CN108776589B - Deployment method of radar signal processing software component - Google Patents
Deployment method of radar signal processing software component Download PDFInfo
- Publication number
- CN108776589B CN108776589B CN201810563967.9A CN201810563967A CN108776589B CN 108776589 B CN108776589 B CN 108776589B CN 201810563967 A CN201810563967 A CN 201810563967A CN 108776589 B CN108776589 B CN 108776589B
- Authority
- CN
- China
- Prior art keywords
- component
- graph
- components
- time
- group
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/60—Software deployment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/60—Software deployment
- G06F8/65—Updates
Abstract
The invention discloses a deployment method of a radar signal processing software component, which comprises the following steps: the method comprises the steps of information acquisition, graph analysis, detail graph component information acquisition, component sequencing, critical path calculation, component and CPU core grouping in the graph, adaptation of a sub-graph component group and a core group, optimal deployment scheme selection and deployment scheme pushing. The invention realizes the dynamic deployment of the radar signal processing software component to the hardware processor, so that the radar signal processor has the capability of quickly responding to the change of the radar working mode, thereby achieving the purpose of self-adaptive adjustment of the working mode.
Description
Technical Field
The invention relates to a radar signal processing method, in particular to a deployment method of a radar signal processing software component.
Background
In modern war, the battlefield environment and the battle object faced by radar are complex and various. Therefore, the research on the radar with stronger adaptability to battlefield targets and environments and the improvement of the battlefield viability of the radar are the main targets of the development of the radar technology. Radars that can intelligently select a transmission waveform, a working mode, and optimal allocation of resources according to a target and an environment are considered as important directions for future radar development. The working mode is an important parameter of the radar signal processor, the selection of the working mode is determined by the transmitting waveform, and the transmitting waveform is changed according to the battlefield scene and the battle target. When the working mode is changed, the signal processor also needs to change the flow of processing the echo data correspondingly. Therefore, a technology capable of adaptively adjusting a processing flow according to a working mode is an important development direction of a radar signal processor software technology, namely, future radar signal processor software needs to have the capability of constructing a new working flow diagram in real time and rapidly deploying a new processing software component to a hardware processor.
The software design process of the present signal processor is that a designer designs a work flow chart according to a radar working mode, comprehensively considers and calculates factors such as platform processing capacity, economic requirements, performance indexes and the like, then manually writes operation codes of corresponding processors, and maps functional components to each processor, thereby realizing the processing of the process, and the software design process mainly has the following defects: firstly, the real-time increase, deletion and change of the working mode of the signal processor cannot be realized; secondly, when the radar works, the function and data flow direction of each working flow are fixed, and the function components, the data flow direction, the software and hardware mapping relation and the like cannot be adjusted in real time.
Disclosure of Invention
The technical problem to be solved by the invention is to overcome the defects of the prior art, provide a deployment method of a radar signal processing software component, realize the dynamic deployment of the radar signal processing software component to a hardware processor, and enable a radar signal processor to have the capability of quickly responding to the change of a radar working mode, thereby achieving the purpose of self-adaptive adjustment of the working mode.
In order to solve the technical problem, the invention provides a deployment method of a radar signal processing software component, which comprises the following steps:
(1) information acquisition: the system receives workflow diagram information input by a front end, wherein the workflow diagram information comprises description information and software and hardware constraint conditions;
(2) and (3) analysis of the graph: analyzing and quantizing the description information in the step (1) to obtain the serial number, the dependency relationship and the data set format information of the components in the graph, and storing the serial number, the dependency relationship and the data set format information of the components in the graph, which are obtained by analyzing and quantizing, by an adjacency list method;
(3) obtaining detailed graph component information: obtaining data dimension information which can be processed by the component and data passing rates of the component under different core numbers from a component database;
(4) and (3) component sequencing: sorting the components in the graph according to an in-degree sequence;
(5) and (3) calculating a critical path: obtaining the critical path length of the graph and a master-slave relationship table of each component by calculating the earliest EFT time and the latest LFT time of the components in each graph;
(6) the components in the figure and the CPU core are grouped as follows: according to the time sequence of the critical path, orderly grouping the components in the graph by adopting an enumeration method, wherein the component group number calculation formula is 2n-1; the CPU cores are grouped out of order, the number of the CPU groups is 2n-1Wherein n is the number of components;
(7) the sub-map component group is matched with the core group: calculating and storing the processing time of each group of the components in the graph in different core numbers to obtain the result of the adaptation of the group of the component of the graph and the core group;
(8) selecting an optimal deployment scheme: screening the results of the adaptation of the sub-image component group and the core group according to the input software and hardware constraint conditions, and selecting the combination with the minimum delay period and hardware scale for output to obtain the optimal combination;
(9) a pushing deployment scheme: and (4) pushing the best combination after the adaptation in the step (8) to each CPU of the round-robin architecture system.
Further, the workflow diagram in step (1) is modeled using a DAG directed acyclic graph model.
Further, the method for adjacency list in step (2) specifically includes the following steps:
1) establishing an adjacent task pool: the adjacent task pool comprises an adjacent matrix and a dependency relationship table, and the dependency relationship of the components is represented by adopting a method of a two-way linked list;
2) resolving the character string describing the dependency relationship: analyzing the character string to obtain the sequence numbers of the pre-adjacent point and the subsequent adjacent point in the graph and the data set transmitted between the pre-adjacent point and the subsequent adjacent point, and adding the pre-adjacent point and the subsequent adjacent point into a double linked list;
3) obtaining an adjacency matrix: and sequentially calculating the doubly linked list representing the dependency relationship to obtain the adjacency relationship among the components in the graph, namely the number of connecting lines between each component and other components.
Further, the critical path calculation in step (5) specifically includes the following steps:
1) and (3) calculating the entrance degree: calculating the entrance degree of each adjacent point (assembly), and obtaining the entrance degree of each assembly by inquiring the adjacent matrix;
2) sorting according to the degree of income: subtracting 1 from the degree of approach in sequence, recording adjacent points (assemblies) with zero degree of approach after subtracting 1 each time, and finally obtaining an assembly sequence ordered according to the degree of approach;
3) calculating the earliest time EFT;
4) calculating the latest occurrence time LFT;
5) obtaining a master node and a slave node: if EFT and LFT of a certain adjacent point (component) are equal, the component is a point on a critical path and is a master node, otherwise, the component is a slave node;
6) obtaining a critical path: and merging the slave nodes into the critical path to obtain a final critical path.
Further, the components and CPU core groups in the graph in step (6) are processed by a round-robin processing architecture method, and the specific steps are as follows:
1) calculating the completion time T of the whole workflow according to the obtained maximum parallelism of the components in each graph, and further obtaining the number Cn of the circulating CPUs, wherein the formula Cn is T/PRT, T is the completion time of the whole workflow, and PRT is a radar repetition period;
2) the resulting CPU number Cn is compared with the system-given maximum number max: if Cn < max, it is not necessary to perform graph division, and if Cn > max, it is necessary to perform graph division.
3) And (4) grouping the sequenced software components in the graph, wherein the sequencing of the software components between groups and in the groups is consistent with the sequence on the key path.
Further, the parallelism refers to the size of a data set input in the multi-core system, which can be processed by segmentation, and the data dimension processed by different software components is different.
Further, the specific steps of the step (7) are as follows:
1) calculating the critical path length: calculating the processing time of the node on the critical path in the single core, wherein the processing time is the longest time L of the systemtScreening the combination of the graph group and the kernel group by taking the time as a reference;
2) selecting a component group;
3) selecting a corresponding core group;
4) compute the time of the component under the current core group: computing group processing time P in assigning corest；
5) Screening and combining: comparison LtAnd Pt，LtGreater than PtRecording is done, otherwise it is discarded.
The invention achieves the following beneficial effects: the intelligent deployment method of the radar signal processing software component can realize real-time online addition, deletion, modification and other operations of working mode level, component level and component level modules in the system, and realize online reconstruction and upgrade of signal processing software by a software redeployment method.
Drawings
FIG. 1 is a schematic flow diagram of the present invention.
FIG. 2 is a schematic diagram of a radar signal processing software component deployment system according to the present invention.
Fig. 3 is a flow diagram of a radar signal processing function of the present invention (a) and a corresponding DAG diagram (b).
FIG. 4 is a schematic diagram of a radar signal processing workflow adjacency list according to the present invention.
FIG. 5 is a schematic diagram of a critical path calculation process of the method for deploying radar signal processing software components according to the present invention.
Detailed Description
Example 1
After receiving the workflow diagram description information input by the front end, the radar system firstly analyzes the description information to obtain the software components and the dependency relationship among the components required by the work, then sorts and divides the functional components in the diagram according to the context dependency relationship, then adapts each divided diagram to the kernel group to obtain the system delay time in various grouping, and selects the optimal component deployment method according to the input delay time and the hardware scale constraint condition, wherein the corresponding deployment system of the radar signal processing software is shown as an analyzer in figure 2 and comprises an external interface manager, a resource manager, a diagram and a scheduling scheme. The deployment method of the radar signal processing software component, as shown in fig. 1, comprises the following specific steps:
(1) and (3) information acquisition: and obtaining description information and software and hardware constraint conditions of the graph.
(2) And (3) analyzing the graph: analyzing and quantizing the input description of the graph to obtain information such as the serial number, the dependency relationship, the data set format and the like of the components in the graph, and storing the information by using an adjacency list method.
The working flow of the radar signal processor is shown in a flow chart mode, a plurality of software components are combined according to a certain dependency relationship, and each functional component of the radar signal processing functional flow chart is a specific processing method of the stream data at each stage in the operation process of the signal processor. A graph formed by connecting a plurality of software components according to a certain dependency relationship may be modeled by a DAG (direct Acyclic graph) Directed Acyclic graph model. The DAG graph is an important tool for modeling and solving the combined optimization problem, and the model can describe the attributes of each function in the flow chart in detail and clearly express the dependency relationship among the functions. Fig. 3 is a radar signal processing function flow diagram (a) and a corresponding DAG diagram (b). Wherein nodes represent task components, edges represent dependencies and communication relationships between tasks, and weights on edges represent traffic. The DAG represents dependencies between tasks using directed edges, G ═ V, E, T, C. (V: task, E: edge of task ni to nj, T represents computation time overhead of task ni, and represents communication time overhead of task ni to nj). The upper half of the circle is the serial number of the component in the schematic diagram (b) in fig. 3, and the lower half is the operation time of the component.
When parsing the input DAG graph, specifically: 1) establishing an adjacent task pool: the pool comprises an adjacency matrix, a dependency relationship table and the like, and the dependency relationship of the components is expressed by adopting a method of a double linked list. 2) Resolving the character string describing the dependency relationship: and analyzing the character string to obtain the sequence numbers of the pre-adjacent point and the subsequent adjacent point in the graph and the data set transmitted between the pre-adjacent point and the subsequent adjacent point, and adding the pre-adjacent point and the subsequent adjacent point into the double linked list. 3) Obtaining an adjacency matrix: and sequentially calculating the doubly linked list representing the dependency relationship to obtain the adjacency relationship among the components in the graph, namely the number of connecting lines between each component and other components. And the adjacency list is quantitatively stored by adopting an adjacency list storage method, so that the computer structure storage of the flow graph is realized. And an adjacency list method is adopted, so that inquiry and scheduling conditions are created for the deployment of the software components to the hardware platform in the subsequent graph. FIG. 4 is a representation of the adjacency list of FIG. 3 (b).
(3) Obtaining detailed component information: and obtaining data dimension information which can be processed by the component, data passing rate of the component under different core numbers and the like from the component database.
(4) Component ordering in the figure: and sorting the components in the graph according to the in-degree sequence.
(5) And (3) calculating a critical path: and calculating the LFT and EFT time of each component to obtain the critical path length of the graph and a master-slave relation table of each component.
The Critical Path (Critical Path) refers to the longest Path for completing all components in the DAG graph, and the completion time of all nodes on the Critical Path determines the time delay amount of the whole flow chart. When the critical path is obtained, the serial and parallel relations among the components in the DAG graph, a task list capable of being processed in parallel, the state of the software component at any time and the like can be obtained. The critical path is obtained mainly by calculating the earliest time (EFT) and the latest time of occurrence (LFT) of each component. EFT and LFT of each component can be calculated by a critical path method, the main node is a node on a critical path, the slave nodes are nodes on a non-critical path, and the EFT and the LFT can be processed in parallel with the main node. The critical path of each DAG graph is not uniform and varies with deployment scenario, data processing time, and communication time.
The steps of the critical path calculation are as follows: 1) calculating the degree of entrance and exit of each adjacent point (assembly): and querying the adjacency matrix to obtain the in-degree of each component. 2) Sorting according to the degree of entry: and sequentially subtracting 1 from the degree of approach, recording adjacent points (assemblies) with zero degree of approach after subtracting 1 each time, and finally obtaining an assembly sequence ordered according to the degree of approach, wherein if the adjacent points with zero degree of approach cannot be found after subtracting 1, a ring and a non-directed acyclic graph are represented in the graph. 3) Calculating the EFT: the earliest time EFT is calculated. 4) Calculating the LFT: the latest time of occurrence LFT is calculated. 5) Obtaining a master node and a slave node: if EFT and LFT of a certain adjacent point (component) are equal, the component is a point on the critical path and is a main node, otherwise, the component is a slave node. 6) Obtaining a critical path: and merging the slave nodes into the critical path to obtain a final critical path.
(6) The components in the figure and the CPU core are grouped as follows: orderly grouping the components by adopting an enumeration method according to the time sequence of the critical pathThe total number of groups is calculated by the formula 2n-1; the CPU cores are grouped out of order with the number of groups being 2n-1。
Because a data set to be processed by the radar signal processing system can be abstracted into a data cube and consists of three dimensions of a channel, a pulse and a distance unit (cpr), the parallelism of a radar signal processing software component mainly refers to the size of the data set input in the multi-core system which can be divided and processed, the data dimensions processed by different software components are different, for example, the smallest dimension which can be processed by an fft component usually is the pulse dimension, and a module solving component can reach the dimension of the distance unit. For the software component of the sliding window calculation, the following formula can be used for calculation. Taking the MTI component as an example, a data set with dimension (6, 10, 1024) is input, and 3-pulse sliding window calculation is performed in a CPU with 20 cores, wherein MTI requires that processable data with dimension (p, r) and the maximum parallelism is calculated as follows:
Ca＝ceil(6×(10-3+1)/ceil(6×(10-3+1)/20))
Pb＝floor(20/Ca)
maximum parallelism: p-Ca × Pb-16
Graph partitioning and core group partitioning when a round robin processing architecture is employed. And calculating the completion time T of the whole workflow according to the obtained maximum parallelism of each assembly, and further obtaining the number Cn of the circulating CPUs, wherein the formula is as follows: cn is T/PRT, where T is the completion time of the entire workflow and PRT is the radar repetition period. The resulting number of CPUs k is then compared with the maximum number max given by the system: if k < max, then graph partitioning is not required; if k > max, graph partitioning is required. And (4) grouping the sequenced software components in the graph, wherein the sequencing of the software components between groups and in the groups is consistent with the sequence on the key path.
Taking the critical path in FIG. 3(b) as an example, the graph can be divided into 2 with sequence n1 group of 31 divisions, the grouping is as follows (123456 is the sequence number of each component in the figure):
when the materials are divided into 1 group: [123456]
When the materials are divided into 2 groups: [1] [23456 ]; [12] [3456 ]; [123] [456 ]; [1234] [56 ]; [12345] [6 ];
…
when the materials are divided into 6 groups: [1][2][3][4][5][6]
When the round-robin processing architecture is adopted, the CPUs process the same DAG graph, and the core number of each CPU is also grouped, except that the core group is an unordered group, that is, the serial number of the core is not considered, the total number of the groups is, assuming that 6 cores are in total, and the number in [ ] is the number of the distributed cores, then:
when the materials are divided into 1 group: [6]
when the materials are divided into 2 groups: [1] [5 ]; [2] [4 ]; [3] [3 ];
…
(7) the sub-map component group is matched with the core group: the processing time of each group of graph components at different core numbers is calculated and stored.
The steps of the sub-map component group and the core group are as follows: 1) calculating the critical path length: calculating the processing time of the node on the critical path in the single core, wherein the processing time is the longest time L of the systemtThe combination of the screening map group and the kernel group is performed with this time as a reference. 2) Selecting a component group: selecting a group in turn in the group of map components, e.g. group when divided into 1 group [123456]. 3) Selecting a corresponding core group: the core groups if the component groups are divided into 1 group are also only one, namely [6]]. 4) Compute the time of the component under the current core group: computing the processing time P of a component 1, 2, 3, 4, 5, 6 when allocating 6 corest. 5) Screening and combining: comparison LtAnd PtAnd if so, recording and otherwise discarding.
(8) Selecting an optimal deployment scheme: and screening the results of the adaptation of the sub-map component group and the core group according to the input software and hardware constraint conditions, and selecting the combination with the minimum delay period and hardware scale for output.
(9) A pushing deployment scheme: and pushing the best combination after the adaptation to each CPU of the round robin architecture system.
Acquiring a component sequence and CPU information, grouping the components in the graph, grouping CPU cores, calculating the processing time of the components in the graph group in each core group, selecting the graph group and the cores meeting the strategy, and outputting.
Claims (7)
1. A deployment method of radar signal processing software components is characterized by comprising the following steps: the method comprises the following steps:
(1) information acquisition: the system receives workflow diagram information input by a front end, wherein the workflow diagram information comprises description information and software and hardware constraint conditions;
(2) and (3) analysis of the graph: analyzing and quantizing the description information in the step (1) to obtain the serial number, the dependency relationship and the data set format information of the components in the graph, and storing the serial number, the dependency relationship and the data set format information of the components in the graph, which are obtained by analyzing and quantizing, by an adjacency list method;
(3) obtaining detailed graph component information: obtaining data dimension information which can be processed by the component and data passing rates of the component under different core numbers from a component database;
(4) and (3) component sequencing: sorting the components in the graph according to an in-degree sequence;
(5) and (3) calculating a critical path: obtaining the critical path length of the graph and a master-slave relationship table of each component by calculating the earliest EFT time and the latest LFT time of the components in each graph;
(6) the components in the figure and the CPU core are grouped as follows: according to the time sequence of the critical path, orderly grouping the components in the graph by adopting an enumeration method, wherein the component group number calculation formula is 2n-1; the CPU cores are grouped out of order, the number of the CPU groups is 2n-1Wherein n is the number of components;
(7) the sub-map component group is matched with the core group: calculating and storing the processing time of each group of the components in the graph in different core numbers to obtain the result of the adaptation of the group of the component of the graph and the core group;
(8) selecting an optimal deployment scheme: screening the results of the adaptation of the sub-image component group and the core group according to the input software and hardware constraint conditions, and selecting the combination with the minimum delay period and hardware scale for output to obtain the optimal combination;
(9) a pushing deployment scheme: and (4) pushing the best combination after the adaptation in the step (8) to each CPU of the round-robin architecture system.
2. The method of deploying a radar signal processing software component of claim 1, wherein: and (2) modeling the workflow diagram in the step (1) by adopting a DAG directed acyclic graph model.
3. The method of deploying a radar signal processing software component of claim 2, wherein: the adjacency list method in step (2) specifically includes the following steps:
1) establishing an adjacent task pool: the adjacent task pool comprises an adjacent matrix and a dependency relationship table, and the dependency relationship of the components is represented by adopting a method of a two-way linked list;
2) resolving the character string describing the dependency relationship: analyzing the character string to obtain the sequence numbers of the pre-adjacent point and the subsequent adjacent point in the graph and the data set transmitted between the pre-adjacent point and the subsequent adjacent point, and adding the pre-adjacent point and the subsequent adjacent point into a double linked list;
3) obtaining an adjacency matrix: and sequentially calculating the doubly linked list representing the dependency relationship to obtain the adjacency relationship among the components in the graph, namely the number of connecting lines between each component and other components.
4. The method of deploying a radar signal processing software component of claim 3, wherein: the critical path calculation in the step (5) specifically includes the following steps:
1) and (3) calculating the entrance degree: calculating the entrance degree of each adjacent point component, and obtaining the entrance degree of each component by inquiring the adjacent matrix;
2) sorting according to the degree of income: subtracting 1 from the degree of approach in sequence, recording adjacent point components with zero degree of approach after subtracting 1 each time, and finally obtaining a component sequence ordered according to the degree of approach;
3) calculating the earliest time EFT;
4) calculating the latest occurrence time LFT;
5) obtaining a master node and a slave node: if the earliest time EFT and the latest occurrence time LFT of a certain adjacent point component are equal, the component is represented as a point on a critical path and is a master node, otherwise, the component is a slave node;
6) obtaining a critical path: and merging the slave nodes into the critical path to obtain a final critical path.
5. The method of deploying a radar signal processing software component of claim 4, wherein: the components and CPU core groups in the graph in the step (6) are processed by adopting a round-robin processing architecture method, and the specific steps are as follows:
1) calculating the completion time T of the whole workflow according to the obtained maximum parallelism of the components in each graph, and further obtaining the number Cn of the circulating CPUs, wherein the formula Cn is T/PRT, T is the completion time of the whole workflow, and PRT is a radar repetition period;
2) the resulting CPU number Cn is compared with the system-given maximum number max: if Cn is less than max, the graph division is not needed, if Cn is more than max, the graph division is needed;
3) and (4) grouping the sequenced software components in the graph, wherein the sequencing of the software components between groups and in the groups is consistent with the sequence on the key path.
6. The method of deploying a radar signal processing software component of claim 5, wherein: the parallelism refers to the size of a data set input in the multi-core system which can be divided and processed, and the data dimension processed by different software components is different.
7. The method of deploying a radar signal processing software component of claim 1, wherein: the specific steps of the step (7) are as follows:
1) calculating the critical path length: calculating the processing time of the node on the critical path in the single core, wherein the processing time is the longest time L of the systemtScreening the combination of the graph group and the kernel group by taking the time as a reference;
2) selecting a component group;
3) selecting a corresponding core group;
4) compute the time of the component under the current core group: computing group processing time P in assigning corest；
5) Screening and combining: comparison LtAnd Pt，LtGreater than PtRecording is done, otherwise it is discarded.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810563967.9A CN108776589B (en) | 2018-06-04 | 2018-06-04 | Deployment method of radar signal processing software component |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810563967.9A CN108776589B (en) | 2018-06-04 | 2018-06-04 | Deployment method of radar signal processing software component |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108776589A CN108776589A (en) | 2018-11-09 |
| CN108776589B true CN108776589B (en) | 2021-07-27 |
Family
ID=64025683
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810563967.9A Active CN108776589B (en) | 2018-06-04 | 2018-06-04 | Deployment method of radar signal processing software component |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108776589B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109613505A (en) * | 2018-12-13 | 2019-04-12 | 航天南湖电子信息技术股份有限公司 | A kind of device and its suppressing method inhibiting dual clutter |
| CN112650505A (en) * | 2020-12-21 | 2021-04-13 | 四川长虹电器股份有限公司 | Web project generation, deployment and configuration method |
| CN113626116B (en) * | 2021-07-20 | 2023-12-15 | 中国电子科技集团公司电子科学研究院 | Intelligent learning system and data analysis method |
| CN113867736B (en) * | 2021-09-03 | 2022-11-08 | 中国人民解放军军事科学院战争研究院 | Deployment scheme generation method and device |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102404381A (en) * | 2011-09-02 | 2012-04-04 | 西安交通大学 | Software deployment system and deployment method based on workflow in cloud computing environment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7069553B2 (en) * | 2003-03-03 | 2006-06-27 | Computer Associates Think, Inc. | Universal deployment tool |
-
2018
- 2018-06-04 CN CN201810563967.9A patent/CN108776589B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102404381A (en) * | 2011-09-02 | 2012-04-04 | 西安交通大学 | Software deployment system and deployment method based on workflow in cloud computing environment |
Non-Patent Citations (1)
| Title |
|---|
| 云计算环境下业务自动化部署的设计与实现;汪源;《软件》;20141230;第35卷(第9期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108776589A (en) | 2018-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108776589B (en) | Deployment method of radar signal processing software component | |
| AU2019284011B2 (en) | Data processing method and related products | |
| CN110515739B (en) | Deep learning neural network model load calculation method, device, equipment and medium | |
| CN110737529B (en) | Short-time multi-variable-size data job cluster scheduling adaptive configuration method | |
| US20150033237A1 (en) | Utility-optimized scheduling of time-sensitive tasks in a resource-constrained environment | |
| Cariño et al. | Dynamic load balancing with adaptive factoring methods in scientific applications | |
| CN107908536B (en) | Performance evaluation method and system for GPU application in CPU-GPU heterogeneous environment | |
| CN113037877B (en) | Optimization method for time-space data and resource scheduling under cloud edge architecture | |
| US8869149B2 (en) | Concurrency identification for processing of multistage workflows | |
| US20190377655A1 (en) | Two-stage distributed estimation system | |
| Ardagna et al. | Predicting the performance of big data applications on the cloud | |
| Al Jawarneh et al. | Spatial-aware approximate big data stream processing | |
| US20210390405A1 (en) | Microservice-based training systems in heterogeneous graphic processor unit (gpu) cluster and operating method thereof | |
| CA2631255A1 (en) | Scalable scheduling of tasks in heterogeneous systems | |
| CN112905317A (en) | Task scheduling method and system under rapid reconfigurable signal processing heterogeneous platform | |
| CN109977497B (en) | Parallel method for accelerating general radar signal processing flow based on CPU | |
| CN109302723B (en) | Multi-node real-time radio monitoring control method based on Internet | |
| CN116700920A (en) | Cloud primary hybrid deployment cluster resource scheduling method and device | |
| Koch et al. | SMiPE: estimating the progress of recurring iterative distributed dataflows | |
| CN113722141B (en) | Method and device for determining delay reason of data task, electronic equipment and medium | |
| CN111966597B (en) | Test data generation method and device | |
| CN114661571A (en) | Model evaluation method, model evaluation device, electronic equipment and storage medium | |
| Begin et al. | Predicting the system performance by combining calibrated performance models of its components: a preliminary study | |
| CN110058812B (en) | Scientific workflow data placement method in cloud environment | |
| CN116719584B (en) | Data processing method, apparatus, computer, storage medium, and program product |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |