CN111694789A - Embedded reconfigurable heterogeneous determination method, system, storage medium and processor - Google Patents

Abstract

The invention belongs to the technical field of reconfigurable computing, and discloses an embedded reconfigurable heterogeneous measurement method, system, storage medium, and processor. The computing power of the boards is centralized to build a task-driven reconfigurable heterogeneous computing platform; the cluster tasks and heterogeneous computing resources are managed in a unified manner by means of dynamic cluster construction, and a reconfigurable virtual computing environment is built using virtualization technology. The invention constructs a heterogeneous computing platform with resource self-organization coordination and unified management capability and a reconfigurable computing environment. The Web visualization module in the user interface layer provides an interactive interface for the user, and the security mechanism in it divides the user level into multiple levels to provide guarantee for the access and authorization of various users; use the graphical interface to complete the management of tasks, resources, users, etc., Reduce user complexity and provide task load balancing capabilities.

Description

Embedded reconfigurable heterogeneous assay method, system, storage medium, processor

technical field

The invention belongs to the technical field of reconfigurable computing, and in particular relates to an embedded reconfigurable heterogeneous measurement method, system, storage medium and processor.

Background technique

At present, in order to meet the needs of diversified computing, the traditional CPU-based general computing system can no longer meet the needs of high computing power such as artificial intelligence. More and more scenarios are beginning to introduce hardware such as GPU and FPGA for acceleration and heterogeneous computing. came into being. Heterogeneous computing refers to a computing system composed of computing units using different types of instruction sets and architectures. Heterogeneous computing is a technology that balances performance, cost, and power consumption. On the basis of the parallelism of computing tasks, tasks are allocated to the computing resources that are most suitable for executing them, so as to minimize the total execution time of computing tasks and achieve performance and cost. optimization. Due to the low integration of special-purpose processor chips, long design cycles, weak computing power, and poor flexibility, they cannot meet the development needs of new technologies. Reconfigurable computing can upgrade or modify the computer hardware structure to better meet the needs of flexible and changeable tasks. Reconfigurable computing-related technologies are developing rapidly, showing the development trend and technical requirements of architecture integration, resource sharing, and collaborative processing. However, a single computing node has limited payload and limited computing power, and task processing is easy to reach its load performance limit, which affects tasks. Therefore, it is necessary to interconnect independent loads in an effective way to form a computing platform with larger scale and higher computing power, so as to better solve the bottleneck problem of load calculation.

Since the concept of reconfigurable computing is proposed at the hardware level to solve the defects of long development cycle and poor flexibility of application-specific integrated circuits, more research is still focused on reconfigurable logic devices and system hardware performance improvement, and less consideration is given to Platform-level architecture, system management software and scheduling strategies improve system performance, thus ignoring system-level management scheduling and software-level reconfigurable computing to improve system performance. For reconfigurable computing to take advantage of its efficiency, reconfigurable hardware is the foundation, platform architecture is the key, self-organization, coordination and dynamic self-evolution theory and technology of resources, communication and architecture are the core, and rapid information storage and retrieval is to improve platform efficiency. It is necessary to conduct in-depth research on the content involved. In addition, for the task-driven model, the process from task access to task execution involves issues such as automatic fault detection, automatic switching of computing nodes, and automatic task recovery. In order to ensure the continuous and efficient execution of tasks, high availability of the computing platform is necessary. Therefore, there is an urgent need for a task-driven, highly available reconfigurable heterogeneous computing platform in which heterogeneous resources can be self-organized, coordinated and managed in a unified manner.

Through the above analysis, the existing problems and defects in the prior art are:

(1) At present, a single computing node has limited payload and limited computing capacity, and task processing is easy to reach its load performance limit, which affects the task processing efficiency.

(2) At present, most of the research on reconfigurable computing is still focused on the improvement of system hardware performance, and less consideration is given to the improvement of system performance by platform-level architecture, system management software and scheduling strategies.

The difficulty of solving the above problems and defects is as follows:

The embedded computing platform constructed by embedded computing nodes is physically distributed and heterogeneous. The use of inter-node communication technology to achieve resource self-organization, coordination and unified management is the basis for cluster construction; computing platforms are no longer strongly dependent on For the main control node, how to ensure the high availability of the platform when the main control node fails is the key to platform construction; the task execution on a single computing node has the risk of a single point of failure, and how to ensure the high availability of task execution at the platform level is the focus of the research. .

The significance of solving the above problems and defects is:

To solve the above problems and defects, multiple embedded computing nodes can be cascaded together by means of communication technology to achieve self-organized coordination and unified management of heterogeneous resources; the cluster is no longer strongly dependent on the main control node, and at the same time, it ensures that the faulty nodes are connected to each other. Automatic migration of tasks, build a "centerless", high-performance, high-availability embedded computing platform, and provide a task-driven, reconfigurable computing environment.

SUMMARY OF THE INVENTION

Aiming at the problems existing in the prior art, the present invention provides an embedded reconfigurable heterogeneous determination method, system, storage medium and processor.

The present invention is implemented in this way, an embedded reconfigurable heterogeneous measurement method, wherein the embedded reconfigurable heterogeneous measurement method combines a plurality of distributed embedded computing boards carrying various heterogeneous computing resources The computing power of the cluster is centralized, and a task-driven reconfigurable heterogeneous computing platform is built; the task and heterogeneous computing resources of the cluster are managed in a unified manner by means of dynamic cluster construction, and a reconfigurable virtual computing environment is built by using virtualization technology.

Further, the embedded reconfigurable heterogeneous measurement method accepts the task input of the user, provides a web visual interactive page, divides the nodes into multiple independent clusters, builds an agent on the upper layer of the cluster, and provides load balancing;

In the proxy node, multiple nodes share a virtual IP based on the VRRP protocol. When the master of the node bound to the virtual IP fails, the virtual IP drifts to a slave node, and the slave node becomes a new master and continues to provide task connections. Access and load balancing functions.

Further, the embedded reconfigurable heterogeneous assay method constructs a cluster through a dynamic cluster building module, constructs a task pool and a resource pool of the cluster through a task pool building module and a resource pool building module, and matches tasks and resources through a virtual computing environment building module. , which generates a virtual computing environment for task execution.

Further, the dynamic cluster construction of the embedded reconfigurable heterogeneous measurement method includes heartbeat detection, database consistency and dynamic center election strategy;

Heartbeat detection In the master-slave service mode of peer-to-peer node status, the survival detection of each node is performed, new computing nodes are added to the cluster, and the faulty nodes are deleted; when the computing nodes are deleted, the system's heartbeat detection will automatically detect Delete the fault information of the development board; after the deletion is completed, synchronize the content of the system resources in the database through the database consistency policy;

The database consistency strategy synchronizes the data in the database, so that each computing node knows the task and resource configuration information of the cluster. When the master node fails, the dynamic center election strategy is used to elect a new master node to synchronize the data in the database;

The dynamic center election realizes the dynamic selection of the master control node in the cluster. On the basis of the database consistency module, the election strategy is used to dynamically select the master node to achieve decentralization; the elected master node is responsible for the distribution of tasks, and assigns specific tasks to specific tasks. and task-related configuration information, distributed to a computing node according to a certain distribution strategy; when a computing node fails, the master node is responsible for failover and re-delivering the tasks on the failed node.

Further, when the embedded reconfigurable heterogeneous determination method has a task X, the master node performs task allocation according to the resource state of the computing node, and sends the task and its configuration information to the computing node i that meets the requirements; if the resources are insufficient , the task is queued up; after the resource allocation is successful, the database of the master node is changed, and the operation is synchronized to other databases; when the computing node i fails, the heartbeat detection will detect the failure and place the node i in the database. After synchronizing the database through the database consistency policy, the master node obtains the information of the failure of node i, and determines that the task X being executed on node i has failed, and the master node re-accords to the information such as the resource status of the surviving nodes in the cluster. Task X is issued, and task X is issued to node j for re-execution.

Further, when the embedded computing board is started, the resource pool building module of the embedded reconfigurable heterogeneous measurement method completes the scanning of resource configuration, obtains the registered device information and performs health detection on the device, and uses the available resource information. It is stored in the resource list of the database, realizes resource discovery and availability detection, and builds the resource pool of the cluster with the help of database consistency strategy;

The construction of the virtual computing environment matches the tasks with the resources required by the tasks. According to the requirements of the tasks, the master node builds the relevant configuration information of the task execution environment, and sends it to the computing nodes along with the tasks. The computing nodes integrate the tasks and the configuration environment to build a virtual computing environment. . Package the application and operating environment as a docker image and upload it to the docker warehouse, and use the docker method to build a virtual computing environment, which starts quickly and belongs to the second level; the computing node pulls the image from the docker warehouse according to the configuration information issued by the master node , it is enough to build a virtual computing environment for task execution, so that the installation of the application and the configuration of the environment can be completed automatically.

Further, the hardware resource of the embedded reconfigurable heterogeneous measurement method adopts a bus-component architecture, and the controller board mounted with the heterogeneous computing resource performs networked interconnection communication through the standard interface defined by the message bus; The unified access of heterogeneous resources on the embedded computing board and the networked interconnection between boards realize the standard scalable high-speed system bus and unified component encapsulation and access of heterogeneous resources.

Another object of the present invention is to provide a program storage medium for receiving user input, and the stored computer program enables an electronic device to perform any one of the following steps: The computing power of the embedded computing boards of the resource is concentrated to build a task-driven reconfigurable heterogeneous computing platform; the tasks and heterogeneous computing resources of the cluster are managed in a unified manner by using the dynamic cluster construction method, and the reconfigurable computing resources are constructed by using the virtualization technology. virtual computing environment.

Another object of the present invention is to provide an embedded reconfigurable heterogeneous assay system for implementing the embedded reconfigurable heterogeneous assay method, the embedded reconfigurable heterogeneous assay system comprising:

The user interface layer is used to provide the task access method of Web visualization and provide the function of load balancing;

The system middleware layer is used for cluster construction and unified management of heterogeneous computing resources;

The hardware layer is used for unified access of heterogeneous resources on embedded computing boards and network interconnection between boards;

The user interface layer includes:

Web visualization module, which provides Web visualization interface to manage tasks, resources and users, and provides guarantee for access and authorization;

The task access module provides a unified access address, uses an agent to ensure high availability of unified access, and distributes tasks among clusters according to the load balancing strategy;

The system middleware layer includes a dynamic cluster building module, a task pool building module, a resource pool building module and a virtual computing environment building module;

Dynamic cluster building modules, including heartbeat detection, database consistency and dynamic center election strategy, are used to build and manage clusters to achieve unified scheduling of tasks and resources;

The task pool building module is used to realize the construction of the cluster task pool;

The resource pool building module is used to realize the discovery and health check of heterogeneous resources;

Another object of the present invention is to provide a processor equipped with the embedded reconfigurable heterogeneous assay system.

Combined with all the above technical solutions, the advantages and positive effects of the present invention are: the present invention provides a task-driven reconfigurable heterogeneous computing platform on an embedded platform, and realizes self-organization, coordination and unification of heterogeneous resources Manage and build a reconfigurable task virtual computing environment to ensure high availability of task access, task distribution, and task execution. The invention concentrates the computing capabilities of the distributed embedded nodes, and constructs a heterogeneous computing platform with resource self-organization coordination and unified management capabilities and a reconfigurable computing environment. The Web visualization module in the user interface layer provides an interactive interface for users, and the security mechanism in it divides the user levels into multiple levels to provide guarantees for the access and authorization of various users; the graphical interface is used to complete the management of tasks, resources, users, etc. Reduce user complexity and provide task load balancing capabilities.

The invention realizes continuous high-availability task access, builds an agent cluster at the upper layer of the cluster at the user interface layer, and realizes that nodes in the agent cluster share a virtual IP (VIP) based on the VRRP protocol. When the node bound to the VIP fails, The VIP automatically drifts to the new proxy node, thereby providing users with a unified and highly available access interface and ensuring the high availability of task access.

The invention realizes the continuous high-availability task distribution, adopts the non-center main control node dynamic election technology, and ensures the high availability of the platform in the task distribution. When the current master node in the cluster fails, the master node will be removed from the cluster through heartbeat detection and database consistency strategy. After synchronizing the database, a new master node will be elected through the dynamic center election strategy, and continue to perform resource allocation and task downlinking. In this way, the decentralization of the cluster is realized, that is, it is no longer necessary to rely on the static master control node, which avoids the collapse of the entire cluster caused by the failure of the master node, ensures the high availability of the master node, and ensures the high availability of task dispatching.

The present invention realizes continuous high-availability task execution. When a certain computing node fails, through heartbeat detection and database consistency strategy, the computing node will be removed from the cluster, and all tasks being executed on the node will fail and fail. Resource allocation and task distribution are carried out by the master node, which realizes task migration and ensures high availability of task execution.

The invention realizes a dynamically reconfigurable task execution environment. On the basis of virtualization technology, an independent virtual computing environment is provided for each task. The computing resources in the virtual computing environment change dynamically according to the task requirements and realize task customization. The task execution environment is a dynamically reconfigurable task execution environment.

The invention realizes the self-organization coordination and unified management of heterogeneous resources. When a new computing node joins the cluster or a node fails (leaves the cluster), the node's survival state is detected through heartbeat detection, and the resource pool The resource discovery and availability detection of building blocks can dynamically add and delete heterogeneous resources, and the database consistency policy synchronizes the status of cluster resources, which realizes the hot swap of computing nodes and the adaptive access and deletion of heterogeneous resources. Unified management is the realization of self-organized coordination and unified management of heterogeneous resources.

Description of drawings

In order to explain the technical solutions of the embodiments of the present application more clearly, the following will briefly introduce the drawings that need to be used in the embodiments of the present application. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

FIG. 1 is a flowchart of an embedded reconfigurable heterogeneous assay method provided by an embodiment of the present invention.

2 is a schematic structural diagram of an embedded reconfigurable heterogeneous assay system provided by an embodiment of the present invention;

In the figure: 1. User interface layer; 2. System middleware layer; 3. Hardware layer.

FIG. 3 is a schematic structural diagram of an embedded reconfigurable heterogeneous measurement system provided by an embodiment of the present invention.

FIG. 4 is a schematic structural diagram of a user interface layer proxy provided by an embodiment of the present invention.

FIG. 5 is a schematic diagram of a dynamic center implementation provided by an embodiment of the present invention.

FIG. 6 is a schematic diagram of task delivery and failover according to an embodiment of the present invention.

FIG. 7 is a schematic diagram of constructing a virtual computing environment provided by an embodiment of the present invention.

FIG. 8 is a schematic diagram of interconnection of heterogeneous computing resource boards according to an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

In view of the problems existing in the prior art, the present invention provides an embedded reconfigurable heterogeneous assay method, system, storage medium, and processor. The present invention is described in detail below with reference to the accompanying drawings.

As shown in Figure 1, the embedded reconfigurable heterogeneous assay method provided by the present invention comprises the following steps:

S101: Concentrate the computing capabilities of multiple distributed embedded computing boards that carry various heterogeneous computing resources (such as CPU, GPU, FPGA, DSP, etc.) to build a task-driven reconfigurable heterogeneous computing platform ;

S102: Use the dynamic cluster construction method to uniformly manage the tasks and heterogeneous computing resources of the cluster, and use the virtualization technology to build a reconfigurable virtual computing environment.

As shown in Figure 2, the embedded reconfigurable heterogeneous assay system provided by the present invention includes:

User interface layer 1, which is used to provide the task access method of Web visualization and provide the function of load balancing.

System middleware layer 2 is used for cluster construction and unified management of heterogeneous computing resources.

Hardware layer 3 is used for unified access of heterogeneous resources on embedded computing boards and network interconnection between boards.

In the present invention, the user interface layer 1 includes:

The Web visualization module provides users with a Web visualization interface to manage tasks, resources, users, etc., and provides guarantees for the access and authorization of various users.

The task access module provides users with a unified access address, uses an agent to ensure the high availability of unified access for users, and distributes tasks among clusters according to a certain load balancing strategy.

The web visualization module provides users with a visual interactive interface for management, and at the same time realizes multiple division of user identities based on the security authentication and authorization mechanism, and provides corresponding management services for different users.

The task access module is used to receive task requests from users. This module is a proxy cluster built by multiple computing nodes. Based on the Virtual Router Redundancy Protocol (VRRP) protocol, the nodes in the proxy cluster share a virtual IP (VIP). When When the node to which the VIP is bound fails, the VIP automatically drifts to a new proxy node, thereby providing users with a unified and highly available access interface and ensuring the high availability of task access. At the same time, the proxy cluster uses the load balancing strategy to allocate tasks to nodes in multiple clusters, thereby improving the concurrency and throughput of tasks and improving resource utilization.

In the present invention, the system middleware layer 2 includes a dynamic cluster building module, a task pool building module, a resource pool building module and a virtual computing environment building module.

Dynamic cluster building modules, including heartbeat detection, database consistency, and dynamic center election strategy, are used to build and manage clusters to achieve unified scheduling of tasks and resources.

Heartbeat detection implements the detection of the health status of nodes in the cluster. In the master-slave service mode with equal node status, the survival detection of each node is performed, new computing nodes are added to the cluster, and faulty nodes are deleted. When a computing node is deleted, the content of the system resources in the database is synchronized through the database consistency policy, so as to ensure that the task will not be sent to the failed computing node;

Database consistency realizes synchronization of data in the database, so that each computing node knows the tasks and resource configuration information of the cluster. When the master node fails, a new master node is elected by the dynamic center election strategy, and the data in the database is synchronized, thus solving the problem of Failover issues. Database consistency ensures that each node in the cluster can become the new master node when the master control node fails, ensuring the high availability of the system;

The dynamic center election realizes the dynamic selection of the master control node in the cluster. On the basis of database consistency, the election strategy is used to dynamically select the master node, which realizes the dynamic construction of the cluster and ensures the high availability of the cluster. The elected master node is responsible for the distribution of tasks, that is, specific tasks and task-related configuration information are distributed to a certain computing node according to a certain distribution strategy. When a computing node fails, the master node is responsible for failover and resends tasks on the failed node.

The task pool building module implements the construction of the cluster task pool.

Resource pool building modules, including resource discovery and availability detection, realize the discovery and health check of heterogeneous resources. Resource discovery, when the embedded computing board is started, completes the scan of resource configuration, obtains registered device information, and obtains heterogeneous computing resource information. Availability detection: perform health detection on the discovered heterogeneous resources, and store the available resource information in the resource list of the database.

Hardware layer 3 is used for unified access of heterogeneous resources on embedded computing boards and networked interconnection between boards. It realizes standard scalable high-speed system bus and unified component encapsulation and access of heterogeneous resources, making heterogeneous Resources can be managed and communicated according to a unified calling interface and protocol, providing hardware support for the system's high scalability and componentized services for heterogeneous resources.

The technical solutions of the present invention will be further described below with reference to the accompanying drawings.

The embedded reconfigurable heterogeneous assay system provided by the present invention includes:

User interface layer 1, as shown in Figure 4, the user interface layer is responsible for accepting the user's task input and providing a web visual interaction page. At the same time, the nodes are divided into multiple independent clusters, and an agent is built on the upper layer of the cluster to provide the function of load balancing. This ensures the relative independence of databases between clusters, similar to the role of cluster sharding. For a single cluster, the concurrency of tasks is premised on the resource allocation of the master node, while for multiple clusters, the concurrency of tasks is allocated to multiple master nodes to improve the concurrency and throughput of tasks. In addition, the utilization of resources is also improved.

In the proxy node, multiple nodes share a virtual IP (VIP) based on the VRRP protocol. When the master of the node bound to the VIP fails, the VIP will drift to a slave node, and the slave node will become a new master and continue to provide Task access and load balancing functions. Therefore, for users, there is only one unified access interface, which ensures the high availability of the proxy node and thus the high availability of task access.

The system middleware layer 2 builds a cluster through the dynamic cluster inter-tool module, builds the task pool and resource pool of the cluster through the task pool building module and the resource pool building module, matches tasks and resources through the virtual computing environment building module, and generates a virtual virtual machine for task execution. computing environment.

Dynamic cluster building blocks consist of heartbeat detection, database consistency, and dynamic center election policies.

The heartbeat detection is responsible for the dynamic addition and deletion of nodes in the cluster. The heartbeat detection performs survival detection on each node under the master-slave service mode with equal node status, adds a new computing node to the cluster, and deletes the faulty node. It is the basis for realizing resource self-organization, coordination and unified management. When the computing node is deleted, the heartbeat detection of the system will automatically detect the fault information of the deleted development board. After the deletion is completed, the content of the system resources in the database is synchronized through the database consistency policy, so as to ensure that the task will not be delivered to the failed computing node.

Database consistency realizes synchronization of data in the database, so that each computing node knows the tasks and resource configuration information of the cluster. When the master node fails, the following dynamic center election strategy is used to elect a new master node to synchronize the data in the database, thereby Resolved an issue with failover. Database consistency ensures that each node in the cluster can become a new master node when the master control node fails, ensuring high system availability.

The dynamic center election module realizes the dynamic selection of the main control node in the cluster. As shown in Figure 5, on the basis of database consistency, the election strategy is used to dynamically select the main node to achieve decentralization. At this time, if the main node fails and goes down, A new master node will be elected to continue to maintain the cluster service, thus solving the cluster's dependence on the static master node, ensuring the high availability of the cluster, and thus ensuring the high availability of task delivery. The elected master node is responsible for the distribution of tasks, that is, specific tasks and task-related configuration information are distributed to a certain computing node according to a certain distribution strategy. When a computing node fails, the master node is responsible for failover and resends tasks on the failed node.

As shown in FIG. 6 , when a task X comes, the master node allocates the task according to the resource status of the computing node, and sends the task and its configuration information to the computing node i that meets the requirements. If resources are insufficient, the task is queued for waiting. After the resource allocation is successful, change the database of the master node, and then synchronize the operation to other databases. As shown in Figure 7, when the computing node i fails, the heartbeat detection will detect the failure and clear the resource information of node i in the database. After synchronizing the database through the database consistency policy, the master node obtains the failure information of node i , then it is determined that the task X being executed on node i has failed, the master node re-issues task X according to information such as the resource status of the surviving nodes in the cluster, and task X is delivered to node j for re-execution.

The task pool building block builds the task pool for the cluster.

When the embedded computing board is started, the resource pool building module completes the scanning of resource configuration, obtains the registered device information, performs health detection on the device, and stores the available resource information in the resource list of the database to realize resource discovery and availability detection. , and build the resource pool of the cluster with the help of the database consistency policy.

The virtual computing environment building module matches the task with the resources required by the task. According to the requirements of the task, the master node builds a virtual computing environment for task execution, that is, the configuration information about the construction task execution environment, and sends it to the computing node along with the task. Integrate tasks and configuration environments to build virtual computing environments. On this task-driven computing platform, the application and operating environment are packaged as a docker image and uploaded to the docker warehouse, and a virtual computing environment is built using the docker method, which starts quickly and belongs to the second level. As shown in Figure 7, the computing node pulls the image from the docker warehouse according to the configuration information issued by the master node, and builds a virtual computing environment for task execution, so that the installation of the application and the configuration of the environment can be automatically completed.

In the construction of the virtual computing environment, the task execution environment is dynamically generated by the task and task configuration information, that is, the task execution environment is driven by the task and is dynamically reconfigurable.

The hardware layer 3 is composed as shown in Figure 8. The hardware resource module adopts the "bus-component" architecture design, and the controller boards mounted with heterogeneous computing resources are networked and interconnected through the standard interface defined by the message bus. communication. When a new computing board is connected, it is only necessary to configure the corresponding board IP in the system to realize the discovery and location of new resources. The same method can be used when the board is removed. The hardware layer is responsible for the unified access of heterogeneous resources on the embedded computing board and the networked interconnection between boards, realizing a standard scalable high-speed system bus and unified component encapsulation and access of heterogeneous resources, so that heterogeneous resources can be The unified calling interface and protocol are used for management and communication, providing hardware support for the system's high scalability and componentized services of heterogeneous resources.

The container-based task allocation and deployment method refines the granularity of the resources occupied by tasks. The granularity is no longer at the slice level, but the size of the resources required by the task, that is, multiple tasks occupy a part of the resources, which greatly improves the resources. utilization rate.

In order to test the difference between the computing performance and resource occupancy of the CPU when executing tasks in Docker and the host (Host), the Linpack test tool was used on the Xilinx UltraScale+MPSoC ZCU102 platform to test the CPU with the floating-point calculation peak as the evaluation index. computing performance, use ApacheBenchmark to stress test the Nginx service, and use Nmon to monitor resource occupancy.

(1) Linpack test

Assuming that the number of floating-point operations per clock cycle of the CPU is 1, the theoretical peak value of CPU floating-point operations is 1.2GHz*1*4 cores=4.8Gflops.

The Linkpack test results are shown in Table 1, where Max represents the actual CPU floating-point calculation peak obtained by the test, and the maximum efficiency is defined as the ratio of Max to the theoretical CPU floating-point calculation peak.

Table 1 Comparison of the peak maximum efficiency of CPU floating-point computing in Host and Docker

Max(Gflops) Maximum Efficiency (Max/Theoretical Peak) Host 4.065e-01 8.47% Docker (Alpine) 5.6567e-01 11.78%

As can be seen in Table 1, there is no obvious performance loss in docker compared to Host.

(2) Apache Benchmark test

The remote access stress test is performed on the Nginx service in the Host and Docker respectively, and the monitoring results are shown in Table 2. Among them, CPU(usr%+sys%) represents the occupancy ratio of user space and kernel space CPU, and CPU-WAvg represents the weighted average proportion of CPU occupancy.

Table 2 Comparison of CPU resource usage in Host and Docker

CPU(usr%+sys%) CPU-WAvg Host 25%+70% 17.6% Docker 20%+50% 11.0%

As can be seen in Table 2, compared with Host, the weighted average proportion of CPU in Docker is reduced by 6.6%, that is, the resource occupancy in Docker is less.

Combining the above two tests, it can be seen that compared with Host, there is no obvious decrease in computing performance in the Docker environment, but it effectively reduces the occupation of resources. In addition, the use of containerization technology improves the deployment efficiency of tasks and reduces the complexity of operation and maintenance.

It should be noted that the embodiments of the present invention may be implemented by hardware, software, or a combination of software and hardware. The hardware portion may be implemented using special purpose logic; the software portion may be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or specially designed hardware. Those of ordinary skill in the art will appreciate that the apparatus and methods described above may be implemented using computer-executable instructions and/or embodied in processor control code, for example on a carrier medium such as a disk, CD or DVD-ROM, such as a read-only memory Such code is provided on a programmable memory (firmware) or a data carrier such as an optical or electronic signal carrier. The device and its modules of the present invention can be implemented by hardware circuits such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., It can also be implemented by software executed by various types of processors, or by a combination of the above-mentioned hardware circuits and software, such as firmware.

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. Any person skilled in the art is within the technical scope disclosed by the present invention, and all within the spirit and principle of the present invention Any modifications, equivalent replacements and improvements made within the scope of the present invention should be included within the protection scope of the present invention.

Claims (10)

1. an embedded reconfigurable heterogeneous assay method, characterized in that, the embedded reconfigurable heterogeneous assay method uses multiple distributed, embedded computing boards that carry various heterogeneous computing resources. The computing power is centralized to build a task-driven reconfigurable heterogeneous computing platform; the cluster tasks and heterogeneous computing resources are managed in a unified manner by means of dynamic cluster construction, and a reconfigurable virtual computing environment is built using virtualization technology. 2. The embedded reconfigurable heterogeneous measurement method according to claim 1, wherein the embedded reconfigurable heterogeneous measurement method accepts a user's task input, provides a web visual interactive page, and divides the nodes into Multiple independent clusters, build an agent on the upper layer of the cluster to provide load balancing; In the proxy node, multiple nodes share a virtual IP based on the VRRP protocol. When the master of the node bound to the virtual IP fails, the virtual IP drifts to a slave node, and the slave node becomes a new master and continues to provide task connections. Access and load balancing functions. 3. The embedded reconfigurable heterogeneous assay method of claim 1, wherein the embedded reconfigurable heterogeneous assay method constructs a cluster through a dynamic cluster building module, and constructs a module through a task pool and a resource pool. The building module builds the task pool and resource pool of the cluster, matches tasks and resources through the virtual computing environment building module, and generates a virtual computing environment for task execution. 4. The embedded reconfigurable heterogeneous measurement method according to claim 3, wherein the dynamic cluster construction of the embedded reconfigurable heterogeneous measurement method comprises heartbeat detection, database consistency and dynamic center election strategy ; In the master-slave service mode with equal node status, heartbeat detection performs survival detection on each node, adds new computing nodes to the cluster, and deletes faulty nodes; when a computing node is deleted, the system's heartbeat detection will automatically detect the deletion. The fault information of the development board; after the deletion is completed, the content of the system resources in the database is synchronized through the database consistency policy; Database consistency realizes the synchronization of data in the database, so that each computing node knows the task and resource configuration information of the cluster. When the master node fails, a new master node is elected by the dynamic center election strategy to synchronize the data in the database; The dynamic center election realizes the dynamic selection of the master control node in the cluster. On the basis of the database consistency module, the election strategy is used to dynamically select the master node to achieve decentralization; the elected master node is responsible for the distribution of tasks, and assigns specific tasks to specific tasks. and task-related configuration information, distributed to a computing node according to a certain distribution strategy; when a computing node fails, the master node is responsible for failover and re-delivering the tasks on the failed node. 5 . The embedded reconfigurable heterogeneous measurement method according to claim 3 , wherein when a task X comes in the embedded reconfigurable heterogeneous measurement method, the master node performs the process according to the resource state of the computing node. 6 . Task allocation, send the task and its configuration information to the computing node i that meets the requirements; if the resources are insufficient, queue the task; after the resource allocation is successful, change the database of the master node, and then synchronize the operation to other databases ; When the computing node i fails, the heartbeat detection will detect the failure and clear the resource information of node i in the database. After synchronizing the database through the database consistency policy, the master node obtains the information of the failure of node i, and then determines that the node i is on the node i. The task X being executed has failed, and the master node re-issues task X according to information such as the resource status of the surviving nodes in the cluster, and task X is delivered to node j for re-execution. 6 . The embedded reconfigurable heterogeneous assay method according to claim 3 , wherein the resource pool of the embedded reconfigurable heterogeneous assay method is constructed when the embedded computing board is started, and the resource configuration is completed. 7 . scan, obtain the registered device information and perform health detection on the device, store the available resource information in the resource list of the database, realize resource discovery and availability detection, and build the resource pool of the cluster with the help of the database consistency policy; The construction of the virtual computing environment matches the tasks with the resources required by the tasks. According to the requirements of the tasks, the master node builds a virtual computing environment for task execution, builds the relevant configuration information of the task execution environment, and sends it to the computing nodes along with the tasks. Integrate with the configuration environment to build a virtual computing environment. Package the application and operating environment as a docker image and upload it to the docker warehouse, and use the docker method to build a virtual computing environment, which starts quickly and belongs to the second level; the computing node pulls the image from the docker warehouse according to the configuration information issued by the master node , to build a virtual computing environment for task execution, so as to automate the installation of applications and the configuration of the environment. 7. The embedded reconfigurable heterogeneous measurement method according to claim 1, wherein the hardware resource of the embedded reconfigurable heterogeneous measurement method adopts a bus-component architecture, The controller board that constitutes computing resources performs networked interconnection and communication through the standard interface defined by the message bus; the heterogeneous resources on the embedded computing board are uniformly accessed, and the boards are networked and interconnected to realize a standard scalable high-speed system bus. Unified component encapsulation and access with heterogeneous resources. 8. A program storage medium for receiving user input, the stored computer program enables electronic equipment to perform any one of the following steps: the embedded computing board of a plurality of distributed, carrying various heterogeneous computing resources The computing power of the card is centralized, and a task-driven reconfigurable heterogeneous computing platform is built; the task and heterogeneous computing resources of the cluster are managed in a unified manner by means of dynamic cluster construction, and a reconfigurable virtual computing environment is built by using virtualization technology. 9 . An embedded reconfigurable heterogeneous assay system for implementing the embedded reconfigurable heterogeneous assay method according to any one of claims 1 to 7 , wherein the embedded reconfigurable heterogeneous assay system include: The user interface layer is used to provide the task access method of Web visualization and provide the function of load balancing; The system middleware layer is used for cluster construction and unified management of heterogeneous computing resources; The hardware layer is used for unified access of heterogeneous resources on embedded computing boards and network interconnection between boards; The user interface layer includes: Web visualization module, which provides Web visualization interface to manage tasks, resources and users, and provides guarantee for access and authorization; The task access module provides a unified access address, uses an agent to ensure high availability of unified access, and distributes tasks among clusters according to the load balancing strategy; The system middleware layer includes a dynamic cluster building module, a task pool building module, a resource pool building module and a virtual computing environment building module; Dynamic cluster building modules, including heartbeat detection, database consistency and dynamic center election strategy, are used to build and manage clusters to achieve unified scheduling of tasks and resources; The task pool building module is used to realize the construction of the cluster task pool; Resource pool building block for discovering and health checking of heterogeneous resources. 10 . A processor, wherein the processor is equipped with the embedded reconfigurable heterogeneous measurement system according to claim 9 . 11 .

Images