CN109857560A - A kind of collaboration parallelization mechanism based on CPU/GPU isomerous environment - Google Patents
A kind of collaboration parallelization mechanism based on CPU/GPU isomerous environment Download PDFInfo
- Publication number
- CN109857560A CN109857560A CN201910082779.9A CN201910082779A CN109857560A CN 109857560 A CN109857560 A CN 109857560A CN 201910082779 A CN201910082779 A CN 201910082779A CN 109857560 A CN109857560 A CN 109857560A
- Authority
- CN
- China
- Prior art keywords
- gpu
- cpu
- node
- collaboration
- worker
- 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.)
- Pending
Links
Landscapes
- Devices For Executing Special Programs (AREA)
Abstract
The collaboration parallelization mechanism based on CPU/GPU isomerous environment that the invention proposes a kind of is realized and greatly improves its speed when big data calculates and handles.One side CPU provides data to GPU and receives the data that GPU is passed back, manages the work of GPU;Another aspect CPU and GPU collaboration are parallel to complete calculating task, in such a way that threshold value is set, compare this different task to request CPU and the different of GPU resource, number can be received under this task using CPU and using GPU by calculating separately out, select the processor that can accommodate most numbers.This mode that CPU or GPU is reasonably selected according to the loading condition of node, the calculating of current big data and processing speed are greatly improved.
Description
Technical field
The present invention relates to industrial equipment state data analysis fields, and in particular to is based on CPU/GPU isomerous environment to a kind of
Collaboration parallelization mechanism.
Background technique
On the one hand collaboration parallelization mechanism based on CPU/GPU isomerous environment, CPU manage the work of GPU, on the other hand join
It carries out CPU-GPU according to the loading condition of node with the calculating task of part and flexibly selects.It can accelerate to the maximum extent
The speed of operation and processing data.
Have closest to technology of the invention:
(1), the cooperated computing mode based on CPU+GPU: CPU is merely responsible for the work of management GPU, provides data simultaneously for GPU
The data that GPU is passed back are received, entire calculating task is undertaken by GPU.The CPU and CPU division of labor is clear, but wastes valuable CPU meter
Calculate resource.
(2), the cooperated computing load equilibrium design based on CPU+GPU: CPU individually undertakes the calculating task of a part,
But load balancing at this time is difficult to accomplish.
Due to the restriction of various history and practical reasons, Heterogeneous Computing still suffers from the problem of all various aspects, wherein most
Distinct issues are program development difficulties, and this problem is more prominent when especially expanding to cluster scale rank.Main performance
Scalability, load balancing, adaptivity, communication, in terms of.
Summary of the invention
To solve shortcoming and defect in the prior art, the invention proposes the collaborations based on CPU/GPU isomerous environment simultaneously
Row mechanism substantially increases calculating and handles the speed of data.
The technical solution of the present invention is as follows:
A kind of collaboration parallelization mechanism based on CPU/GPU isomerous environment, CPU not only manage the work of GPU, also participation portion
The calculating task divided suitably flexibly selects CPU or GPU according to the loading condition of node, comprising the following steps:
Step (1), algorithm count the computing resource of each calculate node in cloud environment before topological operation submission;
Step (2), when this topology submit after, obtain each worker of topology resource request, by this request with it is each
The available resources of a cloud node compare, and when the available resources of certain cloud node are greater than the request of this worker, this worker is advised
It draws and arrives this node;
Step (3) requests CPU and the different of GPU resource by comparing this worker, calculates separately out and is using CPU
With use this worker in the case where GPU that can be received number, select the processor that can accommodate most numbers.
Beneficial effects of the present invention:
(1) management to be worked by CPU GPU, provides data for GPU and receives the data that GPU is passed back, by the fortune of GPU
Efficiency is calculated to be greatly improved on original base;
(2) CPU and GPU cooperates with operation, and the computing capability of CPU is adequately used, cleverer to the processing of data
It is living;
(3) loading condition for passing through each node when operation compares resource and requests the difference of CPU and GPU, selects energy
The processor for accommodating most numbers, solves the problems, such as load imbalance.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is that the present invention is based on the collaboration parallel processing flow charts under CPU/GPU isomerous environment:
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
As shown in Figure 1, the core of the collaboration parallelization mechanism under the isomerous environment of the invention based on CPU/GPU is pair
The reasonable selection of CPU and GPU, by the way that two threshold values: CPU usage α and RAM utilization rate β are arranged, according to the load feelings of node
Condition carries out the flexible choice of CPU-GPU.
Below with reference to figure, the detailed process based on the collaboration parallelization mechanism under CPU/GPU isomerous environment is carried out detailed
Illustrate:
Step (1), algorithm count the computing resource of each calculate node in cloud environment before topological operation submission;
Step (2), when this topology submit after, obtain each worker of topology resource request, by this request with it is each
The available resources of a cloud node compare, and when the available resources of certain cloud node are greater than the request of this worker, this worker is advised
It draws and arrives this node;
Step (3) requests CPU and the different of GPU resource by comparing this worker, calculates separately out and is using CPU
With use this worker in the case where GPU that can be received number, select the processor that can accommodate most numbers.
It is of the invention based on the collaboration parallelization mechanism under CPU/GPU isomerous environment, one side CPU provides data to GPU
And receive the data that GPU is passed back, another aspect CPU and the parallel completion calculating task of GPU collaboration, in such a way that threshold value is set,
CPU or GPU is reasonably selected according to the loading condition of node, the calculating of current big data and processing speed are greatly improved.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (1)
1. a kind of collaboration parallelization mechanism based on CPU/GPU isomerous environment, which is characterized in that can be according to the load feelings of node
The flexible choice of condition progress CPU-GPU, comprising the following steps:
Step (1), algorithm count the computing resource of each calculate node in cloud environment before topological operation submission;
Step (2), when this topology submit after, obtain each worker of topology resource request, by this request and each cloud
The available resources of node compare, and when the available resources of certain cloud node are greater than the request of this worker, this worker is planned for
This node;
Step (3) requests the different of CPU and GPU resource by comparing this worker, and calculating separately out using CPU and makes
It can be received number with this worker in the case where GPU, select the processor that can accommodate most numbers.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910082779.9A CN109857560A (en) | 2019-01-28 | 2019-01-28 | A kind of collaboration parallelization mechanism based on CPU/GPU isomerous environment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910082779.9A CN109857560A (en) | 2019-01-28 | 2019-01-28 | A kind of collaboration parallelization mechanism based on CPU/GPU isomerous environment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109857560A true CN109857560A (en) | 2019-06-07 |
Family
ID=66896599
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910082779.9A Pending CN109857560A (en) | 2019-01-28 | 2019-01-28 | A kind of collaboration parallelization mechanism based on CPU/GPU isomerous environment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109857560A (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103902387A (en) * | 2014-04-29 | 2014-07-02 | 浪潮电子信息产业股份有限公司 | Dynamic load balancing method for CPU+GPU CPPC |
CN105049485A (en) * | 2015-06-09 | 2015-11-11 | 中国石油大学(华东) | Real-time video processing oriented load-aware cloud calculation system |
-
2019
- 2019-01-28 CN CN201910082779.9A patent/CN109857560A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103902387A (en) * | 2014-04-29 | 2014-07-02 | 浪潮电子信息产业股份有限公司 | Dynamic load balancing method for CPU+GPU CPPC |
CN105049485A (en) * | 2015-06-09 | 2015-11-11 | 中国石油大学(华东) | Real-time video processing oriented load-aware cloud calculation system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105912399B (en) | Task processing method, device and system | |
CN102724103B (en) | Proxy server, hierarchical network system and distributed workload management method | |
Rasooli et al. | An adaptive scheduling algorithm for dynamic heterogeneous Hadoop systems | |
CN104239144A (en) | Multilevel distributed task processing system | |
CN104375882B (en) | The multistage nested data being matched with high-performance computer structure drives method of calculation | |
Mahato et al. | On scheduling transactions in a grid processing system considering load through ant colony optimization | |
Lin et al. | A model-based approach to streamlining distributed training for asynchronous SGD | |
Bukhsh et al. | A decentralized edge computing latency-aware task management method with high availability for IoT applications | |
CN111082971A (en) | Shared resource allocation method for cloud load test | |
Bahnasawy et al. | A new algorithm for static task scheduling for heterogeneous distributed computing systems | |
CN109857560A (en) | A kind of collaboration parallelization mechanism based on CPU/GPU isomerous environment | |
CN116089083A (en) | Multi-target data center resource scheduling method | |
Kaur et al. | Improved hyper-heuristic scheduling with load-balancing and RASA for cloud computing systems | |
CN110046046A (en) | A kind of distributed hyperparameter optimization system and method based on Mesos | |
Javadi-Moghaddam et al. | Resource allocation in cloud computing using advanced imperialist competitive algorithm. | |
Goyal et al. | Adaptive and dynamic load balancing methodologies for distributed environment: a review | |
Kumari et al. | A Round-Robin based Load balancing approach for Scalable demands and maximized Resource availability | |
CN109298932B (en) | OpenFlow-based resource scheduling method, scheduler and system | |
CN112346852A (en) | Distributed physical processing of matrix summation operations | |
CN112817761A (en) | Energy-saving method for enhancing cloud computing environment | |
Rai et al. | An efficient distributed dynamic load balancing method based on hybrid approach in cloud computing | |
Tat et al. | A co-ordinate based resource allocation strategy for grid environments | |
Devi et al. | Improving fault tolerant resource optimized aware job scheduling for grid computing | |
CN114706667B (en) | Streaming media forwarding method based on heterogeneous computation | |
Hasan et al. | Successive stage multi-round scheduling for cube based multi-processor systems |
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 | ||
CB03 | Change of inventor or designer information |
Inventor after: Zhang Weishan Inventor after: Zhang Ruicong Inventor after: Fang Kai Inventor before: Zhang Ruicong Inventor before: Zhang Weishan Inventor before: Fang Kai |
|
CB03 | Change of inventor or designer information | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190607 |
|
WD01 | Invention patent application deemed withdrawn after publication |