container's business metrics by intercepting logs and analyzing traffic; and analyzing the system logs output by the application to extract key numerical information as business metrics.
[0033] The beneficial effects of this disclosure are as follows: It provides precise elastic response: The system adopts the K-ELI model to integrate business indicators, security factors, and dynamic weights, resulting in high business sensitivity and decisions that better meet actual needs. This significantly improves the accuracy, timeliness, and effectiveness of scaling decisions, effectively addressing load fluctuations in the nuclear industry. The security policy adapter performs real-time verification and constraints before decision-making and execution, ensuring precise injection of security context. It resolves conflicts between security policies and scaling, and the system is fully compatible with monitoring interfaces, APIs, and security mechanisms of various CPUs, OSs, and container engines, optimized for platform characteristics. It supports process group-level elastic scaling. It meets the refined resource control needs of special modules in nuclear industry scenarios, improving resource utilization. Attached Figure Description
[0034] Figure 1 This is a schematic diagram illustrating a flexible expansion / contraction system for nuclear industry applications according to an exemplary embodiment. Detailed Implementation
[0035] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0036] Figure 1 This is a schematic diagram illustrating a flexible expansion / contraction system for nuclear industry applications according to an exemplary embodiment, such as... Figure 1 As shown, the system includes: a monitoring and acquisition adaptation layer, a security policy adapter, a multi-factor decision engine (setting a policy library for determining K-ELI), a security constraint-driven executor, and a circuit breaker coordination unit. The target application runs on a cluster composed of domestically produced chips and an OS host.
[0037] Monitoring and data collection adaptation layer: used to collect metric data from heterogeneous data sources and aggregate the collected metric data into a monitoring database (such as Prometheus).
[0038] As an example of this embodiment, the monitoring and data acquisition adaptation layer is configured to perform the following metric data acquisition tasks:
[0039] Deploy customized Node-Exporters to collect hardware and operating system metrics of infrastructure (such as physical machines or virtual machines) (e.g., CPU utilization, load, core status, interrupt count; total memory, usage, cache, swap partition status of Mem; disk read / write speed, IOPS, latency, and partition utilization of I / O interfaces; network bandwidth usage, packet transmission rate, error count, and connection count) to ensure the stability and sufficient resources of the infrastructure supporting the business.
[0040] This deployment adapter reads performance data specific to domestic operating systems and converts it into standard indicator formats. Domestic operating systems such as Kylin / Tongxin UOS provide unique performance data interfaces or files (e.g., specific files under / proc or / sys, or output from proprietary performance tools) through their kernels or system tools. This disclosure uses an adapter to parse this unique performance data and convert it into standard indicator formats. This fills monitoring blind spots, ensuring comprehensive monitoring in domestic alternative environments.
[0041] Collect container-level metrics (CPU, memory, disk I / O usage, and runtime status of a single container) by extending Metric-Exporter or calling the iSula / Pouch SDK.
[0042] This involves collecting key business metrics for nuclear industry applications, including: embedding tracking points at critical locations in application code to collect and output application business metrics; using a sidecar to deploy auxiliary containers for application containers, non-intrusively collecting application container business metrics by intercepting logs and analyzing traffic; and analyzing system logs output by the application to extract key numerical information as business metrics. This provides operations and engineering personnel with a complete and observable view from hardware anomalies to business disruptions, ensuring the safe, stable, and efficient operation of nuclear industry systems. This is not merely technical monitoring, but a core infrastructure for ensuring the security of critical business operations.
[0043] The security policy adapter is used to retrieve node security label mapping rules (such as {label combination} <-> security level), application security configuration file templates (such as SELinux context, Seccomp profile), and network policy restrictions from the central policy database (CMDB or security platform) at preset intervals. After loading the retrieved data, the security policy adapter caches it locally and subscribes to change notifications.
[0044] The multi-factor decision engine dynamically determines the K-ELI based on the indicator values of each infrastructure and each nuclear industry operation. Based on the calculated K-ELI and a policy library, it determines the expansion / contraction operation and sends it to the safety constraint-driven actuator. The expansion / contraction operation type includes expansion, contraction, or no operation, and also includes the base step size for expansion and contraction. The policy library contains preset mapping relationships between different K-ELIs and expansion / contraction suggestions. In one possible implementation, the K-ELI can be the sum of the indicator values of each infrastructure and each nuclear industry operation.
[0045] In another possible implementation, the multi-factor decision engine obtains the latest indicator data from the monitoring database, preprocesses the obtained indicator data (e.g., filtering, normalization), and determines the nuclear industry resilience load index K-ELI based on the preprocessed indicator data using the following formula:
[0046]
[0047] F(·) is a comprehensive calculation function used to perform a weighted summation of the terms of the input function; Winfra i For the i-th infrastructure, InfraMetric i Let N be the weight of the i-th infrastructure, and N be the total number of infrastructures. Infrastructure metrics are used to represent resource metrics of infrastructure such as CPU / containers; Wbiz j Let BizMetric be the j-th nuclear industry business indicator value. jis the weight of the j-th nuclear industry business indicator; M is the total amount of nuclear industry business; Security_Factor is the safety compliance factor, whose value range can be (0~1], used to quantify the degree of compliance; Trend_Correction is the trend correction factor, whose value range can be (0.8~1.2), which can be adjusted based on load forecast.
[0048] The multi-factor decision engine queries the elastic strategy template library, determines the corresponding scaling operation and base step size based on the range of K-ELI values, and sends it to the safety constraint-driven actuator. For example, when K-ELI is in the range [0, 0.4), the scaling operation type is shrinking, and the base step size is -1; when K-ELI is in the range [0.4, 1.2], the scaling operation type is no operation, and the base step size is 0; when K-ELI is in the range (1.2, 1.6), the scaling operation type is expanding, and the base step size is 1; when K-ELI is in the range (1.6, 2.0), the scaling operation type is emergency expansion, and the base step size is 2.
[0049] The safety constraint-driven actuator determines the final change in the number of instances of a node using the following formula based on the base step size in the received scaling operation, and adjusts the number of instances of the node based on the final change in the number of instances.
[0050] ΔN1=min(ΔNbase,Smax-Ncurrent);
[0051] ΔN=round(ΔN1×Trend_Correction),N min ≤(ΔN+Ncurrent)≤N max ;
[0052] ΔNbase is the base step size, Smax is the maximum number of compliant nodes that can support instances, and Ncurrent is the current number of instances in the node; N min N is the preset minimum number of node instances. max The preset maximum number of node instances; Trend_Correction is randomly selected within the range of (0.8 to 1.2), such that N min ≤(ΔN+Ncurrent)≤N max In this way, this disclosure fully considers security and compliance requirements, ensuring that the change in the number of node instances is consistent with the actual capacity of the node.
[0053] In one possible implementation, the scaling operations performed by the security constraint-driven executor include Pod-level scaling and process group-level scaling. Pod-level scaling involves: calling the container engine's API, adjusting the number of Deployment / StatefulSet replicas, and injecting scaling operation content (securityContext) obtained from the application security configuration file template into the Pod template.
[0054] Process group scaling includes sending HTTP commands to a pre-installed lightweight agent within the target Pod. The agent is responsible for starting or stopping a specified number of worker processes and synchronously updating the Cgroup resource limits (such as CPU share) of the process group.
[0055] In one possible implementation, the safety constraint-driven actuator employs a rolling update strategy during the scaling operation to ensure smooth service.
[0056] In one possible implementation, the system disclosed herein further includes a circuit breaker coordination module. During the scaling operation performed by the security constraint driver, if the module detects that a node has entered a circuit breaker state, it sends a signal to the security constraint driver to freeze all scaling operations until the node exits the circuit breaker state. The circuit breaker coordination module subscribes to microservice circuit breaker state change messages. When the circuit breaker state changes to Open or Half-Open, the module freezes all scaling operations of the corresponding service. After the circuit breaker state returns to Closed, the module unfreezes the elasticity policy of the corresponding service.
[0057] Compared with the prior art, this application has the following beneficial effects:
[0058] Precise and Resilient Response: The system disclosed herein employs a K-ELI model that integrates business indicators, safety factors, and dynamic weights, resulting in high business sensitivity and decisions that better align with actual needs. This significantly improves the accuracy, timeliness, and effectiveness of capacity expansion and contraction decisions, effectively addressing load fluctuations in the nuclear industry.
[0059] Embedded security safeguards: The security policy adapter performs real-time verification and constraints before decision-making and execution, and accurately injects security context. This resolves conflicts between security policies and scaling up / down operations, ensuring 100% compliance with nuclear industry safety baselines and preventing service interruptions or vulnerabilities due to security compliance issues.
[0060] Deep adaptation and performance optimization: Full-link adaptation to monitoring interfaces, APIs and security mechanisms of various types of CPUs, OSs and container engines, and optimization for platform characteristics.
[0061] Fine-grained resource control: Supports elastic scaling at the process group level. Meets the refined resource control requirements of special modules in nuclear industry scenarios, improving resource utilization.
[0062] Intelligent collaboration enhances stability: The circuit breaker coordination unit proactively freezes relevant scaling-down operations. This avoids inappropriate scaling-down during service capacity degradation (circuit breaker) periods, ensuring service continuity.
[0063] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.
Claims
1. A flexible expansion / contraction system for nuclear industry applications, characterized in that, The system includes: a monitoring and acquisition adaptation layer, a security policy adapter, a multi-factor decision engine, and a security constraint-driven executor; Monitoring data collection and adaptation layer: used to collect indicator data from heterogeneous data sources and aggregate the collected indicator data into the monitoring database; The security policy adapter is used to retrieve node security label mapping rules, application security configuration file templates, and network policy restrictions from the central policy repository at preset intervals; after loading the retrieved data, the security policy adapter caches it locally and subscribes to change notifications. The multi-factor decision engine dynamically determines the K-ELI based on the indicator values of each infrastructure and each nuclear industry business. Based on the calculated K-ELI and the strategy library, it determines the expansion and contraction operations to be sent to the safety constraint drive actuator. The expansion and contraction operations include expansion, contraction, or no operation, and also include the basic step size for expansion and contraction. The strategy library contains the mapping relationship between different K-ELIs and expansion and contraction suggestions. The safety constraint-driven actuator determines the final change in the number of instances of a node using the following formula based on the base step size in the received scaling operation, and adjusts the number of instances of the node based on the final change in the number of instances: ΔN1=min(ΔNbase,Smax-Ncurrent); ΔN=round(ΔN1×Trend_Correction),N min ≤(ΔN+Ncurrent)≤N max ; ΔNbase is the base step size, Smax is the maximum number of compliant nodes that can support instances, and Ncurrnet is the current number of instances in the node; N min N is the preset minimum number of node instances. max The maximum number of nodes is preset; Trend_Correction is the trend correction factor, and Trend_Correction has a preset value range.
2. The system according to claim 1, characterized in that, The multi-factor decision engine obtains the latest indicator data from the monitoring database, preprocesses the obtained indicator data, and determines the nuclear industry resilience load index K-ELI based on the preprocessed indicator data using the following formula: F(·) is a comprehensive calculation function used to perform a weighted summation of the terms of the input function; Winfra i For the i-th infrastructure, InfraMetric i Let N be the weight of the i-th infrastructure, and N be the total number of infrastructures; Wbiz j Let BizMetric be the j-th nuclear industry business indicator value. j is the weight of the j-th nuclear industry business indicator; M is the total amount of nuclear industry business; Security_Factor is the safety compliance factor.
3. The system according to claim 1, characterized in that, The pre-defined mapping relationships between different K-ELI values and scaling suggestions in the strategy library are as follows: when 0 ≤ K-ELI < 0.4, the scaling operation type is shrinking with a base step size of -1; when 0.4 ≤ K-ELI ≤ 1.2, the scaling operation type is no operation with a base step size of 0; when 1.2 < K-ELI ≤ 1.6, the scaling operation type is expanding with a base step size of 1; and when 1.6 < K-ELI ≤ 2.0, the scaling operation type is emergency expanding with a base step size of 2.
4. The system according to claim 1, characterized in that, Safety constraint-driven scaling operations include Pod-level scaling and process group-level scaling. Pod-level scaling includes: calling the container engine's API, adjusting the number of Deployment / StatefulSet replicas, and injecting the security context obtained from the application security configuration file template into the PodSpec; Process group scaling includes sending HTTP commands to a pre-installed lightweight agent within the target Pod. The agent is responsible for starting or stopping a specified number of worker processes and synchronously updating the Cgroup resource limits of the process group.
5. The system according to claim 1, characterized in that, The safety constraint-driven actuator employs a rolling update strategy during the scaling operation.
6. The system according to claim 1, characterized in that, The system also includes a circuit breaker coordination module, which subscribes to microservice circuit breaker status change messages. When the circuit breaker coordination module changes the circuit breaker status to Open or Half-Open, it freezes all scaling-down operations of the corresponding service. After the circuit breaker coordination module returns to Closed status, it unfreezes the elasticity policy of the corresponding service.
7. The system according to claim 1, characterized in that, The monitoring and data acquisition adaptation layer is configured to perform the following metrics data acquisition tasks: Deploy a customized Node-Exporter to collect hardware and operating system metrics of the infrastructure; The deployment adapter reads performance data specific to the domestic operating system and converts the read performance data into data in a standard indicator format; Collect container-level metrics by extending Metric-Exporter or calling the iSula / Pouch SDK; Collect key business metrics for nuclear industry applications, including: embedding tracking points at key locations in the application code to collect and output the application's business metrics; using a sidecar to deploy an auxiliary container for the application container, and non-intrusively collecting the application container's business metrics by intercepting logs and analyzing traffic; and analyzing the system logs output by the application to extract key numerical information as business metrics.
