CN115391048A - Micro-service instance dynamic horizontal expansion and contraction method and system based on trend prediction - Google Patents

Abstract

The disclosure provides a method and system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction, which belongs to the field of cloud computing technology. The solution includes: periodically obtaining the load of the host machine and microservice instances in the cloud platform based on the Exporter process indicator; wherein, the Exporter process is an indicator collection process existing in the host machine and the microservice instance, and provides an indicator data interface to the outside; the obtained load indicator is stored in the time series database according to time series; based on the time series database Load indicators, use the pre-built load trend prediction model to obtain the load trend of the corresponding service; based on the obtained load trend, realize the expansion or contraction of the microservice instance.

Description

Method and system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction

technical field

The disclosure belongs to the technical field of cloud computing, and in particular relates to a method and system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction.

Background technique

The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

With the rapid development of cloud computing, it has gradually become a trend to develop and deploy business systems based on the microservice architecture in the cloud platform. The microservice architecture splits traditional monolithic applications into independent subsystems, reducing the degree of coupling between functional modules. Container technology packages microservices and their dependent environments into mirrors, and shares and downloads and runs the mirrors through mirror warehouses, which reduces the cost of microservices and improves the flexibility of development and deployment. As the number of containers in the microservice architecture continues to increase, the container cloud platform represented by Kubernetes (k8s) has emerged as the times require. The cloud platform can schedule and manage containers in the cluster, and an orchestration capability is necessary for platform-level projects. It is to realize the "horizontal expansion/contraction" of microservice instances (that is, to adjust the number of microservice instances according to the real load level), and to increase or delete the number of microservice instances according to the real load of the business system, while ensuring service quality At the same time, the allocation of resources on demand is realized.

The inventors found that the most basic way to achieve "horizontal expansion/shrinkage" of microservice instances is that the operation and maintenance personnel make reasonable plans for the future according to the business system and manually set the number of instances. There are differences in the capabilities of operation and maintenance personnel and the complexity of business systems, resulting in unsatisfactory results; in addition to this manual prediction method, responsive scaling can also be achieved by setting specific resource thresholds, such as for microservice instances The CPU usage sets a threshold. When the threshold is exceeded, the number of instances is automatically increased to increase the overall service load capacity. However, there is a lag in scheduling in this way. Dynamic scaling will only occur when the threshold is reached, resulting in Increase service response time, waste system resources and other issues.

Contents of the invention

In order to solve the above problems, the present disclosure provides a method and system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction. The solution collects specific load indicators and uses the indicators to predict the trend characteristics of the load. The horizontal expansion/contraction of microservice instances is realized in advance, which has better foresight and flexibility; at the same time, the scheme can make more efficient use of system resources while ensuring the service quality of microservices.

According to the first aspect of the embodiments of the present disclosure, a method for dynamic horizontal expansion and contraction of microservice instances based on trend prediction is provided, including:

Based on the Exporter process, periodically obtain the load indicators of the host computer and the microservice instance in the cloud platform; wherein, the Exporter process is an indicator collection process existing in the host computer and the microservice instance, and provides an indicator data interface externally;

Store the obtained load indicators in the time series database according to the time series;

Based on the load indicators in the time series database, using a pre-built load trend prediction model to obtain the load trend of the corresponding service;

Based on the obtained load trend, the expansion or contraction of the microservice instance is realized.

Further, the load index includes the hardware index of the host machine, the load utilization rate of the container where the microservice instance is located, and the application layer load index reflecting the corresponding capability of the service.

Further, the hardware index of the host machine and the load utilization rate of the container where the microservice instance is located specifically include CPU utilization rate, memory utilization rate and network IO; the application layer load index reflecting the corresponding capability of the service includes the number of queries per second and response time.

Further, the Exporter process is an API service deployed together with related microservices, collects load indicators internally, and provides query interfaces for external indicator monitoring components.

Further, the load trend prediction model adopts a hybrid model composed of multiple time series prediction algorithms, specifically consisting of Holt-Winters, ARIMA and STL models.

Further, the training of the load trend prediction model is specifically:

Build a training set based on pre-acquired historical load indicator data;

Training several time series prediction models respectively based on the training set;

Based on the SMAPE algorithm, the training results of several time series forecasting models are evaluated, and the optimal model is selected as the final load trend forecasting model.

According to the second aspect of the embodiments of the present disclosure, a system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction is provided, including:

A data acquisition unit, which is used to periodically acquire the load indicators of the host computer and the microservice instance in the cloud platform based on the Exporter process; wherein, the Exporter process is an indicator collection process existing in the host computer and the microservice instance, and provides external Indicator data interface;

A data storage unit, which is used to store the obtained load index in a time series database according to time series;

A load trend prediction unit, which is used to obtain the load trend of the corresponding service based on the load index in the time series database by using a pre-built load trend prediction model;

The microservice instance expansion or contraction control unit is used for realizing the expansion or contraction of the microservice instance based on the obtained load trend.

According to a third aspect of the embodiments of the present disclosure, there is provided an electronic device, including a memory, a processor, and a computer program stored on the memory, and the processor implements the trend-based Predictive method for dynamic horizontal scaling of microservice instances.

According to a fourth aspect of the embodiments of the present disclosure, there is provided a non-transitory computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the aforementioned microservice based on trend prediction is implemented. Instance dynamic horizontal scaling method.

Compared with the prior art, the beneficial effects of the present disclosure are:

(1) This disclosure provides a method and system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction. The solution collects specific load indicators and uses the indicators to predict the trend characteristics of the load. Realizing the horizontal expansion/contraction of microservice instances has better foresight and flexibility; at the same time, the solution can utilize system resources more efficiently while ensuring the service quality of microservices.

(2) In the process of trend prediction through index data, the solution described in this disclosure combines the three algorithms of Holt-Winters, ARIMA and STL, and rationally optimizes the training process to obtain an optimal prediction model; the above-mentioned The three selected algorithms complement each other in terms of use scenarios and advantages and disadvantages. The predicted data fitted by each will be further checked to ensure that the prediction is as accurate as possible.

Advantages of additional aspects of the disclosure will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the disclosure.

Description of drawings

The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure, and the exemplary embodiments and descriptions of the present disclosure are used to explain the present disclosure, and do not constitute improper limitations to the present disclosure.

FIG. 1 is a schematic flowchart of a method for dynamic horizontal expansion and contraction of a microservice instance based on trend prediction described in an embodiment of the present disclosure;

FIG. 2 is a flow chart of indicator collection described in an embodiment of the present disclosure;

FIG. 3 is a schematic diagram of a prediction model training process described in an embodiment of the present disclosure.

Detailed ways

The present disclosure will be further described below in conjunction with the accompanying drawings and embodiments.

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the present disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

It should be noted that the terminology used herein is only for describing specific embodiments, and is not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural, and it should also be understood that when the terms "comprising" and/or "comprising" are used in this specification, they mean There are features, steps, operations, means, components and/or combinations thereof.

In the case of no conflict, the embodiments in the present disclosure and the features in the embodiments can be combined with each other.

Embodiment one:

The purpose of this embodiment is to provide a method for dynamic horizontal expansion and contraction of microservice instances based on trend prediction.

A method for dynamic horizontal expansion and contraction of microservice instances based on trend prediction, including:

Based on the Exporter process, periodically obtain the load indicators of the host computer and the microservice instance in the cloud platform; wherein, the Exporter process is an indicator collection process existing in the host computer and the microservice instance, and provides an indicator data interface externally;

Store the obtained load indicators in the time series database according to the time series;

Based on the load indicators in the time series database, using a pre-built load trend prediction model to obtain the load trend of the corresponding service;

Based on the obtained load trend, the expansion or contraction of the microservice instance is realized.

Further, the load index includes the hardware index of the host machine, the load utilization rate of the container where the microservice instance is located, and the application layer load index reflecting the corresponding capability of the service.

Further, the hardware index of the host machine and the load utilization rate of the container where the microservice instance is located specifically include CPU utilization rate, memory utilization rate and network IO; the application layer load index reflecting the corresponding capability of the service includes the number of queries per second and response time.

Further, the Exporter process is an API service deployed together with related microservices, collects load indicators internally, and provides query interfaces for external indicator monitoring components.

Further, the load trend prediction model adopts a hybrid model composed of multiple time series prediction algorithms, specifically consisting of Holt-Winters, ARIMA and STL models.

Further, the training of the load trend prediction model is specifically:

Build a training set based on pre-acquired historical load indicator data;

Training several time series prediction models respectively based on the training set;

Based on the SMAPE algorithm, the training results of several time series forecasting models are evaluated, and the optimal model is selected as the final load trend forecasting model.

Specifically, for ease of understanding, the scheme described in this embodiment will be described in detail below in conjunction with the accompanying drawings:

Based on the problems existing in the prior art, this embodiment provides a method for dynamic horizontal expansion and contraction of microservice instances based on trend prediction, as shown in Figure 1, which is an overall overview of the microservice architecture based on automatic horizontal expansion/shrinkage based on trend prediction Figure, compared with ordinary microservice deployment, the difference is that a load trend prediction module is added, and the data based on this module comes from the indicator data collected by the indicator collection module and exists in the time series database. The indicator collection module obtains the access path of the specific microservice instance through the service discovery module. After the load trend prediction module predicts a specific trend, it will dynamically realize the horizontal expansion/contraction of the microservice instance through the instance controller. Based on this, it can be found that there are two key points in the solution described in this embodiment. One is to define how to collect load indicators that can affect instance expansion/shrinkage, and the other is to use an algorithm to train the best prediction model.

First, as shown in Figure 2, it is a flow chart of indicator collection, where Exporter is an indicator collection process that exists in the host machine and microservice instance, and these Exporter processes collect the load indicators of the host machine and microservice instance inwardly. Provide API interface exposure indicator data to the outside for collection by the indicator collection module, and the indicator collection module finally stores the data in the time series database. The Exporter process is an API service deployed together with related microservices, collects indicators internally, and provides query interfaces to indicator monitoring software such as Prometheus externally. This design idea is widely used in indicator monitoring software represented by Prometheus, and the corresponding Exporter process can be obtained through microservice providers or open source communities.

The definition of load indicators is mainly divided into two categories, one is the hardware indicators of the host machine and the load utilization rate of the container where the instance is located. The collected load indicators usually include CPU usage, memory usage, network IO, etc. These basic load indicators of different microservice instances should be defined as the same. The second is the application layer load indicators that reflect the service responsiveness, such as Query per Second (QPS), Response Time (RT), etc. The load indicators of these service layers vary with different service types.

With the specific indicators for trend prediction, the next step is to determine the specific algorithm for judging whether the microservice instance needs to be expanded/shrunk horizontally based on the load indicators. As shown in Figure 3, it is a graph of the training process of the trend prediction model. The historical data in it are the load index data we collected in the previous step. Most of these load indicators are stable time series data or can be converted into stable time For sequence data, we can use some specific algorithms to train an optimal model based on time series data, and then we can make effective predictions.

As shown in the trend prediction model training process diagram, we mainly combine three time series prediction algorithms, namely Holt-Winters, ARIMA (AutoRegressive Integrated Moving Average) and STL (Seasonal-Trend decomposition using LOESS); specifically:

Holt-Winters is a time series analysis and forecasting method. This method can be used to predict non-stationary sequences with linear trends and periodic fluctuations. The exponential smoothing method (EMA) is used to make the model parameters continuously adapt to the changes of non-stationary sequences, and a prediction function for trend and periodicity is fitted. forecast future trends.

The prediction principle of the ARIMA model is to analyze the input time series itself, extract all the correlation information in the sequence to fit a function, and use this function to make predictions. The ARIMA(p, d, q) model is an extension of the ARMA(p, q) model, and ARMA is a combination of the AR model and the MA model. The full name of the AR(p) model is the p-order regression model; the MA(q ) model, the full name is the q-order moving average model; the ARMA model, the full name is the autoregressive moving average model. On the basis of ARMA, it is necessary to make further differences to make the forecast a stationary sequence, and finally get ARIMA.

STL (Seasonal and Trend decomposition using Loess) is a very general and robust method for decomposing time series. STL can perform local polynomial regression on both seasonal and trend factors. Among them, LOESS (locally weighted scatterplot smoothing, LOWESS or LOESS) is local polynomial regression fitting, which is a common method for smoothing two-dimensional scatterplots. It combines the simplicity of traditional linear regression and the flexibility of nonlinear regression. When estimating the value of a response variable, first take a subset of data from the vicinity of its predictor variable, and then perform linear regression or quadratic regression on the subset, and use weighted least squares method for regression, that is, the value closer to the estimated point The greater its weight, the resulting local regression model is finally used to estimate the value of the response variable. In this way, the point-by-point operation is performed to obtain the entire fitting curve.

There is no "silver bullet" in algorithm selection for trend prediction in the solution described in this example. The three selected algorithms complement each other in terms of usage scenarios and advantages and disadvantages, and the predicted data fitted by each will further improve Carry out relevant checks to ensure that the predictions are as accurate as possible.

The overall training process is to set some timing tasks and train the three algorithms of Holt-Winters, ARIMA and STL at the same time, using 2/3 of the historical data as training data and 1/3 of the data as testing data. For each algorithm, some specific parameters are looped in, the model is trained for each parameter of each algorithm, and then SMAPE (Symmetric Mean Absolute Percentage Error) is used as the evaluation of the model, so that we finally get an optimal model. Trend predictions can then be made based on this model.

Among them, SMAPE is an accuracy measurement method based on percentage (or relative) error, and we use it as a prediction evaluation indicator. In layman's terms, the accuracy measurement is to judge the quality of the model based on the deviation between the predicted value and the accurate value. The most reliable data for evaluating the prediction model is naturally the real time series data that has been obtained. We can divide the historical time series data into two parts, one part is used to train the model, and the other part is used to test the model. If we have 4 weeks (28 days) of data, we use the data of the first 3 weeks (the first 21 days) to adjust the algorithm training model, and then use the model to gradually predict the data of the last 7 days. It is best to use the rolling forecast method. Each forecast advances one day, and a total of 7 simulation tests can be carried out. Through this test method, an optimal forecast model is finally obtained.

Furthermore, if the amount of load index data we collect is large enough, we can also choose to train neural network models such as RNN and LSTM for prediction.

So far, we can predict the load trend of the corresponding service based on the historical load index data, and make an intuitive decision to expand or shrink according to the specific trend, and these decisions are finally realized through the instance controller module .

The above is the whole process of automatic horizontal expansion/shrinkage of microservice instances based on trend prediction. There are two key technical points: First, we need to define index data that can truly reflect the real load level of microservices, including the host And at the level of microservice instances, as well as at the level of physical resources and application layer indicators, and then we collect these time series data into the database through the indicator collection module, which will be used in the next prediction. The second is to select a suitable algorithm to train a high-accuracy prediction model, and then use the index data in the previous step to predict the load trend, and realize the expansion/contraction of the instance in advance according to the future prediction.

Embodiment two:

The purpose of this embodiment is to provide a system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction.

According to the second aspect of the embodiments of the present disclosure, a system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction is provided, including:

A data acquisition unit, which is used to periodically acquire the load indicators of the host computer and the microservice instance in the cloud platform based on the Exporter process; wherein, the Exporter process is an indicator collection process existing in the host computer and the microservice instance, and provides external Indicator data interface;

A data storage unit, which is used to store the obtained load index in a time series database according to time series;

A load trend prediction unit, which is used to obtain the load trend of the corresponding service based on the load index in the time series database by using a pre-built load trend prediction model;

The microservice instance expansion or contraction control unit is used for realizing the expansion or contraction of the microservice instance based on the obtained load trend.

Further, the system described in this embodiment corresponds to the method described in Embodiment 1, and its technical details have been described in detail in Embodiment 1, so details will not be repeated here.

In further embodiments, there is also provided:

An electronic device includes a memory, a processor, and computer instructions stored in the memory and run on the processor. When the computer instructions are run by the processor, the method described in Embodiment 1 is completed. For the sake of brevity, details are not repeated here.

It should be understood that in this embodiment, the processor can be a central processing unit CPU, and the processor can also be other general-purpose processors, digital signal processors DSP, application specific integrated circuits ASIC, off-the-shelf programmable gate array FPGA or other programmable logic devices , discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.

The memory may include read-only memory and random access memory, and provide instructions and data to the processor, and a part of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

A computer-readable storage medium is used for storing computer instructions, and when the computer instructions are executed by a processor, the method described in the first embodiment is completed.

The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The software module may be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware. To avoid repetition, no detailed description is given here.

Those skilled in the art can appreciate that the units of the examples described in this embodiment, that is, the algorithm steps, can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be regarded as exceeding the scope of the present application.

The method and system for dynamic horizontal expansion and contraction of microservice instances based on trend prediction provided by the above embodiments can be realized and have broad application prospects.

The above descriptions are only preferred embodiments of the present disclosure, and are not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure shall be included within the protection scope of the present disclosure.

Claims (10)

1. A dynamic horizontal expansion and contraction method of a micro-service instance based on trend prediction is characterized by comprising the following steps:

periodically acquiring load indexes of a host machine and a micro-service instance in the cloud platform based on an Exporter process; the Exporter process is an index collection process existing in a host and a micro service instance, and provides an index data interface to the outside;

storing the obtained load indexes in a time sequence database according to a time sequence;

based on the load indexes in the time sequence database, acquiring the load trend of the corresponding service by utilizing a pre-constructed load trend prediction model;

based on the obtained load trend, expansion or contraction of the micro-service instance is achieved.

2. The method for micro-service instance dynamic horizontal expansion and contraction based on trend prediction as claimed in claim 1, wherein the load indexes include hardware indexes of a host, load usage rate of a container where the micro-service instance is located, and application layer load indexes embodying corresponding capability of a service.

3. The method according to claim 2, wherein the hardware index of the host and the load utilization of the container in which the micro service instance is located specifically include CPU utilization, memory utilization, and network IO; the application layer load indexes embodying the corresponding capabilities of the service comprise the number of queries per second and the response time.

4. The method as claimed in claim 1, wherein the Exporter process is an API service deployed with related micro-services, and internally collects load metrics and externally provides a query interface for the metrics monitoring component.

5. The method as claimed in claim 1, wherein the load trend prediction model is a hybrid model composed of a plurality of time series prediction algorithms, and is specifically composed of Holt-Winters, ARIMA and STL models.

6. The method according to claim 1, wherein the training of the load trend prediction model specifically comprises:

constructing a training set based on pre-obtained historical load index data;

respectively training a plurality of time sequence prediction models based on the training set;

and evaluating the training results of the plurality of time series prediction models based on a SMAPE algorithm, and selecting the optimal model as a final load trend prediction model.

7. A trend prediction based micro-service instance dynamic horizontal expansion contraction system, comprising:

the data acquisition unit is used for periodically acquiring load indexes of hosts and micro-service instances in the cloud platform based on an Exporter process; the Exporter process is an index collection process existing in a host and a micro service instance, and provides an index data interface to the outside;

the data storage unit is used for storing the obtained load indexes in a time sequence database according to time sequence;

the load trend prediction unit is used for obtaining the load trend of the corresponding service by utilizing a pre-constructed load trend prediction model based on the load indexes in the time sequence database;

and the micro service instance expansion or contraction control unit is used for realizing expansion or contraction of the micro service instance based on the obtained load trend.

8. The trend prediction-based micro-service instance dynamic horizontal expansion contraction system as claimed in claim 7, wherein the load trend prediction model adopts a hybrid model composed of a plurality of time series prediction algorithms, and specifically consists of Holt-Winters, ARIMA and STL models.

9. An electronic device comprising a memory, a processor and a computer program stored and executed on the memory, wherein the processor implements a trend prediction based micro-service instance dynamic level expansion contraction method according to any one of claims 1 to 6 when executing the program.

10. A non-transitory computer-readable storage medium, on which a computer program is stored, wherein the program, when executed by a processor, implements a trend prediction based microservice instance dynamic level expansion contraction method as claimed in any one of claims 1 to 6.

Images