CN111736988A - Heterogeneous acceleration method, equipment and device and computer readable storage medium - Google Patents
Heterogeneous acceleration method, equipment and device and computer readable storage medium Download PDFInfo
- Publication number
- CN111736988A CN111736988A CN202010476916.XA CN202010476916A CN111736988A CN 111736988 A CN111736988 A CN 111736988A CN 202010476916 A CN202010476916 A CN 202010476916A CN 111736988 A CN111736988 A CN 111736988A
- Authority
- CN
- China
- Prior art keywords
- cloud computing
- computing system
- utilization rate
- processing
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 108
- 230000001133 acceleration Effects 0.000 title claims abstract description 79
- 238000012545 processing Methods 0.000 claims abstract description 68
- 230000008569 process Effects 0.000 claims abstract description 50
- 238000012805 post-processing Methods 0.000 claims description 40
- 238000007781 pre-processing Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 11
- 238000013527 convolutional neural network Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 230000002087 whitening effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The invention discloses a heterogeneous acceleration method, equipment, a device and a computer readable storage medium, wherein the method comprises the following steps: acquiring configuration and use information of the whole system; the whole system comprises an edge device and a cloud computing system; determining the current utilization rate of the whole system according to the configuration use information; according to the current utilization rate of the whole system, the edge devices and the cloud computing system are distributed to execute the processing process of the data to be processed acquired by the corresponding edge devices, so that heterogeneous acceleration of the data to be processed is completed; according to the invention, the edge devices and the cloud computing system are distributed to execute the processing process of the data to be processed acquired by the corresponding edge devices according to the current utilization rate of the whole system, and the processing process of each data to be processed, which is required to be executed by the edge devices and the cloud computing system, is distributed on the whole system level of the edge devices and the cloud computing system, so that the processing efficiency of the data to be processed is improved, and the efficiency of the edge devices and the cloud computing system is improved.
Description
Technical Field
The present invention relates to the field of heterogeneous acceleration technologies, and in particular, to a heterogeneous acceleration method, device, and apparatus, and a computer-readable storage medium.
Background
Currently, many companies are providing heterogeneous acceleration services and products of different Edge devices (Edge devices) or cloud computing systems. The Heterogeneous Acceleration (HSA) generally includes a design division configuration and integration of software and hardware Acceleration executed by a processor, and the Heterogeneous System is often an edge device or a cloud computing System. For large-scale Data such as Big Data (Big Data) and real-time images, the current edge device or cloud computing system can achieve different degrees of performance; however, the heterogeneous system acceleration in the prior art does not evaluate how to allocate or combine different edge devices and cloud computing systems in a systematic manner, and cannot efficiently and dynamically allocate the computing performance of the edge devices and the cloud computing systems and dynamically reconfigure the hardware acceleration function.
Therefore, how to efficiently and dynamically allocate the operation performance of the edge device and the cloud computing system and dynamically reconfigure the hardware acceleration function to improve the performance of the edge device and the cloud computing system is a problem that needs to be solved.
Disclosure of Invention
The invention aims to provide a heterogeneous acceleration method, equipment, a device and a computer readable storage medium, which are used for distributing heterogeneous acceleration processes required to be executed by an edge device and a cloud computing system by a whole system level formed by the edge device and the cloud computing system, so that the efficiency of the edge device and the cloud computing system is improved.
In order to solve the above technical problem, the present invention provides a heterogeneous acceleration method, including:
acquiring configuration and use information of the whole system; the whole system comprises an edge device and a cloud computing system;
determining the current utilization rate of the whole system according to the configuration use information;
and according to the current utilization rate of the whole system, distributing the edge devices and the cloud computing system to execute the processing process of the to-be-processed data acquired by the corresponding edge devices so as to finish the heterogeneous acceleration of the to-be-processed data.
Optionally, the allocating, according to the current usage rate of the whole system, a processing procedure in which the edge device and the cloud computing system execute the to-be-processed data acquired by the edge device corresponding to each other includes:
according to the current utilization rate of the whole system, a target edge device and the cloud computing system are allocated to execute the processing process of the data to be processed corresponding to each other; wherein the target edge device is the edge device that obtains the data to be processed.
Optionally, the allocating, according to the current usage rate of the whole system, a target edge device and the cloud computing system to execute a processing procedure of the to-be-processed data corresponding to each other includes:
and distributing the target edge device and the cloud computing system to execute the processing process of the data to be processed corresponding to the target edge device and the cloud computing system according to the current utilization rate of the cloud computing system.
Optionally, when the number of the cloud computing systems is 1 and the to-be-processed data is image data, allocating, according to the current usage rate of the cloud computing system, the target edge device and the cloud computing system to execute respective corresponding processing processes of the to-be-processed data, including:
determining a preset utilization rate range corresponding to the current utilization rate of the cloud computing system; the preset utilization rate range is a high utilization rate range, a medium utilization rate range or a low utilization rate range;
if the preset utilization rate range is the low utilization rate range, controlling the cloud computing system to execute pre-processing, main computing and post-processing in the processing process of the data to be processed;
if the preset utilization rate range is the medium utilization rate range, controlling the target edge device to execute the pre-processing, and controlling the cloud computing system to execute the main computing and the post-processing;
if the preset utilization rate range is the high utilization rate range, controlling the target edge device to execute the pre-processing, the main operation and the first part of post-processing, and controlling the cloud computing system to execute a second part of post-processing; wherein the second part of post-processing is storage and recording in the post-processing; the first part of post-processing is content in the post-processing other than the second part of post-processing.
Optionally, the edge device is specifically an embedded FPGA, and the cloud computing system is specifically a system composed of a server and a PCIe acceleration card.
Optionally, the allocating, according to the current usage rate of the whole system, a processing procedure in which the edge device and the cloud computing system execute the to-be-processed data acquired by the edge device corresponding to each other includes:
and distributing the edge device and the cloud computing system to execute the processing process of the corresponding data to be processed according to the current utilization rate and the computing efficiency of the whole system and the required computing efficiency corresponding to the data to be processed.
Optionally, before allocating the edge device and the cloud computing system to execute the processing process of the respective corresponding to-be-processed data according to the current utilization rate and the computing performance of the whole system and the required computing performance corresponding to the to-be-processed data, the method further includes:
and determining the operation efficiency of the whole system according to the configuration use information.
The present invention also provides a heterogeneous acceleration apparatus, including:
the acquisition module is used for acquiring the configuration and use information of the whole system; the whole system comprises an edge device and a cloud computing system;
the determining module is used for determining the current utilization rate of the whole system according to the configuration use information;
and the distribution module is used for distributing the edge devices and the cloud computing system to execute the processing process of the to-be-processed data acquired by the edge devices corresponding to each other according to the current utilization rate of the whole system so as to finish heterogeneous acceleration of the to-be-processed data.
The invention also provides a heterogeneous acceleration device, which comprises:
a memory for storing a computer program;
a processor for implementing the steps of the heterogeneous acceleration method as described above when executing the computer program.
The invention also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps of the heterogeneous acceleration method as described above.
The invention provides a heterogeneous acceleration method, which comprises the following steps: acquiring configuration and use information of the whole system; the whole system comprises an edge device and a cloud computing system; determining the current utilization rate of the whole system according to the configuration use information; according to the current utilization rate of the whole system, the edge devices and the cloud computing system are distributed to execute the processing process of the data to be processed acquired by the corresponding edge devices, so that heterogeneous acceleration of the data to be processed is completed;
therefore, according to the invention, the edge devices and the cloud computing system are distributed to execute the processing process of the to-be-processed data acquired by the corresponding edge devices according to the current utilization rate of the whole system, and the processing process of each to-be-processed data required to be executed by the edge devices and the cloud computing system is distributed on the whole system level of the edge devices and the cloud computing system, so that the processing efficiency of the to-be-processed data is improved, and the efficiency of the edge devices and the cloud computing system is improved. In addition, the invention also provides a heterogeneous acceleration device, a heterogeneous acceleration device and a computer readable storage medium, and the heterogeneous acceleration device, the heterogeneous acceleration device and the computer readable storage medium also have the beneficial effects.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a flowchart of a heterogeneous acceleration method according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a system architecture according to an embodiment of the present invention;
FIG. 3 is a block diagram of another system architecture according to an embodiment of the present invention;
fig. 4 is a block diagram of a heterogeneous acceleration device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, fig. 1 is a flowchart illustrating a heterogeneous acceleration method according to an embodiment of the present invention. The method can comprise the following steps:
step 101: acquiring configuration and use information of the whole system; the whole system comprises an edge device and a cloud computing system.
The whole system in this step may be a system composed of the edge device and the cloud computing system, that is, the whole system may include the edge device and the cloud computing system. The specific configuration of the whole system in this embodiment, that is, the specific number of the edge devices and the cloud computing systems in the whole system, may be set by a designer or a user according to a practical scenario and a user requirement, for example, the whole system may include one cloud computing system and a plurality of edge devices, and may also include a plurality of cloud computing systems and a plurality of edge devices. The present embodiment does not set any limit to this.
Correspondingly, the specific types of the edge device and the cloud computing system in the whole system in this embodiment may be set by a designer or a user according to a practical scenario and a user requirement, for example, the edge device may be a local device that can perform all or part of heterogeneous system acceleration, such as an embedded FPGA (Field programmable gate Array) in fig. 2, for example, an FPGA including a CPU, and may perform heterogeneous system acceleration (CPU + FPGA), or an FPGA including a GPU; the cloud computing system may be a system capable of performing heterogeneous system acceleration at the cloud, such as a system composed of a server and a PCIe acceleration card, for example, a server + PCIe FPGA acceleration card or a server + PCIe GPU acceleration card. The present embodiment does not set any limit to this.
It should be noted that the system-wide configuration usage information obtained by the processor in this step may be original information obtained from the system-wide configuration information required by the processing procedure for distributing the data to be processed. The specific content of the configuration and use information of the whole system in this step can be set by a designer according to a practical scene and user requirements, for example, the configuration and use information of the whole system can include load use information of each edge device and each cloud computing system in the whole system, which is required for determining the current use rate of the whole system, for example, CPU use rate information of each edge device and each cloud computing system; configuration information for each edge device and each cloud computing system in the system may also be included as needed to determine the overall computing performance of the system. The present embodiment does not set any limit to this.
It is to be understood that the heterogeneous acceleration method provided in this embodiment may be implemented by a processor in any cloud computing system or edge device in the whole system when executing the code of the program of the heterogeneous acceleration method, or may be implemented by a processor in an additional device (such as the cooperative controller in fig. 3) when executing the code of the program of the heterogeneous acceleration method, which is not limited in this embodiment.
Step 102: and determining the current utilization rate of the whole system according to the configuration utilization information.
The current utilization rate of the whole system in this step may be the respective current utilization rates of each edge device and each cloud computing system in the whole system.
Specifically, the specific manner in which the processor determines the current utilization rate of the whole system according to the configuration and use information of the whole system in this step may be set by a designer, and for example, the specific manner may be implemented in a manner the same as or similar to the utilization rate calculation method in the prior art, which is not limited in this embodiment.
It can be understood that the purpose of this step may be to determine, according to the configuration usage information, a current usage rate of the whole system corresponding to the to-be-processed data that needs to be subjected to heterogeneous acceleration, that is, a current usage rate of the edge devices and/or the cloud computing system in the whole system that is required for allocating the processing process of the to-be-processed data. For example, when the processing process of the data to be processed is distributed only by using the current utilization rate of the cloud computing system in the whole system, the current utilization rate of the cloud computing system can be determined according to the configuration utilization information of the cloud computing system in the whole system.
Step 103: according to the current utilization rate of the whole system, the edge devices and the cloud computing system are distributed to execute the processing process of the data to be processed acquired by the corresponding edge devices, so that heterogeneous acceleration of the data to be processed is completed.
The data to be processed in this step may be data that needs to be subjected to heterogeneous acceleration and is acquired by the edge device, such as original images (i.e., image data) sent by the camera in fig. 2 to the edge device. The processing procedure of the data to be processed in this step may be a heterogeneous accelerated processing procedure of the data to be processed that needs to be distributed to the edge device and the cloud computing system, that is, the edge device and the cloud computing system may execute all or part of corresponding processing procedures in all the processing procedures according to the distribution result, so that after the edge device and the cloud computing system execute the processing procedures obtained by respective distribution, heterogeneous acceleration of the data to be processed is completed.
Specifically, the specific content of the processing procedure of the data to be processed in this step may be set by a designer or a user according to a practical scene and a user requirement, for example, when the data to be processed is data of a big data type or image data of a real-time image, the processing procedure of the data to be processed may include pre-processing, main operation, and post-processing; the preprocessing may include data acquisition and simplification, such as clipping, mean subtraction, normalization, PCA (principal components Analysis), Whitening, and other processes of image data; the primary operation may include a course of operations that classify the inference; post-processing may include processes of displaying, recording, and storing.
It can be understood that the purpose of this step may be to allocate, for the processor, the processing procedure of the to-be-processed data acquired by the edge device to the corresponding edge device and the cloud computing system in the entire system according to the current utilization rate of the entire system, so as to complete the heterogeneous system acceleration of the to-be-processed data.
Specifically, the specific manner in which the processor executes the processing process of the to-be-processed data acquired by the edge device and the cloud computing system corresponding to each other according to the current utilization rate of the whole system in this step may be set by a designer according to a practical scene and a user requirement, for example, the processor may allocate any processing process of the to-be-processed data to the edge device (i.e., the target edge device) and the cloud computing system acquiring the to-be-processed data for execution, that is, the step may allocate the target edge device and the cloud computing system to execute the processing process of the to-be-processed data corresponding to each other according to the current utilization rate of the whole system; the target edge device is an edge device for acquiring data to be processed. The processor may also allocate a processing procedure of any data to be processed to the cloud computing system and the edge device corresponding to the data to be processed for execution, for example, in this step, a target edge device may be allocated, and the edge device and the cloud computing system may be selected to execute respective processing procedures of the data to be processed according to the current utilization rate of the whole system; wherein the edge device is selected to be the same as the location of the edge device.
Correspondingly, in the step, the processor can directly allocate the edge devices and the cloud computing system to execute the processing process of the data to be processed acquired by the corresponding edge devices according to the current utilization rate of the whole system; the processor can allocate the target edge device and the cloud computing system to execute the processing process of the data to be processed corresponding to the target edge device and the cloud computing system according to the current utilization rate of the cloud computing system and/or the current utilization rate of the edge devices in the current utilization rate of the whole system; for example, the processor may allocate the target edge device and the cloud computing system to execute the processing process of the respective corresponding to-be-processed data according to the current utilization rate of the cloud computing system in the current utilization rates of the entire system. The processor can also allocate the edge device and the cloud computing system to execute the processing process of the respective corresponding data to be processed according to the current utilization rate and the computing performance (such as the number of floating point operations per second or the number of processed images per second) of the whole system and the required computing performance corresponding to the data to be processed; that is to say, the processor may allocate the target edge device and the cloud computing system to execute the processing process of the corresponding to-be-processed data according to the current utilization rate of the cloud computing system and/or the current utilization rate of the edge device, of which the computing performance of the processor in the current utilization rates of the whole system satisfies the required computing performance.
Correspondingly, the step may further include a step of acquiring, by the processor, the overall system operational performance and the required operational performance corresponding to the data to be processed, for example, the processor may determine the overall system operational performance according to the configuration usage information.
It should be noted that, for the specific process of the processor in this step, according to the current utilization rate of the whole system, allocating the edge device and the cloud computing system to execute the processing process of the to-be-processed data acquired by the respective corresponding edge device, may be set by the designer, if the number of the cloud computing systems is 1 and the to-be-processed data is image data, the process of the processor allocating the target edge device and the cloud computing system to execute the processing process of the respective corresponding to-be-processed data according to the current utilization rate of the cloud computing system may include: determining a preset utilization rate range corresponding to the current utilization rate of the cloud computing system; wherein the preset utilization rate range is a high utilization rate range, a medium utilization rate range or a low utilization rate range; if the preset utilization rate range is the low utilization rate range, controlling the cloud computing system to execute pre-processing, main computing and post-processing in the processing process of the data to be processed; if the preset utilization rate range is the middle utilization rate range, controlling the target edge device to execute pre-processing, and controlling the cloud computing system to execute main computing and post-processing; if the preset utilization rate range is the high utilization rate range, controlling the target edge device to execute pre-processing, main operation and first part post-processing, and controlling the cloud computing system to execute second part post-processing; wherein the second part of post-processing is storage and recording in the post-processing; the first part of post-processing is the content in the post-processing except the second part of post-processing.
As shown in fig. 3 and table 1, the whole system includes a cloud computing system and 4 Edge devices (Edge 0-Edge 3), the first three Edge devices (Edge 0-Edge 2) perform the correlation operation of CIFAR-10 (a deep learning convolutional neural network and image database) CNN (convolutional neural network) classification, and the fourth Edge device performs the correlation operation of AlexNet (a deep learning convolutional neural network) CNN classification; at the time of T +0, the processor of the cooperative controller determines that a preset utilization rate range corresponding to the current utilization rate of the cloud computing system is a low utilization rate range, namely the current utilization rate of the cloud computing system is low, the processor can control Edge0 to only acquire and transmit image data and acquire a final heterogeneous acceleration result, and control the cloud computing system to perform preprocessing, main computing and postprocessing of CIFAR-10; at the time of T +1, the processor of the cooperative controller determines that a preset utilization rate range corresponding to the current utilization rate of the cloud computing system is a medium utilization rate range, namely the utilization rate of the cloud computing system is high, the processor can control Edge1 to acquire image data, pre-process and transmit CIFAR-10 and finally acquire heterogeneous acceleration results, and control the cloud computing system to perform main operation and post-process of CIFAR-10; at the time of T +2, the processor of the cooperative controller determines that a preset utilization rate range corresponding to the current utilization rate of the cloud computing system is a high utilization rate range, namely the utilization rate of the cloud computing system is very high, the processor can control Edge2 to acquire image data, pre-process, main operation, post-process and transmit CIFAR-10, and control the cloud computing system to only store and record a final heterogeneous acceleration result; at the time of T +3, when determining that the image data acquired by Edge3 needs to perform the actions related to AlexNet, the processor of the cooperative controller may control the Edge device (Edge3) and the PCIe of the cloud computing system to recombine into an AlexNet accelerator, so as to control Edge3 and the cloud computing system to complete AlexNet preprocessing, main computing, and post-processing of the image data acquired by Edge3 in the manner of the previous time (T + 0-T +2), so that Edge3 can acquire the final heterogeneous acceleration result; that is, the step may further include the processor adjusting the acceleration algorithm and/or the acceleration circuit of the cloud computing system and the edge device according to the calculation mode information of the data to be processed.
Table 1 table showing heterogeneous acceleration method of the whole system shown in fig. 3
In this embodiment, according to the current utilization rate of the whole system, the edge devices and the cloud computing system are allocated to execute the processing process of the to-be-processed data acquired by the corresponding edge devices, and the processing process of each to-be-processed data required to be executed by the edge devices and the cloud computing system is allocated on the whole system level of the edge devices and the cloud computing system, so that the processing efficiency of the to-be-processed data is improved, and the efficiency of the edge devices and the cloud computing system is improved.
Referring to fig. 4, fig. 4 is a block diagram of a heterogeneous acceleration device according to an embodiment of the present invention, where the device may include:
an obtaining module 10, configured to obtain configuration and usage information of the whole system; the whole system comprises an edge device and a cloud computing system;
a determining module 20, configured to determine a current usage rate of the whole system according to the configuration usage information;
the allocation module 30 is configured to allocate the edge devices and the cloud computing system to execute the processing process of the to-be-processed data acquired by the respective corresponding edge devices according to the current utilization rate of the whole system, so as to complete heterogeneous acceleration of the to-be-processed data.
Optionally, the allocation module 30 may be specifically configured to allocate the target edge device and the cloud computing system to execute a processing procedure of the data to be processed corresponding to each other according to the current utilization rate of the whole system; the target edge device is an edge device for acquiring data to be processed.
Optionally, the allocation module 30 may be specifically configured to allocate the target edge device and the cloud computing system to execute a processing process of respective corresponding to-be-processed data according to the current utilization rate of the cloud computing system.
Optionally, when the number of the cloud computing systems is 1 and the data to be processed is image data, the allocating module 30 may include:
the determining submodule is used for determining a preset utilization rate range corresponding to the current utilization rate of the cloud computing system; wherein the preset utilization rate range is a high utilization rate range, a medium utilization rate range or a low utilization rate range;
the first allocation submodule is used for controlling the cloud computing system to execute pre-processing, main computing and post-processing in the processing process of the data to be processed if the preset utilization rate range is the low utilization rate range;
the second distribution submodule is used for controlling the target edge device to execute pre-processing and controlling the cloud computing system to execute main operation and post-processing if the preset utilization rate range is the middle utilization rate range;
the third sub-distribution sub-module is used for controlling the target edge device to execute pre-processing, main operation and first part post-processing and controlling the cloud computing system to execute second part post-processing if the preset utilization rate range is the high utilization rate range; wherein the second part of post-processing is storage and recording in the post-processing; the first part of post-processing is the content in the post-processing except the second part of post-processing.
Optionally, the allocation module 30 may be specifically configured to allocate the edge device and the cloud computing system to execute the processing process of the respective corresponding to-be-processed data according to the current utilization and the computing performance of the whole system and the required computing performance corresponding to the to-be-processed data.
Optionally, the apparatus may further include:
and the operation efficiency determining module is used for determining the operation efficiency of the whole system according to the configuration use information.
In this embodiment, according to the present invention, the distribution module 30 distributes the processing processes of the to-be-processed data acquired by the edge devices and the cloud computing system corresponding to the edge devices according to the current utilization rate of the whole system, and distributes the processing processes of each to-be-processed data required to be executed by the edge devices and the cloud computing system on the whole system level of the edge devices and the cloud computing system, so as to improve the processing efficiency of the to-be-processed data and improve the performance of the edge devices and the cloud computing system.
An embodiment of the present invention further provides a heterogeneous acceleration apparatus, including: the method comprises the following steps: a processor and a memory, wherein the memory is used for storing a computer program, and the processor is used for implementing the steps of the heterogeneous acceleration method provided by the above embodiments when executing the computer program.
Specifically, the heterogeneous acceleration device provided by the present embodiment may be any edge device in the whole system; the system can also be any cloud computing system in the whole system, such as a server in the cloud computing system; other devices, such as cooperative controllers, communicatively coupled to each edge device and each cloud computing system in the overall system, respectively, may also be used.
The memory in this embodiment includes at least one type of readable storage medium, which includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, and the like. The memory may be an internal storage unit of the heterogeneous acceleration device in some embodiments. The memory may also be an external storage device of the heterogeneous acceleration apparatus in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like. Further, the memory may also include both an internal storage unit of the heterogeneous acceleration apparatus and an external storage device. The memory can be used for storing application software installed in the heterogeneous acceleration device and various types of data, such as: code that executes a heterogeneous acceleration method, etc., may also be used to temporarily store data that has been output or is to be output.
The processor in this embodiment may be a Central Processing Unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip in some embodiments, and is configured to run program codes stored in a memory or process data, such as codes of a program for executing a heterogeneous acceleration method.
Furthermore, an embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the heterogeneous acceleration method provided in the foregoing embodiment are implemented.
Wherein the computer-readable storage medium may include: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The embodiments are described in a progressive manner in the specification, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device, the apparatus and the computer-readable storage medium disclosed in the embodiments correspond to the method disclosed in the embodiments, so that the description is simple, and the relevant points can be referred to the description of the method.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The present invention provides a heterogeneous acceleration method, apparatus, device and computer readable storage medium. The principles and embodiments of the present invention are explained herein using specific examples, which are presented only to assist in understanding the method and its core concepts. It should be noted that, for those skilled in the art, it is possible to make various improvements and modifications to the present invention without departing from the principle of the present invention, and those improvements and modifications also fall within the scope of the claims of the present invention.
Claims (10)
1. A heterogeneous acceleration method, comprising:
acquiring configuration and use information of the whole system; the whole system comprises an edge device and a cloud computing system;
determining the current utilization rate of the whole system according to the configuration use information;
and according to the current utilization rate of the whole system, distributing the edge devices and the cloud computing system to execute the processing process of the to-be-processed data acquired by the corresponding edge devices so as to finish the heterogeneous acceleration of the to-be-processed data.
2. The heterogeneous acceleration method of claim 1, wherein the allocating, according to the current usage rate of the whole system, the edge devices and the cloud computing system to execute a processing procedure of the to-be-processed data obtained by the respective corresponding edge devices comprises:
according to the current utilization rate of the whole system, a target edge device and the cloud computing system are allocated to execute the processing process of the data to be processed corresponding to each other; wherein the target edge device is the edge device that obtains the data to be processed.
3. The heterogeneous acceleration method of claim 2, wherein the allocating, according to the current usage rate of the whole system, a target edge device and the cloud computing system to execute a processing procedure of the to-be-processed data, respectively, comprises:
and distributing the target edge device and the cloud computing system to execute the processing process of the data to be processed corresponding to the target edge device and the cloud computing system according to the current utilization rate of the cloud computing system.
4. The heterogeneous acceleration method of claim 3, wherein when the number of the cloud computing systems is 1 and the to-be-processed data is image data, the allocating the target edge device and the cloud computing system to execute the processing procedure of the to-be-processed data corresponding to each of the target edge device and the cloud computing system according to the current utilization rate of the cloud computing system includes:
determining a preset utilization rate range corresponding to the current utilization rate of the cloud computing system; the preset utilization rate range is a high utilization rate range, a medium utilization rate range or a low utilization rate range;
if the preset utilization rate range is the low utilization rate range, controlling the cloud computing system to execute pre-processing, main computing and post-processing in the processing process of the data to be processed;
if the preset utilization rate range is the medium utilization rate range, controlling the target edge device to execute the pre-processing, and controlling the cloud computing system to execute the main computing and the post-processing;
if the preset utilization rate range is the high utilization rate range, controlling the target edge device to execute the pre-processing, the main operation and the first part of post-processing, and controlling the cloud computing system to execute a second part of post-processing; wherein the second part of post-processing is storage and recording in the post-processing; the first part of post-processing is content in the post-processing other than the second part of post-processing.
5. The heterogeneous acceleration method of claim 1, wherein the edge device is an embedded FPGA, and the cloud computing system is a system consisting of a server and a PCIe acceleration card.
6. The heterogeneous acceleration method of any one of claims 1 to 5, wherein the allocating, according to the current usage rate of the whole system, the processing procedure in which the edge device and the cloud computing system execute the data to be processed acquired by the corresponding edge device includes:
and distributing the edge device and the cloud computing system to execute the processing process of the corresponding data to be processed according to the current utilization rate and the computing efficiency of the whole system and the required computing efficiency corresponding to the data to be processed.
7. The heterogeneous acceleration method of claim 6, wherein before the allocating the edge device and the cloud computing system to perform the processing of the corresponding data to be processed according to the current usage and the computing performance of the whole system and the required computing performance corresponding to the data to be processed, the method further comprises:
and determining the operation efficiency of the whole system according to the configuration use information.
8. A heterogeneous acceleration device, characterized by comprising:
the acquisition module is used for acquiring the configuration and use information of the whole system; the whole system comprises an edge device and a cloud computing system;
the determining module is used for determining the current utilization rate of the whole system according to the configuration use information;
and the distribution module is used for distributing the edge devices and the cloud computing system to execute the processing process of the to-be-processed data acquired by the edge devices corresponding to each other according to the current utilization rate of the whole system so as to finish heterogeneous acceleration of the to-be-processed data.
9. A heterogeneous acceleration device, characterized by comprising:
a memory for storing a computer program;
a processor for implementing the steps of the heterogeneous acceleration method according to any of claims 1 to 7 when executing said computer program.
10. A computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, which computer program, when being executed by a processor, carries out the steps of the heterogeneous acceleration method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010476916.XA CN111736988A (en) | 2020-05-29 | 2020-05-29 | Heterogeneous acceleration method, equipment and device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010476916.XA CN111736988A (en) | 2020-05-29 | 2020-05-29 | Heterogeneous acceleration method, equipment and device and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111736988A true CN111736988A (en) | 2020-10-02 |
Family
ID=72648005
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010476916.XA Withdrawn CN111736988A (en) | 2020-05-29 | 2020-05-29 | Heterogeneous acceleration method, equipment and device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111736988A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112395089A (en) * | 2020-11-19 | 2021-02-23 | 联通智网科技有限公司 | Cloud heterogeneous computing method and device |
| CN113609068A (en) * | 2021-08-10 | 2021-11-05 | 中国人民解放军61646部队 | Cloud service architecture based on hybrid heterogeneous processor |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150195372A1 (en) * | 2012-07-27 | 2015-07-09 | Nokia Corporation | Methods and apparatuses for facilitating utilization of cloud services |
| CN107087019A (en) * | 2017-03-14 | 2017-08-22 | 西安电子科技大学 | A kind of end cloud cooperated computing framework and task scheduling apparatus and method |
| CN108664318A (en) * | 2018-05-11 | 2018-10-16 | 北京邮电大学 | Computation migration method and server-side, boundary server-side for computation migration |
| US20190245806A1 (en) * | 2018-02-07 | 2019-08-08 | Cisco Technology, Inc. | Optimizing fog orchestration through edge compute resource reservation |
| CN110245013A (en) * | 2018-03-09 | 2019-09-17 | 北京京东尚科信息技术有限公司 | Internet of Things managing computing resources method and apparatus |
| CN110399210A (en) * | 2019-07-30 | 2019-11-01 | 中国联合网络通信集团有限公司 | Method for scheduling task and device based on edge cloud |
| CN111124531A (en) * | 2019-11-25 | 2020-05-08 | 哈尔滨工业大学 | Dynamic unloading method for calculation tasks based on energy consumption and delay balance in vehicle fog calculation |
-
2020
- 2020-05-29 CN CN202010476916.XA patent/CN111736988A/en not_active Withdrawn
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150195372A1 (en) * | 2012-07-27 | 2015-07-09 | Nokia Corporation | Methods and apparatuses for facilitating utilization of cloud services |
| CN107087019A (en) * | 2017-03-14 | 2017-08-22 | 西安电子科技大学 | A kind of end cloud cooperated computing framework and task scheduling apparatus and method |
| US20190245806A1 (en) * | 2018-02-07 | 2019-08-08 | Cisco Technology, Inc. | Optimizing fog orchestration through edge compute resource reservation |
| CN110245013A (en) * | 2018-03-09 | 2019-09-17 | 北京京东尚科信息技术有限公司 | Internet of Things managing computing resources method and apparatus |
| CN108664318A (en) * | 2018-05-11 | 2018-10-16 | 北京邮电大学 | Computation migration method and server-side, boundary server-side for computation migration |
| CN110399210A (en) * | 2019-07-30 | 2019-11-01 | 中国联合网络通信集团有限公司 | Method for scheduling task and device based on edge cloud |
| CN111124531A (en) * | 2019-11-25 | 2020-05-08 | 哈尔滨工业大学 | Dynamic unloading method for calculation tasks based on energy consumption and delay balance in vehicle fog calculation |
Non-Patent Citations (1)
| Title |
|---|
| 殷佳等: "基于移动边缘计算的任务迁移和协作式负载均衡机制", 《计算机科学》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112395089A (en) * | 2020-11-19 | 2021-02-23 | 联通智网科技有限公司 | Cloud heterogeneous computing method and device |
| CN113609068A (en) * | 2021-08-10 | 2021-11-05 | 中国人民解放军61646部队 | Cloud service architecture based on hybrid heterogeneous processor |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111950723B (en) | Neural network model training method, image processing method, device and terminal equipment | |
| CN107621973B (en) | Cross-cluster task scheduling method and device | |
| CN108229419B (en) | Method and apparatus for clustering images | |
| CN111736988A (en) | Heterogeneous acceleration method, equipment and device and computer readable storage medium | |
| CN114416352A (en) | Computing resource allocation method and device, electronic equipment and storage medium | |
| US9946645B2 (en) | Information processing apparatus and memory control method | |
| CN112651953A (en) | Image similarity calculation method and device, computer equipment and storage medium | |
| KR101827167B1 (en) | Method and apparatus for high speed images stitching using sift parallization | |
| CN115269118A (en) | Scheduling method, device and equipment of virtual machine | |
| CN114915753A (en) | Architecture of cloud server, data processing method and storage medium | |
| CN112132215B (en) | Method, device and computer readable storage medium for identifying object type | |
| US11281935B2 (en) | 3D object detection from calibrated 2D images | |
| CN112465050A (en) | Image template selection method, device, equipment and storage medium | |
| CN111754401A (en) | Decoder training method, high-definition face image generation device and computer equipment | |
| CN116129496A (en) | Image shielding method and device, computer equipment and storage medium | |
| CN111539281A (en) | Distributed face recognition method and system | |
| CN116129325A (en) | Urban treatment image target extraction method and device and application thereof | |
| CN115909009A (en) | Image recognition method, image recognition device, storage medium and electronic equipment | |
| CN113421317B (en) | Method and system for generating image and electronic equipment | |
| CN117975076A (en) | Image classification method, electronic device and storage medium | |
| CN113221835A (en) | Scene classification method, device, equipment and storage medium for face-check video | |
| CN114296965A (en) | Feature retrieval method, feature retrieval device, electronic equipment and computer storage medium | |
| CN114443873A (en) | Data processing method, device, server and storage medium | |
| CN112766277A (en) | Channel adjustment method, device and equipment of convolutional neural network model | |
| CN114553700B (en) | Device grouping method, device, computer device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20201002 |