CN115617445A

CN115617445A - Container management method and device in Serverless computing

Info

Publication number: CN115617445A
Application number: CN202110790182.7A
Authority: CN
Inventors: 冯悦; 张伟; 王郁文; 朱宇昕; 田康
Original assignee: China Mobile Communications Group Co Ltd; China Mobile Xiongan ICT Co Ltd; China Mobile System Integration Co Ltd
Current assignee: China Mobile Communications Group Co Ltd; China Mobile Xiongan ICT Co Ltd; China Mobile System Integration Co Ltd
Priority date: 2021-07-13
Filing date: 2021-07-13
Publication date: 2023-01-17

Abstract

The invention provides a container management method and device in serverless computation. The method comprises the following steps: creating a pool of reserve containers, the pool of reserve containers comprising at least one target container; distributing a target container to the target function operation request under the condition that the target function operation request is received and the container corresponding to the target function does not exist; wherein the target container is adapted to run a plurality of functions including the target function. The container management method and device in server computing provided by the invention can effectively improve the service processing efficiency and obviously save system resources.

Description

Container management method and device in Serverless computing

Technical Field

The invention relates to the technical field of cloud computing, in particular to a container management method and device in server computing.

Background

In the existing Serverless (Serverless) computation, the scheduling mode of the service is as follows: when the function operation request arrives and a container corresponding to the function does not exist, a container related to the service is created to operate the function task. And after the function operation is finished, deleting the container if no function request corresponding to the container exists within a certain time, and releasing the resource.

Since the time range is difficult to judge, a new function request may be generated after the container is just deleted, and the container needs to be created again in such a situation, which causes resource waste. Meanwhile, due to the delay caused by creating the container, the service processing efficiency is low, and the user experience is poor.

Disclosure of Invention

The invention provides a container management method and device in server computing, which are used for overcoming the defects that unreasonable container creation causes resource waste and low service processing efficiency in the prior art.

In a first aspect, the present invention provides a container management method in server computing, including:

creating a reserve container pool, the reserve container pool comprising at least one target container;

under the condition that an objective function operation request is received and a container corresponding to an objective function does not exist, distributing the objective container to the objective function operation request;

wherein the target container is adapted to run a plurality of functions including the target function.

In one embodiment, after the creating the spare container pool, the method further comprises:

and under the condition that the utilization rate of the target container in the standby container pool is greater than or equal to a first threshold value, newly building a host and newly building a target container for the standby container pool.

releasing at least one host and at least one target container if the usage rate of the target container in the spare container pool is less than or equal to a second threshold;

wherein the first threshold is greater than the second threshold.

In one embodiment, in the case where a plurality of target containers are included in the spare container pool, the method further comprises:

and under the condition that the predicted operation time length of the next objective function is greater than or equal to the first operation time length threshold value, newly building at least one objective container.

In one embodiment, in the case that a plurality of target containers are included in the spare container pool, the method further comprises:

releasing at least one target container in the case that the expected operation time length of the next target function is less than or equal to a second operation time length threshold value;

wherein the first operational duration threshold is greater than the second operational duration threshold.

In one embodiment, the expected operating time of the next objective function is determined by:

inputting the operation time lengths of a first number of continuously operated objective functions into an operation time length prediction model to determine the expected operation time length of the next objective function.

In one embodiment, the run-time prediction model is a Prophet model.

In a second aspect, the present invention provides a container management apparatus in server less computing, including:

a container creation module to create a reserve container pool, the reserve container pool comprising at least one target container;

the container management module is used for distributing the target container to the target function operation request under the condition that the target function operation request is received and the container corresponding to the target function does not exist;

In a third aspect, the present invention provides an electronic device, including a processor and a memory storing a computer program, where the processor implements the steps of the container management method in serverless computing according to the first aspect when executing the program.

In a fourth aspect, the present invention provides a processor-readable storage medium storing a computer program for causing a processor to perform the steps of the method for container management in a serverless computation according to the first aspect.

The container management method and device in server computing provided by the invention effectively shorten the response time of the function operation request by establishing the standby container pool, and realize container multiplexing by distributing the target container capable of operating various functions from the standby container pool for the function operation request, thereby effectively improving the service processing efficiency and obviously saving system resources.

Drawings

In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.

FIG. 1 is a schematic flow chart of a container management method in serverless computing according to the present invention;

FIG. 2 is a schematic structural diagram of a container management device in a serverless computing system according to the present invention;

fig. 3 is a schematic structural diagram of an electronic device provided in the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

To facilitate understanding of the technical solutions provided by the present invention, the following will now be described:

the core idea of Serverless computing is to enable a user to concentrate on business logic and send business-unrelated things to cloud service for doing. The function, namely the service, is an important component of the server computing, the main content uses the function to realize the business logic, after the user uploads the function code, the function computing service can prepare the computing resource and run the code in an elastic and reliable mode, and the function processing of various scenes is supported.

The architecture of Serverless computing still requires a running server, but it is an event-driven architecture. In contrast, containerized applications remain legacy applications in nature, but are broken into many smaller parts or services. With containerized applications, it is never completely shut down. Even if no one accesses it, the container still needs to exist and operate. They can be scaled down to a single instance, but they are still running. A Serverless application may reduce its cost to zero if it does not request its functionality, and in fact they will stop running if it does not, which may significantly reduce cost and facilitate more rapid scaling.

At present, serverless computing is basically based on kubernets, and a kubernets scheduler is used as a brain of a cluster, so that how to improve the resource utilization rate of the cluster and ensure stable operation of services in the cluster becomes more and more important. Kubernets cluster scheduling employs a monolithic-based architecture, originally designed for long service, with less consideration in other scenarios. Although Kubernetes later adds support for short tasks and optimizes container scheduling performance, the problems that container expansion timeliness is poor, service operation needs to wait for container creation and starting to be completed, service online delay is large and the requirement of Serverless calculation cannot be met exist.

Fig. 1 is a schematic flow chart of a container management method in serverless computing according to the present invention. Referring to fig. 1, the present invention provides a container management method in server computing, which may include:

step 110, creating a standby container pool, wherein the standby container pool comprises at least one target container;

step 120, distributing a target container to the target function operation request under the condition that the target function operation request is received and the container corresponding to the target function does not exist;

It should be noted that the execution main body of the container management method in server computing provided by the present invention may be a server computing server. The serverless computing server may be implemented by a computer device, such as a mobile phone, a tablet computer, a laptop computer, a palmtop computer, an in-vehicle electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, or a Personal Digital Assistant (PDA), etc.

It should be noted that, in the process of creating the present invention, the inventor considers the following characteristics of the existing serverless calculation:

1. function first request delay is too long

When there is a function request and there is no container corresponding to the function, the most extreme case requires creating a host and creating a container to run the function. The host is created on the order of minutes and the container is created on the order of seconds, while the function may run on the order of milliseconds, causing the request to be delayed too long, resulting in a poor user experience.

2. Problem of reusability of container

In the cluster scheduling in the conventional manner, each time there is a function request and there is no container corresponding to the function, a container is created, and the created container is related to a service requested by the function. When the service requested by the second function is different from the service requested by the first function, the container related to the second function request needs to be created again, and the container requested by the first function may be deleted after being idle for a period of time, so that a new container needs to be created according to the type of the function under the condition that the empty container exists, the container cannot be reused, and resource waste is caused.

3. If a part of containers are reserved for functions all the time so as to directly run tasks when function requests arrive, the overhead caused by creating the containers is avoided, although the user experience can be improved to a certain extent, the problem of sharp rise of resource cost is brought at the same time, and the cost advantage caused by Serverless is weakened. If the function container which is not reused is released in a short time in order to reduce the resource cost, the problem of repeated creation and deletion of the container can be caused, the function execution delay is increased, and the user experience is influenced.

In view of the above-mentioned features of the serverless computing, the inventor considers that it is impossible for the serverless computing server to prepare one or more always running containers for each function, which greatly increases the cost of the server.

One solution is to prepare the computing resources when the request comes, and release the resources after the execution is finished. The problem with this approach is that it usually takes some time to prepare the context for function execution, so called cold start, which increases the response time of the call.

Similarly, if the resource is released immediately after execution, it is possible that the next request will come soon and prepare the environment again resulting in a cold start, lengthening the response time.

The inventor believes that a good solution should be possible to allow call requests to be processed in a timely manner while using less resources, i.e., balancing host resources and function execution delays and ensuring that the sum of the host resources and the function execution delays is minimal.

Therefore, the inventor introduces the concepts of 'host resource division' and 'function operation delay division':

the host resource is divided into = duration of the host startup state;

the function runtime delay point = host creation time + container creation time + function scheduling time + function runtime.

In order to achieve the minimum sum of the host resource and the function operation delay, the main method is to reduce five parameters, namely the host startup time, the host creation time, the container creation time, the function scheduling time and the function operation time.

First, to reduce host creation time and container creation time, the container and the runtime environment may be separated, such that the runtime environment is dynamically mounted.

Optionally, in step 110, the serverless computing server may first create a pool of standby containers that may contain at least one target container therein.

Wherein the target container is configured to be suitable for use in various operating environments. For example, the target container may be adapted to Java's Runtime Environment "Java Runtime Environment (JRE)", C # and C + +. Net's Runtime Environment "Net Common Language Runtime (CLR)", and the like. Thus, the target container may run various functions written in various languages such as java, C + +, and the like.

In step 120, when the server computing server receives the target function running request but there is no container corresponding to the target function currently, the server computing server may allocate a target container to the target function running request from the standby container pool to run the target function.

By creating the standby container pool in advance in step 110, and not needing to create a container after receiving the operation request of the target function, the function operation request response time can be effectively shortened.

In addition, in step 120, since the target container capable of running various functions is allocated to the target function when the running request of the target function is received (instead of allocating a container only for the target function to the target function), when the target function is run completely and there is no target function running request any more subsequently, the target container can also be reused for running other functions without being released, thereby realizing effective saving of resources.

In summary, the container management method in server computing provided by the present invention effectively shortens the response time of the function operation request by creating the standby container pool, and realizes container multiplexing by allocating the target container capable of operating various functions from the standby container pool for the function operation request, thereby effectively improving the service processing efficiency and significantly saving system resources.

In an embodiment, after creating the standby container pool, the method for managing containers in serverless computing provided by the present invention may further include:

and under the condition that the utilization rate of the target container in the standby container pool is greater than or equal to a first threshold value, newly building the host and newly building the target container for the standby container pool.

It should be noted that when too many hosts and containers remain operational, the host operational duration is greatly increased. In order to ensure that the host running time, the host creating time and the container creating time reach a balanced state, the problem that the time for preparing the container is long when the function is run for the first time needs to be solved.

Therefore, when the usage rate of the target container in the standby container pool exceeds a first threshold, for example, 80%, the serverless computing server can automatically create a new host and create a new target container for the standby container pool.

The size of the first threshold may be adjusted according to actual conditions, which is not specifically limited in the present invention.

releasing the at least one host and the at least one target container when the usage rate of the target container in the standby container pool is less than or equal to a second threshold;

wherein the first threshold is greater than the second threshold.

For example, the serverless computing server may automatically release a portion of the hosts and a portion of the target containers when the usage of the target containers of the pool of spare containers is less than or equal to a first threshold, such as 30%.

The released host and target container are the host and target container which are not used currently.

The size of the second threshold may be adjusted according to actual situations, which is not specifically limited in the present invention.

According to the container management method in server computing, the host and the container are newly built or released according to the utilization rate of the target container in the standby container pool, the dynamic expansion and contraction of the host running time according to the actual resource use condition can be realized, and the system resource is further saved under the condition that the function running response time is short.

In order to reduce the scheduling time of the target function and the running time of the target function, in an embodiment, in the case that the standby container pool includes a plurality of target containers, the method for managing containers in serverless computing provided by the present invention may further include:

and under the condition that the predicted operation time length of the next objective function is larger than or equal to the first operation time length threshold value, newly building at least one objective container.

In an embodiment, in a case that the standby container pool includes a plurality of target containers, the method for managing containers in serverless computing provided by the present invention may further include:

releasing at least one target container in the case that the expected operating time of the next objective function is less than or equal to a second operating time threshold;

Alternatively, in the case where a plurality of target containers are included in the spare container pool, a plurality of (e.g., 2, etc.) target containers may be first allocated to the target function in advance, and the number of the target containers may be scaled based on the allocation.

When the predicted running time of the next target function is predicted to be larger than or equal to the first running time threshold, the serverless computing server can add a new target container for the target function; and when the predicted operation time length of the next target function is predicted to be less than or equal to the second operation time length threshold, the serverless computing server can release the created target container for the target function.

Wherein, the first operation time length threshold value can be 1.1 times, 1.2 times, etc. of the average operation time length of the objective function; the second operating duration threshold may be 0.8 times, 0.9 times, etc. the average operating duration of the objective function.

According to the container management method in server computing, provided by the invention, the container resources are accurately adjusted according to the predicted result of the function running time, and the dynamic adjustment of the container quantity based on the actual resource requirement can be realized, so that the resource use and the user experience are effectively balanced.

In one embodiment, the expected operating time of the next objective function may be determined as follows:

the operating durations of a first number of continuously operated objective functions are input into an operating duration prediction model to determine the expected operating duration of the next objective function.

The first number may be preset, for example, 400, 500, and the like, and may also be determined according to the following manner:

calculating a statistical period of the objective function; the statistical period of the objective function is related to the request interval of the objective function (the larger the request interval is, the longer the statistical period is, and the smaller the request interval is, the shorter the statistical period is, and according to the empirical value, the statistical period may be set to be a second number times, for example, 5 times, the average request interval of the objective function);

the run length of the next objective function is predicted with the average run length of the objective function for a third number, e.g., 100, of consecutive statistical cycles.

The specific sizes of the first number, the second number and the third number may be adjusted according to actual situations, which is not specifically limited in the present invention.

According to the container management method in server computing, the operation duration of the next objective function is predicted by inputting the operation durations of a plurality of continuously operated objective functions into the operation duration prediction model, so that the characteristics of each objective function in the time domain can be fully considered, and the prediction result of the operation duration of the next objective function is very accurate.

In one embodiment, the run-time prediction model is a Prophet model.

The Prophet model is called "foreknowledge" model, which is a data prediction tool based on Python and R languages, and is particularly good at predicting various problems related to time series. The method has the advantages of flexible prediction, very high fitting speed and convenient parameter setting, and can efficiently and accurately predict the running time of the next target function.

The container management device in the serverless computation provided by the present invention is described below, and the container management device in the serverless computation described below and the container management method in the serverless computation described above may be referred to in correspondence.

Fig. 2 is a schematic structural diagram of a container management device in serverless computing according to the present invention. Referring to fig. 2, the present invention provides a container management apparatus in a serverless computing, which may include:

a container creation module 210 for creating a reserve container pool comprising at least one target container;

the container management module 220 is configured to, when an objective function operation request is received and a container corresponding to an objective function does not exist, allocate the objective container to the objective function operation request;

The container management device in server computing provided by the invention effectively shortens the response time of the function operation request by establishing the standby container pool, and realizes container multiplexing by distributing the target container capable of operating various functions from the standby container pool for the function operation request, thereby effectively improving the service processing efficiency and obviously saving system resources.

In one embodiment, after the creating the spare container pool, the container creation module 210 is further configured to:

wherein the first threshold is greater than the second threshold.

In one embodiment, where multiple target containers are included in the spare container pool, the container creation module 210 is further configured to:

In one embodiment, the run-time prediction model is a Prophet model.

It should be noted that the container management apparatus in server computing according to the embodiment of the present invention can implement all the method steps implemented by the container management method in server computing, and can achieve the same technical effects, and detailed descriptions of the same parts and beneficial effects as those of the method embodiment in this embodiment are omitted here.

Fig. 3 illustrates a physical structure diagram of an electronic device, which may include, as shown in fig. 3: a processor (processor) 310, a Communication Interface (Communication Interface) 320, a memory (memory) 330 and a Communication bus 340, wherein the processor 310, the Communication Interface 320 and the memory 330 complete the Communication with each other through the Communication bus 340. Processor 310 may invoke computer programs in memory 330 to perform the steps of the container management method in a serverless computing, including, for example:

wherein the object container is adapted to run a plurality of functions including the object function.

In addition, the logic instructions in the memory 330 may be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention or a part thereof which substantially contributes to the prior art may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, enable the computer to perform the steps of the container management method in a serverless computing provided by the above methods, for example, including:

On the other hand, embodiments of the present application further provide a processor-readable storage medium, where the processor-readable storage medium stores a computer program, where the computer program is configured to enable the processor to perform the steps of the container management method in server computing provided in the foregoing embodiments, for example, the steps include:

distributing a target container to the target function operation request under the condition that the target function operation request is received and the container corresponding to the target function does not exist;

The processor-readable storage medium can be any available medium or data storage device that can be accessed by a processor, including, but not limited to, magnetic memory (e.g., floppy disks, hard disks, magnetic tape, magneto-optical disks (MOs), etc.), optical memory (e.g., CDs, DVDs, BDs, HVDs, etc.), and semiconductor memory (e.g., ROMs, EPROMs, EEPROMs, non-volatile memory (NAND FLASH), solid State Disks (SSDs)), etc.

The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.

Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims

1. A method for managing containers in a serverless computing is characterized by comprising the following steps:

creating a pool of reserve containers, the pool of reserve containers comprising at least one target container;

2. The method for container management in serverless computing as claimed in claim 1, wherein after said creating a pool of standby containers, the method further comprises:

3. The method for container management in serverless computing as claimed in claim 2, wherein after the creating a pool of standby containers, the method further comprises:

wherein the first threshold is greater than the second threshold.

4. The method for managing containers in serverless computing as claimed in claim 1, wherein in case that a plurality of target containers are included in the standby container pool, the method further comprises:

5. The method for managing containers in server computing according to claim 4, wherein in case that a plurality of target containers are included in the standby container pool, the method further comprises:

6. The method for managing containers in a serverless computing system according to claim 4 or 5, wherein the expected operation time of the next objective function is determined by:

7. The method for managing the containers in the serverless computing system according to claim 6, wherein the run-time prediction model is a Prophet model.

8. A container management apparatus in server computing, comprising:

9. An electronic device comprising a processor and a memory storing a computer program, wherein the processor when executing the computer program performs the steps of the method of container management in serverless computing according to any one of claims 1 to 7.

10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of a method of container management in a serverless computing as claimed in any one of claims 1 to 7.