CN114579311B - Method, device, equipment and storage medium for executing distributed computing task - Google Patents

Abstract

The disclosure provides a method, device, device, storage medium, and program product for executing distributed computing tasks, and relates to the technical field of artificial intelligence, and in particular to cloud computing, distributed computing, and other technical fields. The specific implementation scheme is: sending the computing data to be processed related to the distributed computing task to each first node in the first node set, wherein the hardware information of the first node matches the local hardware information; aggregating each first node The node obtains a first aggregation result for the first calculation result of the calculation data to be processed; shares the first aggregation result with each second node in the second node set to obtain a shared result, wherein the hardware information of the second node is the same as the local The hardware information does not match; and determining the shared result as new computing data to be processed, and returning the operation of sending the computing data to be processed to each node in the first node set until the execution of the distributed computing task is completed.

Description

Method, device, device and storage medium for performing distributed computing tasks

technical field

The present disclosure relates to the technical field of artificial intelligence, and in particular to technical fields such as cloud computing and distributed computing.

Background technique

Distributed computing is a computing method, which is relative to centralized computing. With the development of computing technology, some computing tasks require greater computing power to complete. If centralized computing is used, it will take a long time to complete. Distributed computing can decompose the computing task into many small parts and assign them to multiple nodes for processing. This can shorten calculation time and improve calculation efficiency.

Contents of the invention

The present disclosure provides a method, device, device, storage medium and program product for executing distributed computing tasks.

According to an aspect of the present disclosure, there is provided a method for executing a distributed computing task, including: sending to-be-processed computing data related to the distributed computing task to each first node in the first node set, wherein the The hardware information of the first node matches the local hardware information;

Aggregating the first calculation results of each first node for the calculation data to be processed to obtain a first aggregation result; sharing the first aggregation result with each second node in the second node set to obtain a shared result , wherein the hardware information of the second node does not match the local hardware information; and determine the shared result as new computing data to be processed, and return the sending the computing data to be processed to the first set of nodes The operation of each node until the distributed computing task is completed.

According to another aspect of the present disclosure, there is provided a method for executing a distributed computing task, including: receiving computing data to be processed from a master node; performing a computing operation according to the computing data to be processed to obtain a first computing result; and Send the first calculation result to the master node.

According to another aspect of the present disclosure, an apparatus for executing a distributed computing task is provided, including: a first sending module, configured to send pending computing data related to the distributed computing task to each node in the first set of nodes a first node, wherein the hardware information of the first node matches the local hardware information; the aggregation module is configured to aggregate the first calculation results of each first node for the calculation data to be processed to obtain the first Aggregation result; a sharing module, configured to share the first aggregation result with each second node in the second node set to obtain a shared result, wherein the hardware information of the second node does not match the local hardware information and a determining module, configured to determine the shared result as new computing data to be processed, and return the operation of sending the computing data to be processed to each node in the first node set until the execution of the distributed computing task is completed .

According to another aspect of the present disclosure, there is provided an apparatus for performing distributed computing tasks, including: a receiving module, configured to receive computing data to be processed from a master node; a computing module, configured to perform computation based on the computing data to be processed A calculation operation, to obtain a first calculation result; and a sending module, configured to send the first calculation result to the master node.

Another aspect of the present disclosure provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores information executable by the at least one processor. Instructions, the instructions are executed by the at least one processor, so that the at least one processor can execute the method shown in the embodiments of the present disclosure.

According to another aspect of the embodiments of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to make the computer execute the method shown in the embodiments of the present disclosure.

According to another aspect of the embodiments of the present disclosure, there is provided a computer program product, including computer programs/instructions, which is characterized in that, when the computer program/instructions are executed by a processor, the steps of the methods shown in the embodiments of the present disclosure are implemented.

It should be understood that what is described in this section is not intended to identify key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will be readily understood through the following description.

Description of drawings

The accompanying drawings are used to better understand the present solution, and do not constitute a limitation to the present disclosure. in:

1 is a schematic diagram of an exemplary distributed system of a method, an apparatus, an electronic device, and a storage medium for performing a distributed computing task according to an embodiment of the present disclosure;

Fig. 2 schematically shows a flow chart of a method for performing a distributed computing task according to an embodiment of the present disclosure;

FIG. 3 schematically shows a flowchart of a method for determining a first node set and a second node set according to an embodiment of the present disclosure;

FIG. 4 schematically shows a flowchart of a method for determining a master node according to an embodiment of the present disclosure;

Fig. 5 schematically shows a schematic diagram of a method for executing a distributed computing task according to another embodiment of the present disclosure;

Fig. 6 schematically shows a block diagram of an apparatus for performing distributed computing tasks according to an embodiment of the present disclosure;

FIG. 7 schematically shows a block diagram of an apparatus for performing distributed computing tasks according to another embodiment of the present disclosure; and

FIG. 8 schematically illustrates a block diagram of an example electronic device that may be used to implement embodiments of the present disclosure.

Detailed ways

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and they should be regarded as exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The application scenarios of the method and apparatus for performing distributed computing tasks provided by the present disclosure will be described below with reference to FIG. 1 .

FIG. 1 is a schematic diagram of an exemplary distributed system of a method, an apparatus, an electronic device, and a storage medium for performing distributed computing tasks according to an embodiment of the present disclosure. It should be noted that, what is shown in FIG. 1 is only an example of the system architecture to which the embodiments of the present disclosure can be applied, so as to help those skilled in the art understand the technical content of the present disclosure, but it does not mean that the embodiments of the present disclosure cannot be used in other device, system, environment or scenario.

As shown in FIG. 1 , the distributed system 100 includes multiple nodes, such as nodes 111 , 112 , 113 , 114 , 121 , 122 , 123 , 124 , 131 , 132 , 133 and 134 .

The multiple nodes may be based on different hardware architectures. In this embodiment, nodes based on the same hardware architecture may be called homogeneous nodes, and nodes based on different hardware architectures may be called heterogeneous nodes.

For example, nodes 111, 112, 113 and 114 can be based on GPU (Graphics Processing Unit, graphics processing unit), and nodes 121, 122, 123 and 124 can be based on NPU (Natural-network Processing Unit, embedded neural network processor), node 131, 132, 133, and 134 may be based on a CPU (Central Processing Unit, central processing unit). Based on this, the nodes 111, 112, 113 and 114 are mutually isomorphic nodes, the nodes 121, 122, 123 and 124 are mutually isomorphic nodes, and the nodes 131, 132, 133 and 134 are mutually isomorphic nodes. Nodes 111, 112, 113 and 114 and nodes 121, 122, 123 and 124 are mutually heterogeneous nodes, nodes 111, 112, 113 and 114 and nodes 131, 132, 133 and 134 are mutually heterogeneous nodes, Nodes 121, 122, 123, and 124 and nodes 131, 132, 133, and 134 are heterogeneous nodes.

According to the embodiments of the present disclosure, for example, homogeneous nodes can be formed into a homogeneous node set. Then determine the master node in each node set. Combine the master nodes in each node set into a heterogeneous set of nodes. For example, nodes 111 , 112 , 113 and 114 may be formed into a homogeneous node set 10 , and node 111 may be determined as the master node of the homogeneous node set 10 . Nodes 121 , 122 , 123 and 124 may be formed into a homogeneous node set 20 , and node 121 is determined as the master node of the homogeneous node set 20 . The nodes 131 , 132 , 133 and 134 may be formed into a homogeneous node set 30 , and node 131 is determined as the master node of the homogeneous node set 30 .

According to an embodiment of the present disclosure, when executing a distributed computing task, the master node in each homogeneous node set may send the data to be calculated to the nodes in the homogeneous node set to which it belongs. Then the master node can aggregate the calculation results of all nodes in the homogeneous node set to which it belongs, and obtain the aggregation result.

According to the embodiments of the present disclosure, the absolute computing power of heterogeneous nodes is different, and the speed at which the computing tasks are completed is also different. Therefore, after the calculation in the homogeneous node set is completed, it is necessary to perform calculations between the homogeneous node sets. data synchronization. Based on this, each master node can share the aggregation result with other master nodes in the heterogeneous node set. Each master node continues subsequent calculations based on the shared aggregation results. According to the embodiments of the present disclosure, data synchronization can be performed regularly among homogeneous node sets. Wherein, the data synchronization period may be determined according to actual needs. It can be understood that, in each data synchronization cycle, the number of calculations performed in each homogeneous node set may be different. The number of calculations in the homogeneous node set with higher absolute computing power is more, and the number of calculations in the homogeneous node set with smaller absolute computing power is more.

Exemplarily, the optimal or better communication method can be used inside the homogeneous node set. If the nodes in the homogeneous node set 10 are all based on GPU, then the nodes in the isomorphic node set 10 can use a GPU-specific collective communication library for communication. communication. The nodes in the homogeneous node set 20 are all based on the NPU, and the nodes in the homogeneous node set 20 can use the NPU-specific collective communication library for communication. In addition, a general communication method can be used for communication between heterogeneous nodes, for example, a general collective communication library suitable for GPU, CPU and NPU is used among the master nodes 111 , 121 and 131 .

According to the method for executing distributed computing tasks in the embodiments of the present disclosure, each node can be grouped according to hardware information, and isomorphic nodes can be divided into a set of isomorphic nodes, and the sets of isomorphic nodes are divided into sets of isomorphic nodes The master node communicates, which can realize distributed computing using heterogeneous nodes. In addition, a dedicated communication method can be used inside the homogeneous node set, which can make more efficient use of hardware resources and improve computing efficiency.

In the technical solutions of the present disclosure, the collection, storage, use, processing, transmission, provision, and disclosure of calculation data, hardware information and other data involved are all in compliance with relevant laws and regulations, and do not violate public order and good customs.

The method for executing distributed computing tasks provided by the present disclosure will be described below with reference to FIG. 2 . Wherein, the method for executing a distributed computing task may be executed, for example, by a master node.

Fig. 2 schematically shows a flowchart of a method for executing a distributed computing task according to an embodiment of the present disclosure.

As shown in FIG. 2 , the method 200 for executing a distributed computing task includes, at operation S210a, the master node sends the computing data to be processed related to the distributed computing task to each first node in the first node set.

According to an embodiment of the present disclosure, the local hardware information may be, for example, the hardware information of the master node. Wherein, the hardware information may include hardware type, for example. For example, the first node may be a node that matches the hardware information of the master node. The first set of nodes may include at least one first node, for example. Exemplarily, in this embodiment, if the hardware information of two nodes matches, it may indicate that the two nodes are isomorphic nodes.

According to an embodiment of the present disclosure, the distributed computing task may include, for example, a training task of a deep learning model, and the computing data to be processed may include, for example, model parameters.

Then, for each first node, the following operations S210b to S230b are performed.

Wherein, in operation S210b, the computing data to be processed is received from the master node.

In operation S220b, a calculation operation is performed according to the calculation data to be processed to obtain a first calculation result.

According to an embodiment of the present disclosure, the calculation operation may include gradient calculation and the like, for example.

In operation S230b, the first calculation result is sent to the master node.

Next, the master node continues to perform operations S220a-S240a.

In operation S220a, first calculation results of each first node for the calculation data to be processed are aggregated to obtain a first aggregation result.

According to an embodiment of the present disclosure, for example, the master node may perform weighted summation of the first calculation results according to the weight of each first node to obtain the first aggregation result. Wherein, the weight of each first node can be set according to actual needs.

According to another embodiment of the present disclosure, for example, the master node may also average all the first calculation results to obtain the first aggregation result.

In operation S230a, the first aggregation result is shared with each second node in the second node set, the shared result is obtained, and the shared result is determined as new calculation data to be processed.

According to an embodiment of the present disclosure, for example, the second node may be a node whose hardware information does not match the master node. The second set of nodes may, for example, include at least one second node. Exemplarily, in this embodiment, if the hardware information of the two nodes does not match, it may indicate that the two nodes are heterogeneous nodes.

According to an embodiment of the present disclosure, for example, an average value of the first aggregated result and all second aggregated results may be calculated as the shared result.

According to another embodiment of the present disclosure, for example, according to the weight of each master node, the first aggregation result and all the second aggregation results may be weighted and summed to obtain a shared result. Among them, the weight of each master node can be set according to actual needs.

In operation S240a, it is determined whether the execution of the distributed computing task is completed. If the distributed computing task has not been executed yet, return to operation S220a. If the distributed computing task has not been executed yet, it ends.

According to an embodiment of the present disclosure, for example, it may be determined whether the execution of the distributed computing task is completed according to whether the number of execution rounds reaches a predetermined number of times. If the execution rounds reach the predetermined number of times, it is determined that the distributed computing task has been executed. If the number of execution rounds does not reach the predetermined number, it is determined that the execution of the distributed computing task has not been completed.

According to the method for executing distributed computing tasks in the embodiments of the present disclosure, heterogeneous nodes can be used for distributed computing, hardware resources can be used more efficiently, and computing efficiency can be improved.

According to another embodiment of the present disclosure, the master node may also perform calculation operations on the calculation data to be processed to obtain a second calculation result. Based on this, after receiving the first calculation result from each first node, the second calculation result and the first calculation result of each first node may be weighted and summed to obtain the first aggregation result. Wherein, the weight of the master node and each first node can be set according to actual needs. According to another embodiment of the present disclosure, for example, the master node may also average the second calculation result and all first calculation results to obtain the first aggregation result.

The method for determining the first node set and the second node set provided by the present disclosure will be described below with reference to FIG. 3 . Wherein, the method for executing a distributed computing task may be executed, for example, by a master node.

Fig. 3 schematically shows a flowchart of a method for determining a first node set and a second node set according to an embodiment of the present disclosure.

As shown in FIG. 3 , the method for determining the first node set and the second node set may include acquiring local hardware information and hardware information of multiple nodes in operation S310.

According to an embodiment of the present disclosure, the hardware information may include, for example, a hardware type. The hardware type may include GPU, CPU, NPU, etc., for example.

In operation S320, a node whose hardware information matches the local hardware information among the plurality of nodes is determined as the first node to obtain a first node set.

According to an embodiment of the present disclosure, for example, a node having the same hardware type as the master node may be determined as the first node. All first nodes are grouped into a first node set.

In operation S330, among the nodes whose hardware information does not match the local hardware information, a master node is determined as a second node to obtain a second node set.

According to an embodiment of the present disclosure, for example, it is possible to determine a node whose hardware type is different from that of the master node. Then determine the master node among the nodes with different hardware types as the second node. All second nodes are grouped into a second node set.

The method for determining the master node provided by the present disclosure will be described below with reference to FIG. 4 . Wherein, the method for determining the master node may be performed, for example, by the first node.

Fig. 4 schematically shows a flowchart of a method for determining a master node according to an embodiment of the present disclosure.

As shown in FIG. 4, the method for determining a master node may include acquiring hardware information of multiple nodes in operation S410.

According to an embodiment of the present disclosure, the hardware information may include, for example, a hardware type and a node identifier.

In operation S420, a master node is determined among the plurality of nodes according to the hardware information in the plurality of nodes.

According to an embodiment of the present disclosure, a node of the same hardware type as the first node among the multiple nodes may be determined, that is, an isomorphic node. Then, according to the node identifiers of these isomorphic nodes, the master node among these isomorphic nodes is determined. Exemplarily, the node identifier may include a number, and the master node among the isomorphic nodes may be determined according to the size of the number. For example, the homogeneous node with the smallest number acts as the master node.

According to an embodiment of the present disclosure, after the first node set is determined, the master node may establish a first communication domain with each first node in the first node set. Based on this, the master node may, for example, broadcast the calculation data to be processed in the first communication domain, so that the calculation data to be processed can be sent to each first node in the first node set. Correspondingly, each first node may also establish a first communication domain with the master node. Then the first calculation result is sent to the master node through the first communication domain.

According to another embodiment of the present disclosure, nodes may also be grouped in advance. The nodes whose hardware information matches form a homogeneous node set. Then determine the master node in each set of homogeneous nodes. Combine the master nodes in each homogeneous node set into a heterogeneous node set. For each master node, other nodes in the homogeneous node set to which the master node belongs constitute the first node set, and other master nodes in the heterogeneous node set to which the master node belongs constitute the second node set.

Referring to FIG. 5 , the method for executing distributed computing tasks shown above will be further described in conjunction with specific embodiments. Those skilled in the art can understand that the following exemplary embodiments are only for understanding the present disclosure, and the present disclosure is not limited thereto.

Fig. 5 schematically shows a schematic diagram of a method for executing a distributed computing task according to another embodiment of the present disclosure.

As shown in FIG. 5 , the node identifiers may include a0, a1, a2, b0, b1, b2, b3, c0, c1, c2, c3, d0, d1, d2 and d3. The nodes can be started separately, and the hardware type, node identification, network address and other information of their own nodes can be obtained through the built-in program. A network address may include, for example, an IP address.

Each node sends the hardware type, node ID and network address to the registration center, so that the registration center can register the node. After successful registration, each node obtains global information through the registration center, and the global information includes hardware type node identifiers and network addresses of all nodes. Exemplarily, the registration center can be implemented by using a pre-declared master (master control) node or services such as etcd. Among them, etcd is a distributed key-value database.

According to the embodiments of the present disclosure, each node determines the isomorphic nodes that match its own hardware information and the heterogeneous nodes that do not match its own hardware information among other nodes according to its own hardware information and the hardware information of other nodes. The nodes form a set of isomorphic nodes. For example, nodes a0, a1, a2 and a3 form an isomorphic node set g1. Nodes b0, b1, b2 and b3 form an isomorphic node set g2. Nodes c0, c1, c2 and c3 form an isomorphic node set g3. Nodes d0, d1, d2 and d3 form an isomorphic node set g4.

According to the embodiment of the present disclosure, each node can sort all the nodes according to the node identifiers of all the nodes. Exemplarily, in this embodiment, the primary node can be determined for the node according to the number in the node identifier, for example, node 0 in each homogeneous set is the primary node, that is, a0, b0, c0, and d0 are respectively the isomorphic set g1 , the master node of g2, g3 and g4. a0, b0, c0 and d0 constitute a heterogeneous set g5.

Then, each node starts its own working process, and establishes a communication domain through the obtained network address. The master node establishes the homogeneous communication domain and the heterogeneous communication domain, and the non-master node establishes the homogeneous communication domain. Each master node broadcasts the first data in the corresponding homogeneous communication domain, so as to send the first data to each homogeneous node in the homogeneous communication domain.

Next, each node performs a predetermined calculation operation according to the first data to obtain a calculation result.

In response to the arrival of each synchronization period, the non-master node sends the calculation result obtained by itself to the master node of the homogeneous node set to which it belongs.

The master node receives the calculation results from each node in the homogeneous node set to which it belongs. Then aggregate the calculation results calculated by itself and the calculation results of other nodes in the homogeneous node set to which it belongs to obtain aggregated data. The aggregated data is then broadcast across the heterogeneous communication domain to be sent to master nodes of other homogeneous sets.

Master nodes receive aggregated data sent from other master nodes. According to the aggregated data of other master nodes and its own aggregated data, it is determined to update the calculation data to be processed. Then send the calculation data to be processed to each node in the same set, so that the next round of calculation can be performed based on the new calculation data to be processed.

In related technologies, collective communication is designed based on homogeneous hardware, and heterogeneous hardware is not supported for communication.

According to the method for executing distributed computing tasks in the embodiments of the present disclosure, heterogeneous nodes can be used for distributed computing, hardware resources can be used more efficiently, and computing efficiency can be improved.

Fig. 6 schematically shows a block diagram of an apparatus for performing distributed computing tasks according to an embodiment of the present disclosure.

As shown in FIG. 6 , the apparatus 600 for performing distributed computing tasks includes a first sending module 610 , an aggregation module 620 , a sharing module 630 and a determination module 640 .

The first sending module 610 is configured to send the computing data to be processed related to the distributed computing task to each first node in the first node set, wherein the hardware information of the first node matches the local hardware information.

The aggregation module 620 is configured to aggregate the first calculation results of each first node for the calculation data to be processed to obtain the first aggregation result.

The sharing module 630 is configured to share the first aggregation result with each second node in the second node set to obtain the sharing result, wherein the hardware information of the second node does not match the local hardware information.

The determining module 640 is configured to determine the shared result as new computing data to be processed, and return to the operation of sending the computing data to be processed to each node in the first node set until the execution of the distributed computing task is completed.

Fig. 7 schematically shows a block diagram of an apparatus for executing distributed computing tasks according to another embodiment of the present disclosure.

As shown in FIG. 7 , the apparatus 700 for performing distributed computing tasks includes a receiving module 710 , a computing module 720 and a sending module 730 .

The receiving module 710 is configured to receive computing data to be processed from the master node.

The calculation module 720 is configured to perform a calculation operation according to the calculation data to be processed to obtain a first calculation result.

A sending module 730, configured to send the first calculation result to the master node.

According to the embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

FIG. 8 schematically shows a block diagram of an example electronic device 800 that may be used to implement embodiments of the present disclosure. Electronic device is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are by way of example only, and are not intended to limit implementations of the disclosure described and/or claimed herein.

As shown in FIG. 8, the device 800 includes a computing unit 801 that can execute according to a computer program stored in a read-only memory (ROM) 802 or loaded from a storage unit 808 into a random access memory (RAM) 803. Various appropriate actions and treatments. In the RAM 803, various programs and data necessary for the operation of the device 800 can also be stored. The computing unit 801 , ROM 802 , and RAM 803 are connected to each other through a bus 804 . An input/output (I/O) interface 805 is also connected to the bus 804 .

Multiple components in the device 800 are connected to the I/O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, an optical disk, etc. ; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 809 allows the device 800 to exchange information/data with other devices over a computer network such as the Internet and/or various telecommunication networks.

The computing unit 801 may be various general-purpose and/or special-purpose processing components having processing and computing capabilities. Some examples of computing units 801 include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processing processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 executes various methods and processes described above, for example, a method of executing distributed computing tasks. For example, in some embodiments, a method for performing distributed computing tasks may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 808 . In some embodiments, part or all of the computer program may be loaded and/or installed onto the device 800 via the ROM 802 and/or the communication unit 809 . When the computer program is loaded into RAM 803 and executed by computing unit 801, one or more steps of the method for performing distributed computing tasks described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured in any other appropriate way (for example, by means of firmware) to execute a method for performing a distributed computing task.

Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips Implemented in a system of systems (SOC), complex programmable logic device (CPLD), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include being implemented in one or more computer programs executable and/or interpreted on a programmable system including at least one programmable processor, the programmable processor Can be special-purpose or general-purpose programmable processor, can receive data and instruction from storage system, at least one input device, and at least one output device, and transmit data and instruction to this storage system, this at least one input device, and this at least one output device an output device.

Program codes for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special purpose computer, or other programmable data processing devices, so that the program codes, when executed by the processor or controller, make the functions/functions specified in the flow diagrams and/or block diagrams Action is implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer discs, hard drives, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, compact disk read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.

To provide for interaction with the user, the systems and techniques described herein can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user. ); and a keyboard and pointing device (eg, a mouse or a trackball) through which a user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and can be in any form (including Acoustic input, speech input or, tactile input) to receive input from the user.

The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., as a a user computer having a graphical user interface or web browser through which a user can interact with embodiments of the systems and techniques described herein), or including such backend components, middleware components, Or any combination of front-end components in a computing system. The components of the system can be interconnected by any form or medium of digital data communication, eg, a communication network. Examples of communication networks include: Local Area Network (LAN), Wide Area Network (WAN) and the Internet.

A computer system may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by computer programs running on the respective computers and having a client-server relationship to each other.

The server can be a cloud server, also known as cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the problem of traditional physical host and VPS service ("Virtual Private Server", or "VPS") Among them, there are defects such as difficult management and weak business scalability. The server can also be a server of a distributed system, or a server combined with a blockchain.

It should be understood that steps may be reordered, added or deleted using the various forms of flow shown above. For example, each step described in the present disclosure may be executed in parallel, sequentially, or in a different order, as long as the desired result of the technical solution disclosed in the present disclosure can be achieved, no limitation is imposed herein.

The specific implementation manners described above do not limit the protection scope of the present disclosure. It should be apparent to those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made depending on design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present disclosure shall be included within the protection scope of the present disclosure.

Claims (13)

1. A method for performing distributed computing tasks, applied to a master node, the method comprising: Send the computing data to be processed related to the distributed computing task to each first node in the first node set through the first communication domain, wherein the hardware information of the first node matches the local hardware information, and the The first communication domain is established between the master node and each first node in the first node set; the local hardware information is hardware information of the master node; Aggregating first calculation results of each of the first nodes for the calculation data to be processed through the first communication domain to obtain a first aggregation result; Share the first aggregation result with each second node in the second node set through the second communication domain to obtain a shared result, wherein the hardware information of the second node does not match the local hardware information, so establishing the second communication domain between the master node and each second node in the set of second nodes; and determining the shared result as new computing data to be processed, and returning the operation of sending the computing data to be processed related to the distributed computing task to each first node in the first node set, until the distributed The calculation task is completed. 2. The method of claim 1, further comprising: Obtain the local hardware information and hardware information of multiple nodes; Determining a node whose hardware information matches the local hardware information among the plurality of nodes as a first node to obtain the first node set; and Determining a master node among the multiple nodes whose hardware information does not match the local hardware information, as the second node, to obtain the second node set. 3. The method of claim 1 or 2, further comprising: establishing a first communication domain with each first node in the set of first nodes; Wherein, the sending the computing data to be processed related to the distributed computing task to each first node in the first node set includes: The computation data to be processed is broadcast in the first communication domain. 4. The method according to claim 1, wherein the aggregating the first calculation results of each first node for the calculation data to be processed to obtain a first aggregation result comprises: performing calculation operations on the calculation data to be processed to obtain a second calculation result; receiving a first calculation result from each of said first nodes; and performing a weighted summation of the second calculation result and the first calculation result of each first node to obtain the first aggregation result. 5. The method of claim 1 or 2, further comprising: establishing a second communication domain with each second node in the set of second nodes; Wherein, the sharing the first aggregation result with each second node in the second node set, and obtaining the sharing result includes: broadcasting the first aggregation result in the second communication domain; receiving a second aggregated result from each of said second nodes; and The shared result is determined according to the first aggregated result and the second aggregated result. 6. The method according to claim 5, wherein said determining said shared result according to said first aggregated result and said second aggregated result comprises: calculating an average value of the first aggregation result and the second aggregation result as the shared result. 7. A method for performing a distributed computing task, applied to a first node, the method comprising: Receive the computing data to be processed from the master node through the first communication domain, wherein the master node and the first node belong to the same node set, and the hardware information of the nodes in the node set matches each other, and the Establishing the first communication domain between the master node and one or more first nodes in the set of nodes except the master node; performing calculation operations according to the calculation data to be processed to obtain a first calculation result; and sending the first calculation result to the master node through the first communication domain; Wherein, the master node is the master node according to any one of claims 1-6. 8. The method of claim 7, further comprising: Obtain hardware information and node identities for multiple nodes; and Determine the master node among the multiple nodes according to the hardware information and node identifiers in the multiple nodes. 9. The method of claim 8, further comprising: establishing a first communication domain with the master node; Wherein, sending the first calculation result to the master node includes: Send the first calculation result to the master node through the first communication domain. 10. A device for performing distributed computing tasks, applied to a master node, comprising: The first sending module is configured to send the computing data to be processed related to the distributed computing task to each first node in the first node set through the first communication domain, wherein the hardware information of the first node is related to the The local hardware information matches, and the first communication domain is established between the master node and each first node in the first node set; the local hardware information is hardware information of the master node; An aggregation module, configured to aggregate the first calculation results of each first node for the calculation data to be processed through the first communication domain to obtain a first aggregation result; A sharing module, configured to share the first aggregation result with each second node in the second node set through a second communication domain to obtain a shared result, wherein the hardware information of the second node is the same as that of the local hardware The information does not match, and the second communication domain is established between the master node and each second node in the set of second nodes; and A determining module, configured to determine the shared result as new computing data to be processed, and return the operation of sending the computing data to be processed related to the distributed computing task to each first node in the first node set, until the execution of the distributed computing task is completed. 11. An apparatus for performing distributed computing tasks, applied to a first node, comprising: The receiving module is configured to receive computing data to be processed from the master node through the first communication domain, wherein the master node and the first node belong to the same node set, and the hardware information between the nodes in the node set matching, establishing the first communication domain between the master node and one or more first nodes in the node set except the master node; A calculation module, configured to perform a calculation operation according to the calculation data to be processed to obtain a first calculation result; and a sending module, configured to send the first calculation result to the master node through the first communication domain; Wherein, the master node is the master node according to claim 10 . 12. An electronic device comprising: at least one processor; and a memory communicatively coupled to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can perform any one of claims 1-9. Methods. 13. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the method according to any one of claims 1-9.

Images