A method and system for node expulsion based on hierarchical analysis method

Abstract

The present invention proposes a method and system for node eviction based on the Analytic Hierarchy Process (AHP), belonging to the field of node optimization technology. Step S1: Using Pinpoint or Zipkin to track the business links of running software, a business link call graph is drawn based on the business links. Step S2: Dynamically evaluate the current cluster business line based on the AHP, and construct a comprehensive judgment matrix based on the evaluation scores. The comprehensive judgment matrix is ​​used to determine the weights of each business included in the current cluster business line. Step S3: Based on the weights of each business, a three-color marking method is used to mark the node containers in the business link call graph, and node eviction is performed based on the markings.

Description

Technical Field

[0001] The present invention belongs to the technical field of node optimization, and in particular relates to a method and system for node expulsion based on a hierarchical analysis method. Background Art

[0002] Node pressure eviction is the process by which the kubelet proactively terminates pods to reclaim resources on a node. The kubelet monitors resources such as memory, disk space, and file system inodes on cluster nodes. When one or more of these resources reaches a certain level of consumption, the kubelet begins selectively evicting pods. The kubelet configuration file configures a minimum reclaim amount for each resource. When the kubelet notices that a resource is exhausted, it continues to reclaim the resource until the amount you specify is reclaimed. If a node has 10Gi of total memory and you want to trigger eviction when the available memory falls below 1Gi, define the eviction condition as memory.available < 10% or memory.available < 1G.

[0003] If the kubelet's attempts to reclaim node-level resources do not bring the eviction signal below the condition, the kubelet begins evicting user Pods. The kubelet uses the following parameters to determine the order of Pod eviction: (1) whether the Pod's resource usage exceeds its request; (2) the Pod priority; and (3) the Pod's resource usage relative to the request. Therefore, the kubelet arranges and evicts Pods in the following order: BestEffort or Burstable Pods whose resource usage exceeds their request are considered first. These Pods are evicted based on their priority and the extent to which their resource usage level exceeds their request. Guaranteed Pods and Burstable Pods whose resource usage is less than the requested amount are then evicted last based on their priority.

[0004] The current default Kubernetes eviction algorithm cannot guarantee the stable operation of core services. In the worst case, it can cause all services to cease functioning, for example, by evicting a database node or a security service node. While this achieves the goal of node eviction and resource recovery, it can also cause the entire service to cease functioning, a severe blow to any company or organization. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes a solution for node eviction based on hierarchical analysis method.

[0006] The first aspect of the present invention discloses a method for node eviction based on the Analytic Hierarchy Process (AHP). The method comprises: step S1, using Pinpoint or Zipkin to track the business links of related software in operation, and drawing a business link call graph based on the business links; step S2, dynamically evaluating the current cluster business line based on the AHP, and constructing a comprehensive judgment matrix based on the evaluation scores; wherein the comprehensive judgment matrix is ​​used to determine the weights of each business included in the current cluster business line; step S3, using a three-color marking method to mark node containers in the business link call graph based on the weights of each business, and performing node eviction based on the markings; wherein the node containers corresponding to businesses exceeding a weight threshold are marked with a first color, and other node containers that reach the node container marked with the first color through a directed link are determined by reverse traversal, and the other node containers are marked with a second color, and node containers that do not belong to the first and second colors are marked with a third color; wherein, when performing node eviction, the node containers marked with the third color are preferentially evicted.

[0007] According to the method of the first aspect of the present invention, in step S2, the current cluster business line includes: taxi business, shopping mall business, user dynamic business and coupon business, and the dynamic evaluation indicators of each business include revenue, user experience, user activity and user evaluation.

[0008] According to the method of the first aspect of the present invention, in step S2, an evaluation index judgment matrix between each indicator is constructed; an evaluation score matrix of each business relative to each indicator is constructed; the evaluation index judgment matrix and each evaluation score matrix are integrated to obtain the comprehensive judgment matrix; thereby determining the weight of each business included in the current cluster business line.

[0009] According to the method of the first aspect of the present invention, in step S3, the k8s nodes and system nodes in the business link call graph are marked with the first color, and the node containers, k8s nodes and system nodes marked with the first color are all nodes that will not be evicted.

[0010] A second aspect of the present invention discloses a system for node eviction based on the Analytic Hierarchy Process (AHP). The system comprises: a first processing unit configured to: use Pinpoint or Zipkin to track the business links of related software in operation, and draw a business link call graph based on the business links; a second processing unit configured to: dynamically evaluate the current cluster business line based on the AHP, and construct a comprehensive judgment matrix based on the evaluation scores; wherein the comprehensive judgment matrix is ​​used to determine the weights of each business included in the current cluster business line; and a third processing unit configured to: use a three-color marking method to mark node containers in the business link call graph based on the weights of each business, and perform node eviction based on the markings; wherein the node container corresponding to the business exceeding the weight threshold is marked with a first color, and other node containers that reach the node container marked with the first color through a directed link are determined by reverse traversal, and the other node containers are marked with a second color, and node containers that do not belong to the first color or the second color are all marked with a third color; wherein, when performing node eviction, the node container marked with the third color is preferentially evicted.

[0011] According to the system of the second aspect of the present invention, the second processing unit is specifically configured as follows: the various businesses included in the current cluster business line are: taxi business, shopping mall business, user dynamic business and coupon business, and the dynamic evaluation indicators of each business include revenue, user experience, user activity and user evaluation.

[0012] According to the system of the second aspect of the present invention, the second processing unit is specifically configured to: construct an evaluation index judgment matrix between each indicator; construct an evaluation score matrix of each business relative to each indicator; integrate the evaluation index judgment matrix and each evaluation score matrix to obtain the comprehensive judgment matrix; thereby determining the weight of each business included in the current cluster business line.

[0013] According to the system of the second aspect of the present invention, the third processing unit is specifically configured to: mark the k8s nodes and system nodes in the business link call graph with the first color, and the node containers, k8s nodes and system nodes marked with the first color are all nodes that will not be evicted.

[0014] A third aspect of the present invention discloses an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the steps of the method for node eviction based on the hierarchical analysis method described in the first aspect of the present disclosure.

[0015] A fourth aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for node eviction based on the hierarchical analysis method described in the first aspect of the present disclosure.

[0016] In summary, the technical solution proposed in the present invention uses pinpoint or zipkin to track the entire link and draw a link call graph. After the relevant software has been running for a period of time, it can draw the relevant business lines of all businesses through tracking. The present invention classifies and analyzes the businesses in the cluster, retains all system pods, and then uses the tomographic analysis method to determine the priority of the business. In the case of node eviction and insufficient resources, unimportant businesses are expelled first, and eviction is carried out according to business lines to ensure the stable operation of important business lines. For example, when cluster resources are scarce, taxi services give priority to expelling businesses such as user dynamics to ensure the operation of taxi services. When evicting nodes, the present invention uses a three-color marking method to evict business nodes with lower priority. The present invention uses pinpoint or zipkin to track the entire link and draw a link call graph. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0019] Figure 2 A service link diagram according to an embodiment of the present invention;

[0020] Figure 3 is an evaluation index weight diagram according to an embodiment of the present invention;

[0021] Figure 4 is a weight diagram of a benefit judgment matrix according to an embodiment of the present invention;

[0022] Figure 5 A weight diagram of a user experience judgment matrix according to an embodiment of the present invention;

[0023] Figure 6 is a weight diagram of a user activity judgment matrix according to an embodiment of the present invention;

[0024] Figure 7 A weight diagram of a user evaluation judgment matrix according to an embodiment of the present invention;

[0025] Figure 8 is a weight diagram of a comprehensive matrix according to an embodiment of the present invention;

[0026] Figure 9 FIG. 4 is a structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0028] The first aspect of the present invention discloses a method for node eviction based on the Analytic Hierarchy Process (AHP). The method comprises: step S1, using Pinpoint or Zipkin to track the business links of related software in operation, and drawing a business link call graph based on the business links; step S2, dynamically evaluating the current cluster business line based on the AHP, and constructing a comprehensive judgment matrix based on the evaluation scores; wherein the comprehensive judgment matrix is ​​used to determine the weights of each business included in the current cluster business line; step S3, using a three-color marking method to mark node containers in the business link call graph based on the weights of each business, and performing node eviction based on the markings; wherein the node containers corresponding to businesses exceeding a weight threshold are marked with a first color, and other node containers that reach the node container marked with the first color through a directed link are determined by reverse traversal, and the other node containers are marked with a second color, and node containers that do not belong to the first and second colors are marked with a third color; wherein, when performing node eviction, the node containers marked with the third color are preferentially evicted.

[0029] In step S1, the present invention uses pinpoint or zipkin to track the entire link and draw a link call graph. After the relevant software has been running for a period of time, it can draw the relevant business lines of all businesses through tracking, such as Figure 2 shown.

[0030] In some embodiments, in step S2, the current cluster business line includes various businesses: taxi business, shopping mall business, user dynamic business and coupon business, and the dynamic evaluation indicators of each business include revenue, user experience, user activity and user evaluation.

[0031] In some embodiments, in step S2, an evaluation index judgment matrix between each indicator is constructed; an evaluation score matrix of each business relative to each indicator is constructed; the evaluation index judgment matrix and each evaluation score matrix are integrated to obtain the comprehensive judgment matrix; thereby determining the weight of each business included in the current cluster business line.

[0032] Specifically, the Analytic Hierarchy Process (AHP) refers to a decision-making method that decomposes decision-making elements into a hierarchy of goals, criteria, and options, and then conducts qualitative and quantitative analysis based on these factors. This method was proposed by American operations researcher and University of Pittsburgh professor Satti in the early 1970s while researching the "distribution of electricity based on the contribution of various industrial sectors to national welfare" for the US Department of Defense. He applied network systems theory and multi-objective comprehensive evaluation methods to develop a hierarchical weighted decision analysis approach.

[0033] Use the Analytic Hierarchy Process (AHP) to dynamically score the current cluster's business lines. These include ride-hailing, shopping malls, user activity, and coupons. This business reaches a bottleneck during rush hour, and memory usage is lower than the configured capacity. Evaluation metrics for this business include revenue, user experience, user activity, and user reviews.

[0034] Construct the current matrix: The judgment matrix represents the relative importance of all factors in the current layer relative to a factor in the previous layer. The elements of the judgment matrix are given using Saaty's 1-9 scale.

[0035] The Analytic Hierarchy Process (AHP) is a systematic, hierarchical analytical method that combines qualitative and quantitative analysis. This method, based on in-depth research into the nature, influencing factors, and their inherent relationships, uses limited quantitative information to mathematically analyze the decision-making process. This provides a simple approach for complex decision-making problems involving multiple objectives, multiple criteria, or unstructured issues. It is a model and method for making decisions in complex systems that are difficult to fully quantify.

[0036] The steps of the hierarchical analysis method can be roughly divided into the following four steps when using the hierarchical analysis method to construct a system model: establishing a hierarchical structure model; constructing a judgment (paired comparison) matrix; hierarchical single sorting and consistency test; hierarchical total sorting and consistency test.

[0037] Evaluation index judgment matrix

[0038] The current business evaluation metrics include revenue, user experience, user activity, and user reviews. Constructing the current matrix: A judgment matrix compares the relative importance of all factors at one level relative to a single factor at the previous level. Elements of the judgment matrix are expressed using Saaty's 1-9 scale.

[0039]

[0040] As can be seen from the table above, a 4-order judgment matrix was constructed for revenue, user experience, user activity, and user evaluation, and the AHP hierarchical analysis (calculation method: square root method) was used. The eigenvectors obtained were (2.943, 0.473, 0.669, 1.075), and the corresponding weights of the 4 items were 57.043%, 9.166%, 12.963%, and 20.829%, respectively. In addition, the maximum eigenroot (4.065) was calculated based on the eigenvectors. The CI value (0.022) was then calculated using the maximum eigenroot value. CI = (maximum eigenroot - n) / (n - 1). The CI value is used for the consistency test described below. Figure 3 .

[0041] Profit Judgment Matrix

[0042] The current cluster business lines include: taxi, shopping mall, user dynamics, and coupons. The current business evaluation indicators include: revenue, user experience, user activity, and user evaluation. Figure 4 .

[0043]

[0044] User Experience Judgment Matrix

[0045] The current cluster business lines include: taxi, shopping mall, user dynamics, coupons. The current business evaluation indicators include: user experience. Figure 5 , construct the matrix as follows:

[0046]

[0047] User activity judgment matrix

[0048] The current cluster business lines include: taxi, shopping mall, user dynamics, and coupons. The current business evaluation indicators include: user activity. Construct the current matrix:

[0049]

[0050] The judgment matrix is ​​a comparison of the relative importance of all factors in this layer to a factor in the previous layer. Figure 6 .

[0051] User evaluation judgment matrix

[0052] The current cluster business lines include: taxi, shopping mall, user dynamics, and coupons. The current business evaluation indicators include: user evaluation. Construct the current matrix:

[0053]

[0054] The judgment matrix is ​​a comparison of the relative importance of all factors in this layer to a factor in the previous layer. Figure 7 .

[0055] Comprehensive summary

[0056] The current cluster business lines include: taxi, shopping mall, user dynamics, and coupons. The current business evaluation indicators include: revenue, user experience, user activity, and user evaluation. Figure 8 , calculate the weight index of the current business:

[0057]

[0058] In step S3, the node containers in the service link call graph are marked using a three-color marking method according to the weights of the various services, and nodes are evicted based on the markings.

[0059] Among them, the node container corresponding to the business exceeding the weight threshold is marked with a first color, and other node containers that reach the node container marked with the first color through a directed link are determined by reverse traversal. The other node containers are marked with a second color, and the node containers that do not belong to the first color and the second color are all marked with a third color.

[0060] When executing the node eviction, the node container marked with the third color is first evicted.

[0061] In some embodiments, in step S3, the k8s nodes and system nodes in the business link call graph are marked with the first color, and the node containers, k8s nodes, and system nodes marked with the first color are all nodes that will not be evicted.

[0062] Node eviction uses a three-color marking method, called Mark-and-Sweep. This algorithm marks all containers on the current link. High-priority container links that don't need to be evicted are marked black, while containers reachable by the current link are marked gray. After marking is complete, the remaining pods are marked white. White pods indicate links that require eviction. Kubernetes' own nodes and critical system nodes are marked directly black. When resources are insufficient, Kubernetes begins evicting nodes, first based on the rankings and user activity.

[0063] Mark the entry nodes for ride-hailing, shopping malls, and coupons in black. Then, based on the request link, mark the reachable pods in gray. Finally, clear all unmarked pods. This prioritizes clearing user activity when resources are insufficient.

[0064] This invention avoids the situation where multiple business lines become unavailable due to the default eviction policy of K8s. Even when resource defaults are exceeded, the operation of the main business is still guaranteed. This invention can also achieve dynamic adjustment. For example, if a company's main taxi business stagnates and is replaced by social business, user dynamics can be used through the hierarchical analysis method to increase the priority of the taxi business and clear it out.

[0065] A second aspect of the present invention discloses a system for node eviction based on the Analytic Hierarchy Process (AHP). The system comprises: a first processing unit configured to: use Pinpoint or Zipkin to track the business links of related software in operation, and draw a business link call graph based on the business links; a second processing unit configured to: dynamically evaluate the current cluster business line based on the AHP, and construct a comprehensive judgment matrix based on the evaluation scores; wherein the comprehensive judgment matrix is ​​used to determine the weights of each business included in the current cluster business line; and a third processing unit configured to: use a three-color marking method to mark node containers in the business link call graph based on the weights of each business, and perform node eviction based on the markings; wherein the node container corresponding to the business exceeding the weight threshold is marked with a first color, and other node containers that reach the node container marked with the first color through a directed link are determined by reverse traversal, and the other node containers are marked with a second color, and node containers that do not belong to the first color or the second color are all marked with a third color; wherein, when performing node eviction, the node container marked with the third color is preferentially evicted.

[0066] In some embodiments, the second processing unit is specifically configured as follows: the current cluster business line includes: taxi business, shopping mall business, user dynamic business and coupon business, and the dynamic evaluation indicators of each business include revenue, user experience, user activity and user evaluation.

[0067] In some embodiments, the second processing unit is specifically configured to: construct an evaluation indicator judgment matrix between each indicator; construct an evaluation score matrix for each business relative to each indicator; integrate the evaluation indicator judgment matrix and each evaluation score matrix to obtain the comprehensive judgment matrix; thereby determining the weight of each business included in the current cluster business line.

[0068] In some embodiments, the third processing unit is specifically configured to: mark the k8s nodes and system nodes in the business link call graph as the first color, and the node containers, k8s nodes and system nodes marked with the first color are all nodes that will not be evicted.

[0069] A third aspect of the present invention discloses an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the steps of the method for node eviction based on the hierarchical analysis method described in the first aspect of the present disclosure.

[0070] Figure 9 FIG. 1 is a structural diagram of an electronic device according to an embodiment of the present invention. Figure 9 As shown, the electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the electronic device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, near field communication (NFC) or other technologies. The display screen of the electronic device can be a liquid crystal display or an electronic ink display screen, and the input device of the electronic device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the electronic device housing, or an external keyboard, touchpad or mouse.

[0071] Those skilled in the art will understand that Figure 9 The structure shown in the figure is only a structural diagram of the part related to the technical solution of the present disclosure, and does not constitute a limitation on the electronic device to which the solution of the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0072] A fourth aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for node eviction based on the hierarchical analysis method described in the first aspect of the present disclosure.

[0073] In summary, the technical solution proposed in the present invention uses pinpoint or zipkin to track the entire link and draw a link call graph. After the relevant software has been running for a period of time, it can draw the relevant business lines of all businesses through tracking. The present invention classifies and analyzes the businesses in the cluster, retains all system pods, and then uses the tomographic analysis method to determine the priority of the business. In the case of node eviction and insufficient resources, unimportant businesses are expelled first, and eviction is carried out according to business lines to ensure the stable operation of important business lines. For example, when cluster resources are scarce, taxi services give priority to expelling businesses such as user dynamics to ensure the operation of taxi services. When evicting nodes, the present invention uses a three-color marking method to evict business nodes with lower priority. The present invention uses pinpoint or zipkin to track the entire link and draw a link call graph.

[0074] Please note that the various technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above-mentioned embodiments only express several implementation methods of the present application. The description is relatively specific and detailed, but it cannot be understood as a limitation on the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of this application, several variations and improvements can be made, which all fall within the scope of protection of this application. Therefore, the scope of protection of the patent in this application shall be based on the attached claims.

Claims

1. A method for node expulsion based on hierarchical analysis method, characterized in that: The method comprises: Step S1: Use Pinpoint or Zipkin to track the business links of the relevant software in the running state, and draw a business link call graph based on the business links; Step S2: Dynamically evaluate the current cluster business line based on the analytic hierarchy process, and construct a comprehensive judgment matrix based on the evaluation scores; The comprehensive judgment matrix is ​​used to determine the weight of each business included in the current cluster business line; In step S2, an evaluation index judgment matrix between the various indicators is constructed; an evaluation score matrix of the various businesses relative to each indicator is constructed; the evaluation index judgment matrix and the various evaluation score matrices are integrated to obtain the comprehensive judgment matrix; thereby determining the weights of the various businesses included in the current cluster business line; Step S3: Marking the node containers in the service link call graph using a three-color marking method according to the weight of each service, and performing node eviction based on the marking; The node container corresponding to the service exceeding the weight threshold is marked with a first color. Other node containers that reach the node container marked with the first color through directed links are determined by reverse traversal. The other node containers are marked with a second color. Node containers that do not belong to the first color or the second color are all marked with a third color. When executing the node eviction, the node container marked with the third color is first evicted.

2. The method for node expulsion based on hierarchical analysis method according to claim 1, characterized in that: In step S2, the current cluster business line includes the following businesses: taxi-hailing business, shopping mall business, user dynamics business and coupon business. The dynamic evaluation indicators of each business include revenue, user experience, user activity and user evaluation.

3. The method for node expulsion based on hierarchical analysis method according to claim 2, characterized in that: In step S3, the k8s nodes and system nodes in the business link call graph are marked with the first color, and the node containers, k8s nodes and system nodes marked with the first color are all nodes that will not be evicted.

4. A system for node eviction based on hierarchical analysis method, characterized in that: The system comprises: The first processing unit is configured to: use Pinpoint or Zipkin to track the business links of the relevant software in the running state, and draw a business link call graph based on the business links; The second processing unit is configured to: dynamically evaluate the current cluster business line based on the hierarchical analysis method and construct a comprehensive judgment matrix according to the evaluation scores; The comprehensive judgment matrix is ​​used to determine the weight of each business included in the current cluster business line; The second processing unit is specifically configured to: construct an evaluation indicator judgment matrix between various indicators; construct an evaluation score matrix for each business relative to each indicator; integrate the evaluation indicator judgment matrix and various evaluation score matrices to obtain the comprehensive judgment matrix; thereby determining the weight of each business included in the current cluster business line; The third processing unit is configured to: mark the node containers in the service link call graph using a three-color marking method according to the weight of each service, and evict nodes based on the marking; The node container corresponding to the service exceeding the weight threshold is marked with a first color. Other node containers that reach the node container marked with the first color through directed links are determined by reverse traversal. The other node containers are marked with a second color. Node containers that do not belong to the first color or the second color are all marked with a third color. When executing the node eviction, the node container marked with the third color is first evicted.

5. The system for node eviction based on hierarchical analysis method according to claim 4, characterized in that: The second processing unit is specifically configured as follows: the current cluster business line includes: taxi business, shopping mall business, user dynamic business and coupon business, and the dynamic evaluation indicators of each business include revenue, user experience, user activity and user evaluation.

6. The system for node eviction based on hierarchical analysis method according to claim 5, characterized in that: The third processing unit is specifically configured to: mark the k8s nodes and system nodes in the business link call graph as the first color, and the node containers, k8s nodes and system nodes marked with the first color are all nodes that will not be evicted.

7. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it implements the method for node eviction based on hierarchical analysis method according to any one of claims 1 to 3.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for performing node eviction based on the hierarchical analysis method according to any one of claims 1 to 3 is implemented.

Images