A service resource monitoring method and device, electronic equipment and storage medium

Abstract

The application provides a service resource monitoring method and device, electronic equipment and storage medium, the method comprises the following steps: monitoring data processing tasks in a graphics processor computing power container running environment; obtaining a first termination process signal sent by a container closing instruction; triggering a signal processing function; obtaining a processing time delay parameter and a configuration time parameter matched with the signal processing function through the signal processing function, obtaining a comparison result of the processing time delay parameter and the configuration time parameter; based on the comparison result of the processing time delay parameter and the configuration time parameter, monitoring the recovery process of the service resource corresponding to the graphics processor computing power container running environment. Thus, the recovery process of the service resource can be monitored, the service resource is recovered in time and comprehensively, the graceful exit of the container is ensured, the use efficiency of the service resource is improved, the service data processing speed is ensured, and the use experience of the user is improved.

Description

Technical Field

[0001] This invention relates to cloud network service resource monitoring and processing technology, and more particularly to service resource monitoring methods, devices, electronic equipment, and storage media. Background Technology

[0002] With the continuous development of computer technology, cloud servers (CVMs) can provide secure and reliable elastic computing services, and can also provide different instance types to meet specific user scenarios. These instance types consist of different combinations of CPU, memory, storage, and network. When a cloud server is running a neural network model or performing a neural network model training task, if the user needs to interrupt the training task or a system exception occurs during training, regardless of whether the container is performing data I / O read operations or other training calculations, the container needs to gracefully exit. It is crucial to avoid situations where business resources are occupied and not released, otherwise, the efficiency of business resource utilization will be affected. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a business resource monitoring method, apparatus, electronic device, and storage medium, which can monitor the recycling process of business resources corresponding to the graphics processor computing power container operating environment based on the comparison results of processing latency parameters and configuration time parameters, ensuring timely and comprehensive recycling of business resources and graceful exit of the container, thereby improving the utilization efficiency of business resources, ensuring business data processing speed, and improving the user experience.

[0004] The technical solution of this invention is implemented as follows:

[0005] This invention provides a method for monitoring business resources, the method comprising:

[0006] Monitor data processing tasks in the graphics processor computing power container's operating environment;

[0007] When the data processing task is detected to be interrupted, the first process termination signal sent by the container shutdown command is obtained.

[0008] In response to the first process termination signal, a signal handling function is triggered;

[0009] The processing delay parameters and configuration time parameters that match the signal processing function are obtained through the signal processing function.

[0010] The processing delay parameter and the configuration time parameter are compared to obtain the comparison result of the processing delay parameter and the configuration time parameter;

[0011] Based on the comparison results of the processing latency parameter and the configuration time parameter, the recycling process of the business resources corresponding to the graphics processor computing power container operating environment is monitored.

[0012] This invention also provides a business resource monitoring device, comprising:

[0013] The information transmission module is used to monitor data processing tasks in the graphics processor computing power container's operating environment;

[0014] The information processing module is used to acquire the first termination process signal sent by the container shutdown command when the interruption of the data processing task is detected.

[0015] The information processing module is used to trigger a signal processing function in response to the first process termination signal;

[0016] The information processing module is used to obtain processing delay parameters and configuration time parameters that match the signal processing function through the signal processing function;

[0017] The information processing module is used to compare the processing delay parameter and the configuration time parameter to obtain the comparison result of the processing delay parameter and the configuration time parameter;

[0018] The information processing module is used to monitor the recycling process of business resources corresponding to the graphics processor computing power container operating environment based on the comparison results of the processing delay parameter and the configuration time parameter.

[0019] In the above scheme,

[0020] The information processing module is used to execute a business resource reclamation process corresponding to the trigger signal processing function when the processing delay parameter is less than the configured time parameter.

[0021] The information processing module is used to reclaim the memory corresponding to the data processing task through the business resource reclamation process;

[0022] The information processing module is used to close the bidirectional communication connection corresponding to the data processing task through the business resource recycling process.

[0023] In the above scheme,

[0024] The information processing module is used to close the user space file system client corresponding to the data processing task and terminate the user space file system client through the business resource recycling process when the graphics processor computing power container is running in a cloud server network.

[0025] The information processing module is used to close the file descriptor corresponding to the data processing task through the business resource recycling process.

[0026] In the above scheme,

[0027] The information processing module is used to trigger a second termination process signal corresponding to the graphics processor computing power container when the processing delay parameter is greater than or equal to the configuration time parameter.

[0028] The information processing module is used to shut down the graphics processor computing power container by means of a shutdown container instruction corresponding to the second termination process signal.

[0029] In the above scheme,

[0030] The information processing module is used to determine the environment type of the graphics processor computing power container's operating environment;

[0031] The information processing module is used to adjust the configuration time parameter based on the environment type of the graphics processor computing power container's operating environment, so as to match the configuration time parameter with the environment type of the graphics processor computing power container's operating environment.

[0032] In the above scheme,

[0033] The information processing module is used to determine the complexity of the neural network model running in the graphics processor computing power container;

[0034] The information processing module is used to adjust the configuration time parameter when the complexity of the neural network model is greater than the complexity threshold, so as to match the configuration time parameter with the complexity of the neural network model running in the graphics processor computing power container.

[0035] In the above scheme,

[0036] The information processing module is used to parse the data processing task and obtain the priority identifier of the data processing task.

[0037] The information processing module is used to sort the priorities of the received data processing tasks according to the priority identifier of the data processing task.

[0038] The information processing module is used to create a corresponding data processing task queue according to the priority of the data processing task.

[0039] The information processing module is used to adjust the processing order of the graphics processor computing power container according to the data processing task queue.

[0040] This invention also provides an electronic device, the electronic device comprising:

[0041] Memory, used to store executable instructions;

[0042] The processor, when running executable instructions stored in the memory, implements the preceding business resource monitoring method.

[0043] This invention also provides a computer-readable storage medium storing executable instructions, which, when executed by a processor, implement a preceding business resource monitoring method.

[0044] This application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the business resource monitoring method provided in this application.

[0045] The embodiments of the present invention have the following beneficial effects:

[0046] This invention monitors data processing tasks within the GPU computing power container's runtime environment. When an interruption of the data processing task is detected, a first termination signal sent by a container shutdown command is acquired. In response to the first termination signal, a signal processing function is triggered. Through the signal processing function, processing latency parameters and configuration time parameters matching the function are acquired. The processing latency parameters and configuration time parameters are compared to obtain a comparison result. Based on the comparison result, the recycling process of business resources corresponding to the GPU computing power container's runtime environment is monitored. This allows for monitoring of the recycling process of business resources corresponding to the GPU computing power container's runtime environment based on the comparison result of the processing latency parameters and configuration time parameters, ensuring timely and comprehensive recycling of business resources and a graceful exit of the container. This improves the utilization efficiency of business resources, guarantees business data processing speed, and enhances the user experience. Attached Figure Description

[0047] Figure 1 This is a schematic diagram illustrating a usage scenario of the business resource monitoring method provided in this embodiment of the invention;

[0048] Figure 2 A schematic diagram of the composition structure of an electronic device provided in an embodiment of the present invention;

[0049] Figure 3This is an optional flowchart illustrating the business resource monitoring method provided in this embodiment of the invention.

[0050] Figure 4 This is an optional flowchart illustrating the business resource monitoring method provided in this embodiment of the invention.

[0051] Figure 5 This is a schematic diagram illustrating the graceful exit process of the graphics processor computing power container in an embodiment of the present invention;

[0052] Figure 6 This is an optional flowchart illustrating the business resource monitoring method provided in this embodiment of the invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0054] In the implementation of this application, the collection and processing of relevant data should strictly comply with the requirements of relevant laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0055] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0056] Before providing a further detailed description of the embodiments of the present invention, the nouns and terms involved in the embodiments of the present invention will be explained, and the nouns and terms involved in the embodiments of the present invention shall be interpreted as follows.

[0057] 1) In response to, used to indicate the conditions or states on which the operation performed depends. When the conditions or states on which it depends are met, one or more operations performed may be performed in real time or with a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations are performed.

[0058] 2) Terminals, including but not limited to: ordinary terminals and dedicated terminals, wherein the ordinary terminals maintain a long connection and / or a short connection with the transmission channel, and the dedicated terminals maintain a long connection with the transmission channel.

[0059] 3) Client: The carrier that implements specific functions in the terminal. For example, a mobile client (APP) is a carrier of specific functions in a mobile terminal, such as performing report creation or report display functions.

[0060] 4) A server cluster refers to a group of servers working together to provide the same service, appearing to the client as a single server. Server clusters can utilize multiple computers for parallel computing to achieve high computing speeds, and can also use multiple computers for backup, ensuring the entire system continues to operate normally even if one machine fails. The server cluster hard disk failure handling method provided in this application can be applied to cloud server and distributed server scenarios, enabling status detection and fault repair of server hard disks in different usage scenarios. Specifically, a cloud server (CVM Cloud Virtual Machine) is a simple, efficient, secure, reliable, and elastically scalable computing service. Its management is simpler and more efficient than traditional single physical servers. Users do not need to purchase hardware in advance to quickly create or release any number of cloud servers for their business processes and to store user data. In a distributed server environment, user data and programs may not reside on a single server but are distributed across multiple servers. Similarly, distributed server environments also require a large number of hard disks, and the server cluster hard disk failure handling method provided in this application is needed to achieve server hard disk status detection and fault repair.

[0061] 5) Graceful container exit: When a graphics processor computing power container is running an AI training task, if the container is performing data I / O operations or other training computation operations, and the user needs to interrupt the task for some reason or the task is interrupted due to a system abnormality, the process environment, network environment and file system environment inside the container need to exit completely, without leaving dirty data and without generating zombie or frozen processes, so as to avoid the exiting container occupying resources without releasing them. This complete destruction of the container environment is called graceful container exit.

[0062] Figure 1 This is a schematic diagram illustrating a use case of the business resource monitoring method provided in this embodiment of the invention. (See attached diagram.) Figure 1With the continuous development of computer technology, cloud servers (Cloud Virtual Machines, CVMs) can provide secure and reliable elastic computing services, and can also provide different instance types to meet specific user scenarios. Terminals (including terminals 10-1 and 10-2) are equipped with corresponding clients capable of performing different functions. These clients (including terminals 10-1 and 10-2) obtain different information from the corresponding cloud server 200 via network 300 and can deploy different services on the cloud server. The terminals connect to the cloud server 200 via network 300, which can be a wide area network (WAN), a local area network (LAN), or a combination of both, using a wireless link for data transmission. The instance types provided by the cloud server consist of different combinations of CPU, memory, storage, and network, and store user business data on the cloud server's hard drive. The graphics processing unit (GPU) computing power containers deployed in the cloud server can perform different data processing tasks, such as training neural network models in a game environment to determine different parameters in the neural network model. The cloud server 200 can deploy a single-card instance of a JPEG HW decoder, and each single-card instance also includes five video decoders (NVDECs) and one Optical Flow Accelerator (OFA). These resources allow users to offload time-consuming computational loads that would otherwise require preprocessing on the CPU to the GPU, thereby accelerating the training speed of neural networks. In the embodiments provided in this invention, the cloud server application running on the cloud server 200 can be written in software code environments of different programming languages, and the code objects can be different types of code entities. For example, in C language software code, a code object can be a function. In JAVA language software code, a code object can be a class; in iOS Objective-C, it can be a piece of object code. In C++ language software code, a code object can be a class or a function to execute processing instructions from different terminals. This application does not distinguish the source of the cloud server's compilation environment.

[0063] The structure of the business resource monitoring device according to an embodiment of the present invention will be described in detail below. The business resource monitoring device can be implemented in various forms, such as a dedicated terminal with business resource monitoring device processing function, or a server equipped with business resource monitoring device processing function, such as the preceding one. Figure 1 Server 200 in the middle. Figure 2 This is a schematic diagram of the composition structure of the business resource monitoring device provided in an embodiment of the present invention. It can be understood that... Figure 2 This is only an exemplary structure of the business resource monitoring device, not the entire structure; implementation can be carried out as needed. Figure 2 The structure shown may be part or all of the structure.

[0064] The electronic device provided in this embodiment of the invention includes: at least one processor 201, a memory 202, a user interface 203, and at least one network interface 204. The various components in the service resource monitoring device are coupled together via a bus system 205. It can be understood that the bus system 205 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 205 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 2 The general labeled all buses as Bus System 205.

[0065] The user interface 203 may include a monitor, keyboard, mouse, trackball, click wheel, buttons, touchpad, or touch screen.

[0066] It is understood that memory 202 can be volatile memory or non-volatile memory, or both. In this embodiment of the invention, memory 202 is capable of storing data to support the operation of a terminal (such as 10-1). Examples of this data include any computer programs used to operate on the terminal (such as 10-1), such as operating systems and applications. The operating system includes various system programs, such as the framework layer, core library layer, driver layer, etc., used to implement various basic services and handle hardware-based tasks. Applications can include various applications.

[0067] In some embodiments, the service resource monitoring device provided in this invention can be implemented using a combination of hardware and software. For example, the service resource monitoring device provided in this invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the service resource monitoring method provided in this invention. For instance, the processor in the form of a hardware decoding processor can employ one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0068] As an example of the business resource monitoring device provided in this embodiment of the invention, which is implemented using a combination of software and hardware, the business resource monitoring device provided in this embodiment of the invention can be directly embodied as a combination of software modules executed by processor 201. The software modules can be located in a storage medium, which is located in memory 202. Processor 201 reads the executable instructions included in the software modules in memory 202 and combines them with necessary hardware (e.g., including processor 201 and other components connected to bus 205) to complete the business resource monitoring method provided in this embodiment of the invention.

[0069] As an example, processor 201 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., wherein the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0070] As an example of the hardware implementation of the business resource monitoring device provided in this embodiment of the invention, the device provided in this embodiment of the invention can be directly executed by a processor 201 in the form of a hardware decoding processor. For example, it can be executed by one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components to implement the business resource monitoring method provided in this embodiment of the invention.

[0071] In this embodiment of the invention, the memory 202 is used to store various types of data to support the operation of the business resource monitoring device. Examples of such data include: any executable instructions for operation on the business resource monitoring device, such as executable instructions, where a program implementing the business resource monitoring method of this embodiment of the invention may be included in the executable instructions.

[0072] In other embodiments, the service resource monitoring device provided in this invention can be implemented in software. Figure 2A service resource monitoring device stored in memory 202 is shown. This device can be software in the form of programs and plug-ins, and includes a series of modules. As an example of a program stored in memory 202, it may include the service resource monitoring device. The service resource monitoring device includes the following software modules: an information transmission module 2081 and an information processing module 2082. When the software modules in the service resource monitoring device are read into RAM and executed by processor 201, the service resource monitoring method provided in this embodiment of the invention will be implemented. The functions of each software module in the service resource monitoring device include:

[0073] The information transmission module 2081 is used to monitor data processing tasks in the graphics processor computing power container operating environment.

[0074] The information processing module 2082 is used to acquire the first process termination signal sent by the container shutdown instruction when the interruption of the data processing task is detected.

[0075] The information processing module 2082 is used to trigger a signal processing function in response to the first process termination signal.

[0076] The information processing module 2082 is used to obtain processing delay parameters and configuration time parameters that match the signal processing function through the signal processing function.

[0077] The information processing module 2082 is used to compare the processing delay parameter and the configuration time parameter to obtain the comparison result of the processing delay parameter and the configuration time parameter.

[0078] The information processing module 2082 is used to monitor the recycling process of business resources corresponding to the graphics processor computing power container operating environment based on the comparison results of the processing delay parameter and the configuration time parameter.

[0079] according to Figure 2 The electronic device shown, in one aspect of this application, also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in various optional implementations of the above-described business resource monitoring method.

[0080] Before introducing the business resource monitoring method to be provided in this application, we will first introduce the shortcomings of the exit process of the graphics processing unit (GPU) computing power container in traditional technology. For the abnormal exit of GPU computing power containers, the traditional solution relies on executing the docker stop command. However, from the perspective of platform usage, it does not standardize the capture of the stop signal and the execution of the corresponding operation function. It relies too much on the user program to perform resource release operations when the container exits. However, the quality of user code varies, and it is easy for the SIGTERM signal sent by the stop command to be ignored. The container may be forcibly released by the SIGKILL signal, which may result in zombie processes. As a result, the resources in the container are not released in time, and the container cannot complete the graceful exit, which affects the efficiency of the allocation and use of new GPU computing power containers.

[0081] To address this deficiency, refer to Figure 3 , Figure 3 This is an optional flowchart illustrating the business resource monitoring method provided in this embodiment of the invention. It can be understood that... Figure 3 The steps shown can be performed by various electronic devices running business resource monitoring equipment, such as dedicated terminals with business resource monitoring functions, servers or server cluster controllers, or control terminals of cloud network servers. The dedicated terminal with business resource monitoring equipment can be encapsulated within... Figure 1 In the server 200 shown, to execute the pre-processor Figure 2 The corresponding software modules in the business resource monitoring device shown. The following section addresses... Figure 3 The steps shown are explained.

[0082] Step 301: The business resource monitoring device monitors the data processing tasks in the graphics processor computing power container's operating environment.

[0083] In this invention, embodiments can be implemented using cloud technology. Cloud technology refers to a hosting technology that unifies hardware, software, and network resources within a wide area network (WAN) or local area network (LAN) to achieve data computation, storage, processing, and sharing. It can also be understood as a general term for network technologies, information technologies, integration technologies, management platform technologies, and application technologies based on cloud computing business models. The backend services of network systems require substantial computing and storage resources, such as video websites, image websites, and many portal websites; therefore, cloud technology needs cloud computing as its support.

[0084] It's important to note that cloud computing is a computing model that distributes computing tasks across a resource pool comprised of numerous computers, enabling various application systems to access computing power, storage space, and information services as needed. The network providing these resources is called the "cloud." From the user's perspective, resources in the "cloud" are infinitely scalable, readily available, and can be used on demand, expanded at any time, and paid for based on usage. As the foundational providers of cloud computing capabilities, they establish cloud resource pool platforms, often referred to as cloud platforms or Infrastructure as a Service (IaaS). These platforms deploy various types of virtual resources within the resource pool for external customers to choose from. The cloud resource pool primarily includes: computing devices (which can be virtualized machines containing operating systems), storage devices, and network devices.

[0085] Step 302: When the business resource monitoring device detects that the data processing task is interrupted, it obtains the first termination process signal sent by the container shutdown command.

[0086] The data processing tasks in the GPU computing power container runtime environment can be diverse, such as deep learning, image processing, video encoding and decoding, graphics databases, high-performance databases, computational fluid dynamics, computational finance, seismic analysis, molecular modeling, genomics, and other different types of tasks. Taking the neural network model training task in deep learning as an example, when the user's training task is terminated, the GPU computing power container will send the dockerstop command (i.e., the container shutdown command). The container shutdown command triggers the first termination process signal (i.e., the SIGTERM signal event). The GPU computing power container startup command needs to capture this signal event and cannot ignore or block the first termination process signal.

[0087] Step 303: The business resource monitoring device responds to the first termination process signal and triggers the signal processing function.

[0088] In some embodiments of the present invention, when the graphics processor computing power container captures the first termination process signal, it can call the SIG_Function function in the startup command to complete the signal processing, and reclaim and release the container's resources through the signal processing function.

[0089] Step 304: The business resource monitoring device obtains the processing delay parameters and configuration time parameters that match the signal processing function through the signal processing function.

[0090] Step 305: The business resource monitoring device compares the processing delay parameter and the configuration time parameter to obtain the comparison result of the processing delay parameter and the configuration time parameter.

[0091] In some embodiments of the present invention, the configuration time parameter can be flexibly adjusted according to different usage environments. Specifically, the environment type of the graphics processor computing power container's operating environment can be determined; based on the environment type of the graphics processor computing power container's operating environment, the configuration time parameter is adjusted to match the configuration time parameter with the environment type of the graphics processor computing power container's operating environment. For example, for training neural network models in a cloud network, the configuration time parameter is usually in the second range. If task data storage is performed via cloud storage mounted on a cloud network, the configuration time parameter can be increased to ensure complete exit of cloud storage data I / O and avoid creating zombie processes.

[0092] In some embodiments of the present invention, due to the complexity of the neural network model running in the graphics processing unit (GPU) computing power container, the training task is complex for neural network models with complex structures. Therefore, when determining the configuration time parameters, the complexity of the neural network model running in the GPU computing power container can be determined first. When the complexity of the neural network model is greater than a complexity threshold, the configuration time parameters are adjusted to match the complexity of the neural network model running in the GPU computing power container. For example, the complexity of a bidirectional attention neural network model for analyzing behavioral information in game scenes such as shooting games, parkour games, racing games, multiplayer online battle arena (MOBA), racing games (RCG), and sports games (SPG) is much greater than that of a graph neural network model for analyzing behavioral information in card games. When the complexity threshold is 1, the configuration time parameter for training the bidirectional attention neural network model in the GPU computing power container can be greater than the configuration time parameter for the graph neural network model to ensure the graceful exit of the GPU computing power container.

[0093] Step 306: Based on the comparison results of the processing latency parameter and the configuration time parameter, the business resource monitoring device monitors the recycling process of the business resources corresponding to the graphics processor computing power container operating environment.

[0094] refer to Figure 4 , Figure 4 This is an optional flowchart illustrating the business resource monitoring method provided in this embodiment of the invention. It can be understood that... Figure 4 The steps shown can be performed by various electronic devices running business resource monitoring equipment, such as dedicated terminals with business resource monitoring functions, servers or server cluster controllers, or control terminals of cloud network servers. The dedicated terminal with business resource monitoring equipment can be encapsulated within... Figure 1 In the server 200 shown, to execute the pre-processor Figure 2 The corresponding software modules in the business resource monitoring device shown. The following section addresses... Figure 4 The steps shown are explained.

[0095] Step 401: Obtain the comparison results of processing delay parameters and configuration time parameters.

[0096] Step 402: Determine whether the processing delay parameter is less than the configured time parameter. If yes, proceed to step 403; otherwise, proceed to step 405.

[0097] Step 403: When the processing delay parameter is less than the configured time parameter, execute the service resource reclamation process corresponding to the trigger signal processing function.

[0098] Step 404: Through the business resource reclamation process, reclaim the memory corresponding to the data processing task and close the bidirectional communication connection corresponding to the data processing task.

[0099] In some embodiments of the present invention, since the operating environments of the graphics processing unit (GPU) computing power containers are different, in order to ensure the graceful exit of the GPU computing power containers, the resource reclamation process needs to be adjusted according to the operating environment of the GPU computing power containers. Specifically, when the GPU computing power container is operating in a cloud server network, the user space file system client corresponding to the data processing task is closed through the business resource reclamation process, and the user space file system client is terminated; the file descriptor corresponding to the data processing task is closed through the business resource reclamation process.

[0100] refer to Figure 5 , Figure 5 This is a schematic diagram illustrating the graceful exit process of the graphics processor computing power container in an embodiment of the present invention. The signal handling function is the SIG_Function function, which is embedded in the startup command. When the stop signal of abnormal exit is captured, the SIG_Function function will be executed. The function mainly performs memory reclamation operations, deleting and releasing the allocated memory to avoid memory leaks; and actively initiating the closing operation of the socket connection to avoid the situation where the other end is stuck waiting for data.

[0101] In cloud server networks, it's necessary to close sockets and terminate the userspace filesystem client. The userspace filesystem client (fuse filesystem in userspace) is a Linux module used to mount certain network spaces, such as SSH, to the local filesystem. Related content can be found on SourceForge. This ensures the complete exit of cloud drive data I / O. If the fuse client process does not exit gracefully, data I / O may freeze, preventing the graphics processing unit (GPU) container from fully releasing resources and affecting the creation of new GPU containers. Simultaneously, actively closing file descriptors prevents resource leaks or unexpected anomalies. Even though the operating system closes file descriptors after container exit, this is a forced closure. Therefore, resource reclamation of GPU containers is centralized at the platform level, lowering the barrier to entry for user programmers and avoiding the need for constant consideration of resource reclamation and complete release during program development.

[0102] Step 405: When the processing delay parameter is greater than or equal to the configured time parameter, trigger the second termination process signal corresponding to the graphics processor computing power container.

[0103] Among them, the second process termination signal corresponding to the graphics processor computing power container can be the SIGKILL signal, which is a signal that cannot be ignored or blocked. It should be noted that the SIGTERM process termination signal is different from SIGKILL in that it can be blocked and handled; it is usually used to request the program to exit normally by itself. The shell command kill generates this signal by default.

[0104] Step 406: Close the graphics processor computing power container by using the container closure command corresponding to the second termination process signal.

[0105] When a graphics processing unit (GPU) computing power container receives a kill command signal, it can immediately perform a forced exit operation without waiting for the cleanup and resource release of the process environment within the GPU computing power container. Therefore, when processing the second process termination signal, a time parameter can be configured so that the kill signal will not be triggered within the configured time period, ensuring that the cleanup of the container environment and the release of resources are completed within this time period. Thus, by providing the unified and standardized SIGTERM signal processing method in this application, the situation of incomplete cleanup due to user programs is eliminated, improving the operability of the GPU computing power system and facilitating the management of the GPU computing power container's lifecycle.

[0106] In some embodiments of the present invention, the data processing tasks can be parsed to obtain their priority identifiers; the priorities of the received data processing tasks can be sorted according to their priority identifiers; and corresponding data processing task queues can be created based on their priorities. This enables differentiated processing for different customers. Simultaneously, the data processing task queues are traversed to determine the highest-priority data processing task; the link quality of each link in the network resources is determined; and the link with the highest link quality in the network resources is configured for the highest-priority data processing task, so that the data processing tasks in the data processing task queue can be transmitted and processed through the configured link.

[0107] The following uses the training of a neural network model in a game processing environment as an example to illustrate the business resource monitoring method involved in this invention, wherein, in conjunction with Figure 1 The diagram illustrates the usage environment of the business resource monitoring method according to an embodiment of the present invention. The terminals (including terminals 10-1 and 10-2) are equipped with corresponding clients capable of performing different functions. The clients are used by the terminals (including terminals 10-1 and 10-2) to obtain neural network models from the corresponding cloud server 200 via network 300 for training, in order to determine the parameters of the neural network models. For example, in the case of card and board games, game operators can proactively obtain historical data to monitor the use of game cheats, or automatically obtain historical data when trigger conditions are met. For instance, the neural network model can be used to analyze game logs, and when the information in the game logs meets the trigger conditions, the operation of obtaining historical data is triggered. The terminals connect to the server 200 via network 300, which can be a wide area network (WAN), a local area network (LAN), or a combination of both, using a wireless link for data transmission. (See reference...) Figure 6 , Figure 6 The following is an optional flowchart illustrating the business resource monitoring method provided in this embodiment of the invention. Figure 6 The steps shown are explained.

[0108] Step 601: Deploy graphics processor computing power containers through different service clusters of cloud computing servers.

[0109] Step 602: The command `docker stop` to shut down the container triggers the SIGTERM signal.

[0110] Step 603: The startup command of the graphics processor computing power container obtains the SIGTERM signal.

[0111] Step 604: Trigger the signal handling function and execute the SIG Function.

[0112] Step 605: Determine whether the delay parameter exceeds the configured time parameter. If yes, proceed to step 606; otherwise, proceed to step 604.

[0113] Step 606: `docker kill` triggers the SIGKILL signal.

[0114] Step 607: In response to the SIGKILL signal, the graphics processor computing power container is destroyed to reclaim business resources.

[0115] This effectively reduces the difficulty of using third-party components in graphics processing unit (GPU) computing power containers. In particular, through the distributed file storage of cloud server networks, when executing training tasks and reading and writing data within the container, there is no need to worry about abnormal task interruption of training, resulting in the container not being able to exit completely or resources not being released, thus ensuring the graceful exit of GPU computing power containers.

[0116] Beneficial technical effects:

[0117] This invention monitors data processing tasks within the GPU computing container's runtime environment. When an interruption is detected in the data processing task, a first termination signal from a container shutdown command is acquired. In response to the first termination signal, a signal processing function is triggered. Through the signal processing function, processing latency parameters and configuration time parameters matching the function are acquired. The processing latency parameters and configuration time parameters are compared to obtain a comparison result. Based on this comparison result, the recycling process of business resources corresponding to the GPU computing container's runtime environment is monitored. This allows for monitoring of the recycling process of business resources based on the comparison result of processing latency parameters and configuration time parameters, ensuring timely and comprehensive recycling of business resources and a graceful exit of the container. This improves the utilization efficiency of business resources, guarantees business data processing speed, and enhances the user experience.

[0118] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring business resources, characterized in that, The method includes: Monitor data processing tasks in the graphics processor computing power container's operating environment; When the data processing task is detected to be interrupted, the first process termination signal sent by the container shutdown command is obtained. In response to the first process termination signal, a signal handling function is triggered; The processing delay parameters and configuration time parameters that match the signal processing function are obtained through the signal processing function. The processing delay parameter and the configuration time parameter are compared to obtain the comparison result of the processing delay parameter and the configuration time parameter; Based on the comparison result, when the processing delay parameter is less than the configured time parameter, the business resource reclamation process corresponding to the signal processing function is executed to reclaim the memory corresponding to the data processing task and close the bidirectional communication connection corresponding to the data processing task. Based on the comparison result, when the processing delay parameter is greater than or equal to the configuration time parameter, a second termination process signal corresponding to the graphics processor computing power container is triggered, and the graphics processor computing power container is shut down by a container shutdown command corresponding to the second termination process signal.

2. The method according to claim 1, characterized in that, The method further includes: When the graphics processor computing power container is running in a cloud server network, the user space file system client corresponding to the data processing task is shut down through the business resource recycling process, and the user space file system client is terminated. The file descriptor corresponding to the data processing task is closed through the business resource reclamation process.

3. The method according to claim 1, characterized in that, The method further includes: Determine the environment type of the graphics processor computing power container's operating environment; Based on the environment type of the graphics processor computing power container's operating environment, the configuration time parameters are adjusted to match the configuration time parameters with the environment type of the graphics processor computing power container's operating environment.

4. The method according to claim 1, characterized in that, The method further includes: Determine the complexity of the neural network model running in the graphics processor's computing power container; When the complexity of the neural network model exceeds a complexity threshold, the configuration time parameter is adjusted to match the complexity of the neural network model running in the graphics processor's computing power container.

5. The method according to claim 1, characterized in that, The method further includes: The data processing task is parsed to obtain its priority identifier; The data processing tasks are sorted according to their priority identifiers. Create a corresponding data processing task queue based on the priority of the data processing task; The processing order of the graphics processor computing power container is adjusted according to the data processing task queue.

6. A business resource monitoring device, characterized in that, The device includes: The information transmission module is used to monitor data processing tasks in the graphics processor computing power container's operating environment; The information processing module is used to acquire the first termination process signal sent by the container shutdown command when the interruption of the data processing task is detected. The information processing module is used to trigger a signal processing function in response to the first process termination signal; The information processing module is used to obtain processing delay parameters and configuration time parameters that match the signal processing function through the signal processing function; The information processing module is used to compare the processing delay parameter and the configuration time parameter to obtain the comparison result of the processing delay parameter and the configuration time parameter; The information processing module is used to execute a business resource reclamation process corresponding to the signal processing function when the processing delay parameter is less than the configured time parameter, based on the comparison result, to reclaim the memory corresponding to the data processing task and close the bidirectional communication connection corresponding to the data processing task. The information processing module is configured to, based on the comparison result, trigger a second termination process signal corresponding to the graphics processor computing power container when the processing delay parameter is greater than or equal to the configuration time parameter, and close the graphics processor computing power container by means of a container shutdown command corresponding to the second termination process signal.

7. The apparatus according to claim 6, characterized in that, The information processing module is also used for: When the graphics processor computing power container is running in a cloud server network, the user space file system client corresponding to the data processing task is shut down through the business resource recycling process, and the user space file system client is terminated. The file descriptor corresponding to the data processing task is closed through the business resource reclamation process.

8. The apparatus according to claim 6, characterized in that, The information processing module is also used for: Determine the environment type of the graphics processor computing power container's operating environment; Based on the environment type of the graphics processor computing power container's operating environment, the configuration time parameters are adjusted to match the configuration time parameters with the environment type of the graphics processor computing power container's operating environment.

9. The apparatus according to claim 6, characterized in that, The information processing module is also used for: Determine the complexity of the neural network model running in the graphics processor's computing power container; When the complexity of the neural network model exceeds a complexity threshold, the configuration time parameter is adjusted to match the complexity of the neural network model running in the graphics processor's computing power container.

10. The apparatus according to claim 6, characterized in that, The information processing module is also used for: The data processing task is parsed to obtain its priority identifier; The data processing tasks are sorted according to their priority identifiers. Create a corresponding data processing task queue based on the priority of the data processing task; The processing order of the graphics processor computing power container is adjusted according to the data processing task queue.

11. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; When a processor is used to execute executable instructions stored in the memory, it implements the business resource monitoring method according to any one of claims 1 to 5.

12. A computer-readable storage medium storing executable instructions, characterized in that, When the executable instructions are executed by the processor, they implement the business resource monitoring method according to any one of claims 1 to 5.

13. A computer program product, characterized in that, The computer program product includes computer instructions stored in a computer-readable storage medium, which, when executed by a processor, implement the business resource monitoring method according to any one of claims 1 to 5.

Images